{"entries":[{"key":"cn.lambda","quantity":"Initial abstraction ratio","value":0.2,"unit":"dimensionless","class":"C1","confidence":"Verified","worth":"Switching to 0.05 with the converted CN roughly doubles runoff at P = 1 in and changes it about 4 percent at P = 3 in.","displaced_by":"Not displaceable by evidence: no measurement of a site changes the ratio a table was fitted at. A person may state the other published convention, 0.05, and the retention is then converted with it through cn.lambda005_coefficient and cn.lambda005_exponent. A ratio with no published conversion is refused under CN-MTH-006 rather than computed.","source":"NEH Part 630 (2004) Chapter 10","reasoning":"No site measurement changes the ratio a published table was fitted at, so evidence cannot displace this value. What a person may do is choose the other published convention, 0.05, and the retention is then converted to it. The 4 percent figure above compares the two conventions after that conversion.","overridable":false},{"key":"cn.lambda005_coefficient","quantity":"Coefficient of the 0.2 to 0.05 retention conversion","value":1.33,"unit":"dimensionless","class":"C4","confidence":"Moderate","worth":"The whole of the difference between a defensible 0.05 answer and an indefensible one. On woods in fair condition on group B, CN 60 at P = 3.16 in, converting takes S from 6.6667 to 11.7855 in and the depth from 0.8416 to 0.4603 in: the unconverted figure is 82.8 percent high against the converted one, and 35.3 percent high over the seven segment site this was found on.","displaced_by":"A fit published against a stated range of retentions with its scatter reported, or a calibration of this relation against measured event data. The relation is a regression and a better regression replaces it.","source":"Woodward, Hawkins, Jiang, Hjelmfelt, Van Mullem & Quan, Runoff Curve Number Method: Examination of the Initial Abstraction Ratio, ASCE World Water and Environmental Resources Congress (2003) the S(0.05) = 1.33 S(0.2) ** 1.15 retention conversion; page not established by the transcriber","reasoning":"","overridable":true},{"key":"cn.lambda005_exponent","quantity":"Exponent of the 0.2 to 0.05 retention conversion","value":1.15,"unit":"dimensionless","class":"C4","confidence":"Moderate","worth":"It is what makes the conversion more than a scale factor, and it is what puts a bound on the domain: because it exceeds one, the relation returns a retention smaller than the one it converts below S = 0.1494 in, which is CN 98.53, and the engine refuses there under CN-MTH-006 rather than extrapolating.","displaced_by":"The same evidence that would displace the coefficient. The two are one fit and neither is separately measurable.","source":"Woodward, Hawkins, Jiang, Hjelmfelt, Van Mullem & Quan, Runoff Curve Number Method: Examination of the Initial Abstraction Ratio, ASCE World Water and Environmental Resources Congress (2003) the S(0.05) = 1.33 S(0.2) ** 1.15 retention conversion; page not established by the transcriber","reasoning":"","overridable":true},{"key":"cn.impervious","quantity":"Curve number, connected impervious","value":98.0,"unit":"CN","class":"C2","confidence":"Verified","worth":"Against direct depression storage subtraction at Sd = 0.05 in, CN 98 understates impervious runoff by 34.6 percent at P = 0.4 in, 25.5 percent at 0.6 in and 16.7 percent at 1.0 in.","displaced_by":"Not displaceable as a table value; the engine routes impervious segments to direct subtraction instead.","source":"USDA NRCS TR-55 (1986, 2nd ed.) Chapter 2","reasoning":"","overridable":true},{"key":"condition.default","quantity":"Hydrologic condition","value":1.0,"unit":"class index","class":"C5","confidence":"Low","worth":"5 to 12 CN units. One of the two largest silent levers in the method.","displaced_by":"A site visit, current aerial imagery, or a measured ground cover fraction against the NEH Chapter 9 mapping.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. Fair is the mid-range assignment and avoids systematic bias. Recorded as a judgment, not a citation, because a user told this is 'the published default' will not think to change it, and a user told it is a mid-range assumption worth 5 to 12 CN units will.","overridable":true},{"key":"hsg.dual_resolution","quantity":"Dual hydrologic soil group resolution","value":1.0,"unit":"policy flag","class":"C5","confidence":"Low","worth":"Enormous. Open space in good condition on A/D is CN 39 drained against CN 80 undrained. At P = 3 in, pasture on B against D is 0.365 in against 1.250 in, a factor of 3.42.","displaced_by":"User attestation that artificial drainage is present, or a drain tile record.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. The conservative choice, matching the physical condition absent evidence of drainage.","overridable":true},{"key":"impervious.connectivity","quantity":"Impervious connectivity","value":0.0,"unit":"dimensionless","class":"C5","confidence":"Low","worth":"Bounded at 10.2 CN units. Figure 2-4 adjusts the composite as CN = CNp + (Pimp/100)(98 - CNp)(1 - R/2), so the whole span from fully connected to fully unconnected is 0.5 (Pimp/100)(98 - CNp), largest at the 30 percent total impervious ceiling the figure is valid to and at the lowest pervious curve number TR-55 publishes, which is 30: 0.15 x 68 = 10.2 units. At a pervious CN of 61 the same bound is 5.55 units, and at 80 it is 2.7.","displaced_by":"A delineated drainage path showing impervious area discharging to pervious ground. Automated detection from LiDAR routing is an open gap.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. The conservative assumption.","overridable":true},{"key":"impervious.depression_storage","quantity":"Depression storage, impervious","value":0.05,"unit":"in","class":"C4","confidence":"Moderate","worth":"On a 1,500 sq ft roof at P = 0.6 in, moving 0.05 to 0.10 in changes inflow by 9 percent.","displaced_by":"A jurisdictionally published value, which supersedes.","source":"EPA/600/R-15/162A (2016) Section 3.8.7, quoting ASCE (1992)","reasoning":"","overridable":true},{"key":"ga.sandy_loam.K","quantity":"Green-Ampt conductivity, sandy loam","value":0.4291,"unit":"in/hr","class":"C2","confidence":"Verified","worth":"K is the dominant Green-Ampt sensitivity. On 1 in in 20 minutes, dropping to the compacted value raises runoff from 0.333 to 0.550 in, an increase of 65 percent.","displaced_by":"A field infiltration test by a method the governing jurisdiction accepts.","source":"Rawls, Brakensiek & Miller (1983) J. Hydraul. Eng. 109(1):62, Table 2 (1983) via EPA/600/R-15/162A Table 4-7","reasoning":"Primary value 1.09 cm/hr. Conductivity column sourced to Rawls, Brakensiek & Saxton (1981), not the 1983 paper.","overridable":true},{"key":"ga.sandy_loam.K_compacted","quantity":"Green-Ampt conductivity, sandy loam, compacted by construction","value":0.2,"unit":"in/hr","class":"C5","confidence":"Low","worth":"On 1 in in 20 minutes it gives 0.550 in of runoff against 0.334 in on the undisturbed value, 64.7 percent more, and it is the difference between 0.218 in and nothing at all on 1 in in an hour. It is also the least conservative value consistent with its own justification: 0.20 in/hr is a 53.4 percent reduction from 0.4291, where the 70 to 99 percent reduction the compaction literature reports would put K at 0.129 to 0.0043 in/hr and runoff on the same storm at 0.642 to 0.938 in, 16.9 to 70.7 percent above what this value gives. A departure of 0.05 in/hr from it is worth about 11 percent of the runoff depth on that storm.","displaced_by":"A field infiltration test on the compacted surface, taken after construction rather than before it. There is nothing to weigh this against: any measurement at all displaces it.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. This value was a bare constant in the Green-Ampt calculation, justified in a comment as the one-level reduction that is the encodable published precedent, and it matches no row of the Rawls table: sandy loam is 0.4291 in/hr, loam is 0.1339 and silt loam is 0.2559, so it is not a texture class either. It is a judgment, and recording it as a C2 table value by proximity to ga.sandy_loam.K, which is what leaving it in the code did, is the one failure this registry exists to prevent. The magnitude is carried over unchanged, because changing it is a calibration decision and this is a reclassification.","overridable":true},{"key":"ga.sandy_loam.psi","quantity":"Green-Ampt suction head, sandy loam","value":4.3346,"unit":"in","class":"C2","confidence":"Verified","worth":"Secondary to K. Affects ponding time more than volume on storms above about an hour.","displaced_by":"Rarely measured directly. Displaced only alongside a full soil characterization.","source":"Rawls, Brakensiek & Miller (1983) J. Hydraul. Eng. 109(1):62, Table 2 (1983) via EPA/600/R-15/162A Table 4-7","reasoning":"Primary value 11.01 cm. The two EPA manuals disagree on suction head for nine of eleven texture classes; the Reference Manual matches the primary source and the User's Manual does not.","overridable":true},{"key":"ga.moisture_deficit","quantity":"Initial moisture deficit","value":0.3,"unit":"dimensionless","class":"C5","confidence":"Low","worth":"Enters the wetting front term directly; a shift of 0.1 moves ponding time materially.","displaced_by":"A measured antecedent moisture content.","source":null,"reasoning":"NO EXTERNAL SOURCE. A stated assumption pending calibration.","overridable":true},{"key":"method.band_lower","quantity":"Curve number suppression band, lower","value":2.0,"unit":"multiples of Ia","class":"C6","confidence":"Low","worth":"A five CN unit uncertainty is worth a factor of 18.1 at 1.5 Ia, 4.4 at 2 Ia, 2.3 at 3 Ia and 1.4 at 9 Ia.","displaced_by":"Expert panel calibration.","source":null,"reasoning":"REASONED, NOT MEASURED. Proposed default awaiting calibration.","overridable":false},{"key":"method.band_upper","quantity":"Curve number caution band, upper","value":5.0,"unit":"multiples of Ia","class":"C6","confidence":"Low","worth":"Above this the method is in the regime it was fitted for.","displaced_by":"Expert panel calibration.","source":null,"reasoning":"REASONED, NOT MEASURED.","overridable":false},{"key":"method.gate_per_surface","quantity":"Depth band judged on each area's own curve number","value":1.0,"unit":"switch (1 = each area on its own curve number, 0 = the site's composite)","class":"C6","confidence":"Low","worth":"On a half paved 1 ac lot at P = 1.0 in the per area gate gives 0.79 in on the pavement (CN 98), 0 on the lawn (CN 61) and 0.40 in area weighted; judged on the composite CN of 79.5 the same site was refused. With 1 the 95 percent impervious switch to direct runoff (method.impervious_threshold) does not apply: impervious cover stays at CN 98.","displaced_by":"A jurisdiction rule that judges the site on its composite curve number.","source":null,"reasoning":"TR-55 Chapter 2: the runoff equation and its initial abstraction are per cover; the depth bands are written in multiples of each surface's own Ia.","overridable":false},{"key":"method.surface_r1_share","quantity":"Share of the site's runoff at which an area between Ia and 2 Ia withholds the headline","value":0.1,"unit":"fraction of the site's runoff volume","class":"C6","confidence":"Low","worth":"Judged area by area (method.gate_per_surface), an area whose storm is between Ia and 2 Ia withholds the site's curve number figure only where its runoff is at least this share of the site's. Below it the figure stands with a sensitivity band. 0 withholds on any such area.","displaced_by":"A jurisdiction rule, or a site study of the minor surface's runoff.","source":null,"reasoning":"REASONED, NOT MEASURED. The same 10 percent of runoff the product uses to flag a large change.","overridable":false},{"key":"method.impervious_threshold","quantity":"Imperviousness at which direct subtraction governs","value":0.95,"unit":"fraction","class":"C6","confidence":"Low","worth":"Not used while each area is judged on its own curve number (method.gate_per_surface at 1, the default): impervious cover then stays at CN 98. With that set to 0, it decides which method governs an area, and the two disagree by up to 35 percent at small depths.","displaced_by":"Calibration against benchmark sites.","source":null,"reasoning":"REASONED, NOT MEASURED.","overridable":false},{"key":"practice.min_area_acres","quantity":"Practice drainage area ceiling for curve number primacy","value":0.5,"unit":"acre","class":"C6","confidence":"Low","worth":"Below it the curve number is not primary and the practice chain governs.","displaced_by":"Not evidence-displaceable.","source":null,"reasoning":"REASONED, NOT MEASURED.","overridable":false},{"key":"peak.min_tc_hours","quantity":"Minimum time of concentration used for a peak discharge","value":0.1,"unit":"hr","class":"C6","confidence":"Low","worth":"A computed Tc below it is raised to it before the unit hydrograph is drawn, and the raise is stated on the result, for example 'Tc raised from 3.2 minutes to the TR-55 minimum of 6 minutes'. This is most small sites. A longer Tc gives a later, lower peak, so on those sites the peak is lower than one drawn at the shorter computed Tc. The volume does not depend on Tc.","displaced_by":"Not evidence-displaceable.","source":null,"reasoning":"TR-55 applies its peak discharge procedures with a Tc of 0.1 hr at least, and 6 minutes is the minimum in common use on small sites. Applied as a minimum rather than as a refusal. Some manuals set 5 or 10 minutes. A site cannot change it yet, because nothing would carry a site's own minimum into the peak computation, and a recorded departure the number did not honor would be a false entry in the report.","overridable":false},{"key":"method.small_watershed_limit_acres","quantity":"Drainage area above which the site is not presented as a single unsubdivided unit","value":16000.0,"unit":"ac","class":"C6","confidence":"Low","worth":"Whether one curve number over the whole boundary is offered as a design value or whether the product operates in subdivided mode with a mandatory applicability warning. 16,000 ac is 25 square miles exactly, at 640 ac to the square mile. Nothing about the runoff depth relation changes at this area: what changes is that above it the composite bias of a single area weighted curve number over a heterogeneous catchment is no longer bounded by anything this engine computes, and the peak discharge and hydrograph procedures that do have stated bounds are past them.","displaced_by":"The [VERIFY] that the CurveNumber method notes Section 11.5(a) leaves open: which limit attaches to which procedure. The 2,000 acre figure and WinTR-55's 25 square miles are different constraints on different procedures, and the runoff depth relation itself has no stated upper area limit at all. A resolution of that question replaces this single number with three, one per procedure.","source":null,"reasoning":"REASONED, NOT MEASURED. Portal policy per the CurveNumber method notes Section 11.5(a), set at the larger of the two cited figures pending the verification that section marks open. The larger was chosen because the product does not refuse above it, it subdivides and warns, and choosing the smaller would have warned on sites WinTR-55 itself computes without comment.","overridable":false},{"key":"method.derivation_refusal_pixels","quantity":"Source pixel count below which no curve number is derived automatically from a dataset unless waived","value":3.0,"unit":"pixels","class":"C6","confidence":"Low","worth":"Whether a derived figure is withheld (waivable, CN-SEG-006) and the engineer pointed to the aerial photograph classification and manual surface entry. Three pixels is 2,700 m2 (0.6672 ac) on the 30 m annual land cover product, 300 m2 (0.0741 ac) at 10 m and 3 m2 at 1 m, a span of 900 to 1 across products describing the same ground. A quarter acre lot is 1,011.7 m2, n = 1.124 on the 30 m product (refused unless waived) and n = 10.12 at 10 m (neither refused nor warned). Below 10 pixels a site is warned about (the setting derive_warn_pixels).","displaced_by":"Owner decision D-08 (Sep 24, 2026): one small-site rule for the derivation and the assistant, the settings derive_min_pixels (3) and derive_warn_pixels (10), which take precedence over this entry. Before D-08 the assistant used n = 8 from the CurveNumber validation plan Section 3.5.","source":null,"reasoning":"REASONED, NOT MEASURED. The n = 3 refusal and n = 10 warning of the round 1 review default, confirmed for both the derivation and the assistant by the owner (D-08). method.derivation_refusal_acres is this number carried onto the 30 m product for a report that records no pixel count. The rule's second term in the validation plan, the edge fraction, is not derivable from what this product records.","overridable":false},{"key":"method.derivation_refusal_acres","quantity":"Area of interest below which no curve number is derived automatically from the 30 m land cover product unless waived","value":0.67,"unit":"ac","class":"C6","confidence":"Low","worth":"The small-site refusal on one dataset: it is method.derivation_refusal_pixels carried onto the 30 m annual land cover product. Three pixels of 900 m2 is 2,700 m2, which is 0.6672 ac, rounded here to 0.67 ac. Below it the engineer is pointed to the aerial photograph classification and manual surface entry.","displaced_by":"Nothing about this dataset; the figure is arithmetic on its pixel size once the count is chosen. What displaces the count is method.derivation_refusal_pixels and the setting derive_min_pixels (owner decision D-08).","source":null,"reasoning":"REASONED, NOT MEASURED. An acreage is the wrong shape for this rule and this entry is the fallback rather than the rule: method.derivation_refusal_pixels holds the count and is what a detector tests where the report records the pixel count. 3 x 900 m2, divided by 4046.8564224 m2 to the acre, is 0.6672 ac, rounded to 0.67.","overridable":false},{"key":"crosswalk.nlcd_21","quantity":"Cover type, NLCD 21 developed open space","value":1.0,"unit":"cover mapping","class":"C5","confidence":"Moderate","worth":"Against using the TR-55 residential 1 acre lot row as the pervious cover, which already contains 20 percent impervious, group B composite is 65.0 against 71.3 at the measured 10.9 percent impervious, and 0.507 in against 0.773 in, 52 percent more runoff. The family choice itself is nearly free, because the open space and pasture rows are identical digit for digit; the condition axis is not, at 69 against 61 on group B and 0.670 in against 0.365 in.","displaced_by":"Site imagery or a site visit establishing the condition of the pervious cover, or a high resolution land cover product that separates turf from tree canopy.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. Class 21 is vegetation on lots under 20 percent impervious: parks, golf courses, large lot residential, roadside verge. TR-55 open space, described as lawns, parks, golf courses and cemeteries, is the row that names that vegetation. Maps to open_space with the condition left to condition.default, which deliberately does not follow TR-55's own assumption that urban pervious area is in good condition.","overridable":true},{"key":"crosswalk.nlcd_22","quantity":"Cover type, NLCD 22 developed low intensity","value":1.0,"unit":"cover mapping","class":"C5","confidence":"Moderate","worth":"9.2 CN units on group B against the double counted alternative. The residential 1/4 acre row is exactly 0.38 x 98 + 0.62 x 61 = 75.06, so it already holds its impervious share; using it as the pervious cover and adding the measured 34.6 percent impervious as its own segment gives 83.0 where the correct composite is 73.8, and 1.444 in against 0.898 in, 61 percent high. Both sides are on TR-55's own good condition assumption for urban pervious area, so the figure isolates the double counting from the condition choice; under the defaulted fair condition it is 3.9 CN units and 21.5 percent. Preferring the class midpoint of 38 percent to the measured 34.6 is worth a further 1.3 CN units and 7.3 percent.","displaced_by":"A high resolution land cover product that resolves roofs and driveways, or a site plan giving the actual lot coverage.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. Maps to open_space for the PERVIOUS remainder only, with the measured impervious fraction carried as its own segment. The composite residential rows are not used at all, because they are a class midpoint for imperviousness with a cover assumption stapled to it.","overridable":true},{"key":"crosswalk.nlcd_23","quantity":"Cover type, NLCD 23 developed medium intensity","value":1.0,"unit":"cover mapping","class":"C5","confidence":"Moderate","worth":"8.1 CN units on group B against reading the 1/8 acre lot row as the pervious cover and adding the measured 66.4 percent impervious on top: 93.6 against 85.6, and 2.314 in against 1.630 in, 42 percent high. Class midpoint against measured imperviousness is small on this class, 0.55 CN units and 2.5 percent at the defaulted fair condition, because 64.5 and the measured 66.4 nearly agree.","displaced_by":"A high resolution land cover product, or a site plan. The pervious condition is displaced by imagery showing whether the remnant is established turf or compacted fill.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. Maps to open_space for the pervious remainder. At this intensity the remnant is a third of the area and is more often compacted than established, which argues for Poor rather than the defaulted Fair, at 79 against 69 on group B.","overridable":true},{"key":"crosswalk.nlcd_24","quantity":"Cover type, NLCD 24 developed high intensity","value":1.0,"unit":"cover mapping","class":"C5","confidence":"Low","worth":"The cover row is nearly irrelevant on this class and the impervious measurement is nearly everything. Against the commercial row used as the pervious cover with the measured 83.1 percent added on top, 97.0 against 91.7 and 2.657 in against 2.138 in, 24 percent high. Against the class definition midpoint of 90 percent impervious, 94.3 against 91.7, which is 2.55 CN units, 11.3 percent of the runoff depth and 27.6 percent at P = 1.2 in.","displaced_by":"A high resolution land cover product, or a site plan. This is the class where a 1 meter product changes the answer most.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. Maps to open_space for the sixth of the area that is pervious, which is street tree pits, verge and lot edges rather than lawn. Confidence is LOW on the cover row and that matters little, which is the finding: on this class the measured impervious fraction carries the answer.","overridable":true},{"key":"crosswalk.nlcd_31","quantity":"Cover type, NLCD 31 barren land","value":1.0,"unit":"cover mapping","class":"C5","confidence":"Low","worth":"17 CN units on group B against open space in fair condition, 86 against 69, which is 1.663 in against 0.670 in, a factor of 2.5. 12 units against treating barren ground as impervious, 86 against 98 and 1.663 in against 2.768 in.","displaced_by":"A site visit or imagery saying which barren this is. Bedrock outcrop and a paved quarry floor belong at the impervious end; coarse dune sand infiltrates better than any row here.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. TR-55 has no barren row. Maps to newly_graded, the developing urban areas row for pervious ground with no vegetation, which is the only published row for bare ground that still infiltrates. That row has no condition axis, so condition.default does not apply to this class.","overridable":true},{"key":"crosswalk.nlcd_forest","quantity":"Cover type, NLCD 41, 42 and 43 forest","value":1.0,"unit":"cover mapping","class":"C5","confidence":"Moderate","worth":"The condition axis carries this mapping and the family does not. Woods Poor 66 against Good 55 on group B is 0.545 in against 0.195 in, a factor of 2.8; on group A it is 45 against 30, which is 0.024 in against exactly zero. Distinguishing deciduous from evergreen is worth nothing in this method, because there is one woods row.","displaced_by":"Litter depth, canopy cover and evidence of grazing observed on site, read against the NEH Chapter 9 woods condition descriptions.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. Classes 41, 42 and 43 all map to woods with the condition left to condition.default. What is lost is real and unrepresentable: evergreen canopy intercepts through the winter and deciduous litter changes with the season, and the method has no term for either.","overridable":true},{"key":"crosswalk.nlcd_52","quantity":"Cover type, NLCD 52 shrub and scrub","value":1.0,"unit":"cover mapping","class":"C5","confidence":"Low","worth":"16 CN units on group B against the desert shrub row that is right for the same class in the arid west: brush Fair 56 against desert shrub Fair 72, which is 0.220 in against 0.808 in, a factor of 3.7. On group A it is 35 against 55, exactly zero against 0.195 in.","displaced_by":"A user naming the vegetation displaces it directly. The site's location already moves it: since round one of the 2026 review the crosswalk is given the site's coordinates, and west of the 100th meridian (outside the humid Pacific coast north of 40 degrees) class 52 maps to desert shrub, Table 2-2d, stated as an assumption on the result.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS, and this is the weakest row in the crosswalk. One class code covers eastern successional brush and western sagebrush and desert shrub, which TR-55 puts in different tables. Maps to brush, the humid region row, east of the 100th meridian, and to desert shrub in fair condition west of it; the meridian is a coordinate rule standing in for a climate map, not a measurement of the cover.","overridable":true},{"key":"crosswalk.nlcd_71","quantity":"Cover type, NLCD 71 grassland and herbaceous","value":1.0,"unit":"cover mapping","class":"C5","confidence":"Low","worth":"21 CN units on group B and 38 on group A across the range this one class spans. Pasture in poor condition is 79 on group B against meadow at 58, which is 1.189 in against 0.274 in, a factor of 4.3; on group A it is 68 against 30, 0.627 in against exactly zero.","displaced_by":"Imagery or a site visit distinguishing mown, grazed and unmanaged grass, which is the distinction TR-55 prices and NLCD does not record.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. This is the row the land cover reader names as the reason the crosswalk is a judgment rather than a lookup: the same code covers a mown lawn, an unmanaged old field and a hayfield. Maps to pasture with the condition left to condition.default, which is the middle of that range rather than an assertion about the site.","overridable":true},{"key":"crosswalk.nlcd_81","quantity":"Cover type, NLCD 81 pasture and hay","value":1.0,"unit":"cover mapping","class":"C5","confidence":"Moderate","worth":"11 CN units on group B against the meadow row: pasture Fair 69 against meadow 58, which is 0.670 in against 0.274 in, a factor of 2.4.","displaced_by":"Whether the field is grazed or cut for hay, which the operator knows and no raster records.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS, though this is the closest thing to a name match in the crosswalk. NLCD merges grazed pasture with hay and TR-55 separates them: meadow is protected from grazing and mowed, and is published as a single Good condition row. Maps to pasture with the condition left to condition.default, because choosing meadow would assert a good condition as well as a cover type.","overridable":true},{"key":"crosswalk.nlcd_82","quantity":"Cover type, NLCD 82 cultivated crops","value":1.0,"unit":"cover mapping","class":"C5","confidence":"Low","worth":"3 CN units on group B and 5 on group A between the two published conditions on the straight row row: 81 against 78 is 1.314 in against 1.129 in, 16.4 percent, and 72 against 67 on group A is 0.808 in against 0.585 in, 38.1 percent. The treatment axis, straight row against contoured and terraced, is worth more than the condition axis and NLCD does not record it at all.","displaced_by":"The crop, the tillage and the residue management, which the county soil and water conservation district holds, or a contour and terrace layer.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. Class 82 is one code for every row crop, small grain and close seeded legume, and TR-55 prices crop, treatment, condition and residue separately. Maps to row_crops_sr_poor. Straight row is pinned because it is the treatment that assumes no practice, and Poor is pinned because TR-55 publishes only Poor and Good for cultivated crops, so condition.default has no Fair row to land on here; Poor is the conservative of the two.","overridable":true},{"key":"crosswalk.water_as_storage","quantity":"Open water is refused rather than given a curve number","value":1.0,"unit":"policy flag","class":"C5","confidence":"Low","worth":"On a site otherwise at CN 69 on group B, a pond over 10 percent of the area treated as CN 98 gives 0.880 in against 0.670 in for the same site with the pond excluded, 31 percent high, and 0.116 in against 0.019 in at P = 1.2 in, a factor of 6.1. The error scales with the pond's share and it lands on exactly the sites that have one.","displaced_by":"A stage, area and discharge relationship for the water body, which makes it a routing element and not a curve number at all.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS for a curve number for open water, and that is the finding rather than a gap in the transcription. TR-55 has no open water row. Assigning 98 or 100 asserts that a pond spills its whole surface rainfall instantly, which is a claim about an outlet nobody has looked at. The crosswalk refuses NLCD class 11 and names what the user must supply instead.","overridable":true},{"key":"crosswalk.impervious_allocation","quantity":"Allocation of the measured impervious fraction across classes","value":1.0,"unit":"policy flag","class":"C5","confidence":"Low","worth":"On a site half medium intensity developed and half forest with a measured 30 percent impervious on group B, allocating uniformly gives composite CN 74.6 and 1.182 in, against 73.2 and 1.131 in if the whole impervious share is taken out of the developed class. That is 1.4 CN units and 4.46 percent of the runoff depth.","displaced_by":"A per class impervious mean, which the same service can produce by masking the fractional impervious raster with the class raster. That is a real improvement and it is not implemented.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. The impervious fraction is measured over the whole area of interest and the crosswalk has no per class breakdown, so the impervious area is removed in equal proportion from every class that has a cover row. That is wrong in a known direction, since impervious cover concentrates in classes 21 to 24 and is close to absent under forest and crops, and it moves pervious area away from the classes whose pervious cover has the higher curve number. Classes with no cover row are left whole instead: the impervious raster reads zero over open water, so taking a share out of a pond would invent impervious cover and lose pervious area at the same time.","overridable":true},{"key":"crosswalk.sliver_fraction","quantity":"Land cover share below which a class is folded into the dominant one","value":0.03,"unit":"fraction","class":"C6","confidence":"Low","worth":"Larger than the word sliver suggests. A class at the threshold folded across the widest pair in this crosswalk, brush in fair condition at CN 56 into newly graded barren at 86 on group B, moves distributed runoff from 0.220 to 0.263 in, 19.7 percent, and 5.3 percent at P = 6 in. On a typical pair, woods fair 60 into open space fair 69, it is 3.0 percent. This is why every suppression is recorded with its class and its area instead of being folded silently.","displaced_by":"A finer land cover product. At 1 meter the same 3 percent is hundreds of pixels and is a feature of the site rather than a raster edge.","source":null,"reasoning":"REASONED, NOT MEASURED. Three percent is one pixel of the thirty that the land cover reader treats as the minimum for a reportable share, so a class below it on a site at that minimum is a single pixel and sits inside the boundary discretization error that module already quantifies. The worth above says the threshold is not free at the low end either, so the choice is between two visible errors and the recorded one was preferred.","overridable":true},{"key":"crosswalk.sliver_pixels","quantity":"Land cover area, in 30 m pixels, below which a class is folded into the dominant one","value":1.0,"unit":"pixels (900 square meters each) of the soil group's area","class":"C6","confidence":"Low","worth":"A class smaller than one land cover pixel on its soil group is below what the 30 m map can resolve, so it is folded into the largest class and listed with its acres. Larger classes keep their own area. The share this is on a soil group of A acres is 0.2224 / A.","displaced_by":"A finer land cover product.","source":null,"reasoning":"REASONED, NOT MEASURED. Areas are clipped to the boundary, so the old 3 percent rule (crosswalk.sliver_fraction, still used where this is 0), which dated from pixel counting, folded classes the map does resolve. One pixel is the resolution of the source.","overridable":true},{"key":"practice.media_porosity","quantity":"Effective porosity, engineered bioretention media","value":0.25,"unit":"dimensionless","class":"C5","confidence":"Low","worth":"The media holds 259.4 cu ft of the reference cell's 1,167.8, which is 22.2 percent of the storage the practice is credited with. Moving 0.25 to 0.20, which is the low end of what manuals credit, removes 51.9 cu ft, or 4.4 percent of the total. The whole range in use, 0.20 to 0.40, spans 207.5 cu ft, 17.8 percent of the storage and four times that step, and it lands on the layer whose porosity nobody measures on the job.","displaced_by":"A porosity or a water holding capacity measured on the mix actually specified, which the media supplier can produce and almost never is asked for.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS for a design storage credit. Total porosity of a sand based media is measurable and is not the quantity wanted here: what the practice can store and then give back is the drainable fraction between saturation and field capacity, which is smaller than total porosity and depends on the mix. 0.25 is the middle of the 0.20 to 0.40 range that manuals credit and is recorded as a judgment.","overridable":true},{"key":"practice.gravel_porosity","quantity":"Effective porosity, clean open graded stone reservoir","value":0.4,"unit":"dimensionless","class":"C4","confidence":"Moderate","worth":"The stone holds 276.6 cu ft of the reference cell's 1,167.8, 23.7 percent of the storage. Moving 0.40 to 0.35 removes 34.6 cu ft, 3.0 percent of the total. Small on a bioretention cell and not small on an infiltration trench or a permeable pavement, where the stone is the only storage there is and the same 0.40 to 0.35 is 12.5 percent of everything the practice holds.","displaced_by":"A void ratio measured on the specified gradation, which is a routine aggregate test, or a compacted in place measurement, which is the honest one because placement closes voids.","source":null,"reasoning":"NOT VERIFIED AGAINST A SOURCE. The void ratio of clean open graded stone is a material property, it is measurable, and 0.40 is the figure in common use for a washed No. 57 type gradation. No published table was checked in setting this value, so it is recorded as unverified rather than given a citation. The value is optimistic in a known direction: it assumes the stone is clean and stays clean, and fines washed in from an unstabilised contributing area reduce it permanently.","overridable":true},{"key":"practice.media_conductivity","quantity":"Design filtration rate through the engineered media","value":2.0,"unit":"in/hr","class":"C5","confidence":"Low","worth":"It sets the drawdown of every practice with an underdrain, because the media and not the pipe is the constraint: on the reference cell with a 4 in underdrain the media passes 115.3 cu ft/hr against the pipe's 2,699 cu ft/hr at full stage, a factor of 23.4. Halving it to 1 in/hr takes the drawdown from 7.5 to 13.0 hours, 73 percent longer, and it is halved by nothing more exotic than three years of sediment.","displaced_by":"A measured infiltration rate on the installed media, which is the acceptance test some jurisdictions already require, or a long term rate from a maintenance record.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS for a single design figure. Specified media commonly tests between 2 and 8 in/hr when new and falls with age, and manuals variously require a minimum, a maximum, or both. 2 in/hr is the low end of the new condition range and is chosen because a design that works only at the new rate stops working, which is the failure this value exists to prevent.","overridable":true},{"key":"practice.infiltration_safety_factor","quantity":"Safety factor applied to the measured native infiltration rate","value":2.0,"unit":"dimensionless","class":"C5","confidence":"Low","worth":"The largest single lever in practice sizing. Because storage and infiltration both scale with the bottom area, the drawdown time does not depend on the footprint at all, and the safety factor sets the maximum depth of water a practice may store: at a measured 1.0 in/hr and a 48 hour limit it is 24 in of water at a factor of 2 and 12 in at a factor of 4. On half an acre of pavement at a 1 in treatment depth, 1,815 cu ft, that is a bottom area of 908 sq ft against 1,815 sq ft, exactly double, and 4.2 against 8.3 percent of the drainage area. On the reference cell it is the difference between 40.5 hours, which passes, and 81.1 hours, which this module refuses.","displaced_by":"A factor named by the governing jurisdiction, which supersedes outright. Failing that, the number and spread of the field tests: a factor of 2 on a single boring is a different claim from a factor of 2 on one test per 5,000 sq ft of bottom area.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS for a national default. Jurisdictional factors that supersede this one run from 2 to 4, and some scale the factor with the number of tests or with the consequence of failure. 2 is the low end and is therefore the least conservative choice consistent with common practice, which is recorded here rather than left to be discovered: a factor of 2 on a single test in a heterogeneous soil is not a safety factor, it is an average of one measurement.","overridable":true},{"key":"practice.sidewall_credit","quantity":"Fraction of the wetted sidewall area credited with infiltration","value":0.0,"unit":"fraction","class":"C5","confidence":"Low","worth":"On the reference cell the walls below the media surface are 320.1 sq ft against a bottom of 691.6, so crediting them adds 46.3 percent to the infiltrating area and takes the drawdown from 40.5 to 30.6 hours. It decides the answer at the margin: at a safety factor of 4 the same cell goes from 81.1 hours, which is refused, to 61.2 hours, which is still refused, and at a measured 0.75 in/hr it moves a 54 hour cell to 41 hours and turns a failure into a pass.","displaced_by":"A jurisdictional rule, which several publish in both directions, or a measured rate taken on the excavation wall rather than on the bottom.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. Sidewall infiltration is real and the default declines to credit it, for three reasons that are stated because the choice costs the designer footprint. The walls are the surface most likely to be smeared and sealed by the excavator bucket, and nobody tests them. The credit is largest when the practice is fullest, which is exactly the condition the design case checks. And the wetted wall area falls as the practice drains, so a credit taken at full stage flatters the drawdown by more than the area ratio suggests.","overridable":true},{"key":"practice.drawdown_limit_hours","quantity":"Maximum time to empty a practice from full","value":48.0,"unit":"hr","class":"C3","confidence":"Moderate","worth":"It is the criterion that sizes infiltration practices, not the capture volume. Moving 48 to 72 hours raises the storable depth at a design rate of 0.5 in/hr from 24 to 36 in and drops the bottom area needed for 1 in over half an acre of pavement from 908 to 605 sq ft, a third less land. In the other direction it is the whole of the refusal: the reference cell at a safety factor of 4 empties in 81.1 hours, and a cell that stands full for three days is a mosquito habitat, a dead planting and a practice with no capacity when the next storm arrives.","displaced_by":"The governing jurisdiction's own figure, which supersedes. Where a manual sets different limits for the surface ponding and for the subsurface reservoir, both apply and this single limit is not sufficient.","source":null,"reasoning":"NOT VERIFIED AGAINST A SOURCE. 48 hours is the most widely required figure and 72 hours is the common alternative, and no specific manual text was checked in setting this default, so it is shipped as a jurisdictional value that the user must confirm rather than as one this product has read. The governing manual supersedes it and the engine records which was used.","overridable":true},{"key":"practice.treatment_depth_in","quantity":"Depth of runoff over the contributing impervious area to be retained","value":1.0,"unit":"in","class":"C3","confidence":"Moderate","worth":"It scales the required volume exactly linearly and therefore scales the footprint. On half an acre of pavement it is 1,815 cu ft at 1.0 in against 2,722 cu ft at 1.5 in, and a bottom area of 908 against 1,361 sq ft at a design rate of 0.5 in/hr and a 48 hour limit. The common 1.2 in variant is 20 percent on both.","displaced_by":"The governing jurisdiction's water quality volume, which supersedes, including where it is expressed as a percentile storm depth or as a runoff reduction target rather than as a depth over impervious area.","source":null,"reasoning":"NOT VERIFIED AGAINST A SOURCE. One inch over the contributing impervious area is the most common form of the requirement, and the depth, the area it is taken over and whether pervious contributing area counts all vary by jurisdiction. No specific manual was checked. The engine takes the depth over impervious area only, which understates the required volume wherever pervious area also drains to the practice, and says so on the result rather than absorbing it into the default.","overridable":true},{"key":"practice.loading_ratio_warn","quantity":"Contributing area to practice area ratio above which the practice is flagged","value":15.0,"unit":"dimensionless","class":"C5","confidence":"Low","worth":"At 15 to 1 the treatment volume is 15 in of water over the bottom area against the reference cell's 20.3 in of storage, so 74 percent of the practice is spoken for by the design storm alone, with nothing left for the sediment that arrives with it. The flag is worth what pretreatment is worth, which is the difference between a cell that lasts fifteen years and one that clogs in three.","displaced_by":"A jurisdictional loading ratio, which several publish, or a designed pretreatment train with its own sizing, which is what the flag is asking for.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS for a national figure. The conventional range quoted for bioretention is roughly 5 to 1 up to 15 to 1 of contributing area to filter bed area, and the upper end is where the surface loading rate, not the storage volume, starts to govern: clogging, scour across the media surface and pretreatment capacity are all functions of the ratio, and none of the three is modeled anywhere in this engine. The flag says so instead of implying the model covers it.","overridable":true},{"key":"practice.loading_ratio_refuse","quantity":"Contributing area to practice area ratio above which sizing is refused","value":20.0,"unit":"dimensionless","class":"C5","confidence":"Low","worth":"This is the ratio at which the arithmetic runs out rather than a preference. The required volume and the available storage both scale with the practice's bottom area, so the ratio at which they are equal is a property of the profile alone: the reference cell stores 20.3 in of water over its bottom, so at 20.3 to 1 the treatment volume fills it to the overflow and no choice of footprint changes that. Above 20 to 1 the practice as profiled cannot hold the volume however it is proportioned, and the only remedies are a deeper profile, a shallower treatment depth or a second practice.","displaced_by":"A jurisdictional maximum, which supersedes. A deeper profile moves the crossing point and the engine recomputes it from the practice's own effective storage depth rather than from this number, which is the default ceiling and not the physics.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. 20 to 1 is set at the crossing point computed above for the default bioretention profile, which is where the conventional upper end of 15 to 1 also sits once a margin for sediment and for the frustum loss on the ponding volume is allowed. A practice loaded beyond this is outside what the practice does, and approximating it would produce a capture volume that is arithmetically correct and professionally wrong.","overridable":true},{"key":"practice.pavement_run_on_ratio_max","quantity":"Maximum run-on area to pavement area ratio for permeable pavement","value":5.0,"unit":"dimensionless","class":"C5","confidence":"Low","worth":"The stone reservoir has to hold the treatment depth over everything that drains to it, so at a porosity of 0.40 the depth of stone required for 1 in of treatment is 2.5 in of reservoir per unit of ratio: 12.5 in at 5 to 1, which a typical 12 to 18 in section can just carry, and 25 in at 10 to 1, which it cannot. The ceiling is where the section runs out, not where the pavement stops working.","displaced_by":"A jurisdictional limit, which several publish and which is usually lower than this one, or a reservoir sized deeper with the subgrade bearing and the frost depth checked.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS for a single figure. Run-on to permeable pavement carries sediment onto a surface whose whole function is to stay open, and manuals variously cap the ratio at 3 to 1 or 5 to 1, prohibit run-on from pervious areas entirely, or both. 5 to 1 is the permissive end of that and is recorded as a judgment. Run-on from a pervious or unstabilised area is worse than the ratio suggests and nothing in this model sees the difference.","overridable":true},{"key":"practice.side_slope","quantity":"Side slope of the ponding area, horizontal per vertical","value":3.0,"unit":"H per V","class":"C5","confidence":"Low","worth":"It costs storage twice and it is usually taken as free. On the reference cell 3H:1V turns a 1,000 sq ft footprint at grade into a 691.6 sq ft bottom, so 30.8 percent of the plan area is side slope, and the ponding volume is a frustum of 631.8 cu ft against the 750 cu ft a prism on the footprint would claim. On a small 400 sq ft cell with the same 9 in of ponding the bottom is 53.1 percent of the footprint and the prism overstates the ponding volume by 32.1 percent and the total storage by 18.6 percent. Flatter is not safer: at 4H:1V the same cell holds 1,058 cu ft against 1,167.8 at 3H:1V, and because the storage sits over a smaller bottom the drawdown lengthens from 36.6 hours at a vertical wall to 40.5 at 3H:1V and 42.5 at 4H:1V.","displaced_by":"The plan. A side slope is a drawn dimension and this default exists only for a cell described by its area alone.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. 3H:1V is the flattest slope commonly drawn for a mown and maintained cell and the steepest commonly accepted for public safety, so it is where two constraints meet rather than a published value. The engine computes the ponding volume as a frustum on it, because a prism on the footprint is the error this default exists to prevent.","overridable":true},{"key":"practice.orifice_coefficient","quantity":"Discharge coefficient for an underdrain outlet or restrictor","value":0.61,"unit":"dimensionless","class":"C4","confidence":"Moderate","worth":"Almost nothing on an unthrottled underdrain and everything on a throttled one. On the reference cell the media passes 115.3 cu ft/hr and a 4 in underdrain would pass 2,699 cu ft/hr at full stage, so the coefficient changes the answer by exactly nothing until the controlling opening is smaller than 0.82 in, which is where the orifice takes over from the media. Below that it scales the orifice discharge linearly, and on the same cell lined and throttled to a 1 in restrictor the spread between a sharp edged 0.61 and a short tube at 0.80 is 13.24 hours against 11.40, 13.9 percent of the drawdown. It is 13.9 rather than the 23.7 percent the ratio of the coefficients alone would give, because the media still governs while water stands on the surface and the coefficient reaches only the part of the drawdown that happens below it.","displaced_by":"The rating curve for the outlet structure actually specified, which the manufacturer publishes for a proprietary control and which the designer computes for a plate orifice.","source":null,"reasoning":"NOT VERIFIED AGAINST A SOURCE. 0.61 is the standard sharp edged orifice coefficient and appears in every hydraulics text; no specific text was checked here. It is the wrong value for the geometry it is most often applied to, a row of perforations in a pipe wall backed by stone, where the perforation pattern and the stone both matter and the effective coefficient is not a constant at all. The engine models the perforated underdrain as a single orifice and that simplification is worth more than the coefficient is.","overridable":true},{"key":"practice.weir_coefficient","quantity":"Weir coefficient for the overflow or bypass","value":3.0,"unit":"ft^0.5 per s","class":"C4","confidence":"Moderate","worth":"Small, and the overflow is not there to be accurate. On a 4 ft wide overflow passing 5 cfs the head is 0.558 ft at 3.0 against 0.520 ft at the sharp crested 3.33, a difference of 0.46 in of stage, which matters only against the freeboard. What the overflow is worth is structural: without it a storm larger than the design storm accumulates to a stage the practice does not have, and the model would return a capture volume for water that is in fact running down the street.","displaced_by":"The rating for the overflow structure as designed, which for a grate or a domed riser is not a weir at all once it submerges.","source":null,"reasoning":"NOT VERIFIED AGAINST A SOURCE. 3.0 is the standard broad crested weir coefficient in US customary units and 3.33 is the sharp crested value; no specific text was checked. 3.0 is chosen because an earthen or concrete overflow lip is broad crested, and because it gives the higher stage of the two.","overridable":true},{"key":"practice.kind_geometry_defaults","quantity":"Layer depths that each practice kind starts from","value":1.0,"unit":"template flag","class":"C5","confidence":"Low","worth":"The template depths are the whole capture answer, because the effective storage depth over the bottom area is what the loading ratio is checked against. The bioretention template gives 20.3 in of water storage over its bottom, the infiltration trench template 19.2 in, the permeable pavement template 7.2 in and the dry swale template 21.1 in, all four on a 1,000 sq ft footprint. A designer who takes the bioretention template and drops the gravel to 6 in loses 2.4 in of that, 11.8 percent of the storage, and the required footprint grows in the same proportion. The permeable pavement figure is the one to read twice: 600 cu ft on that footprint against the bioretention template's 1,167.8 cu ft on the same ground, because a pavement section has no ponding and no media and the stone is all of it.","displaced_by":"The section on the drawing. These are starting dimensions for a cell described by its area alone, and every one of them is replaced by a designed profile before anything is stamped.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. The templates follow. Bioretention: 9 in ponding over 18 in of media over 12 in of gravel on 3H:1V sides. Infiltration trench: 48 in of stone with vertical walls and no media and no underdrain. Permeable pavement: 18 in of stone with vertical walls, no ponding and no media. Dry swale: 9 in ponding over 18 in of media over 9 in of gravel on 3H:1V sides. They are the middle of what is commonly drawn and they are not a recommendation. A template is applied only where the caller supplies no dimension, and every value it supplies is recorded as defaulted so it prints as an assumption rather than as a design.","overridable":true},{"key":"hydrograph.peak_rate_factor","quantity":"NRCS unit hydrograph peak rate factor","value":484.0,"unit":"cfs hr per sq mi per in","class":"C5","confidence":"Low","worth":"The largest single judgment in the peak computation. The UNIT hydrograph peak scales linearly with it: 62.0 percent of the 484 value at 300 and 124.0 percent at 600. The peak of a storm hydrograph convolved from it moves by less, because a lower peak rate factor is also a wider hydrograph and gathers more of the rainfall excess into its peak: 68.6 percent at 300 and 117.3 percent at 600, so the span of the conventionally published range is worth a factor of 1.71 on the reported peak rather than the factor of two the unit hydrograph peaks suggest. It does not touch the volume at all.","displaced_by":"The NRCS dimensionless curve published for the factor the watershed actually warrants, which this engine does not hold, or a gauged local calibration. Flat, swampy and coastal watersheds are conventionally run at 300 and steep ones at 600.","source":"NEH Part 630 (edition not established by the transcriber) Chapter 16, dimensionless unit hydrograph and peak rate factor","reasoning":"NO EXTERNAL SOURCE EXISTS FOR THE PART THAT IS A JUDGMENT. 484 itself is published as the standard factor, and it is recorded here as C5 because what gets defaulted is not the number but the claim that THIS watershed is a 484 watershed, and nothing published says that about any particular site. 484 is equivalent to a triangular hydrograph whose time base is 2.667 times its time to peak, which is a statement that the recession takes 1.667 times as long as the rise. The factor is a free parameter of the triangular shape only: the dimensionless table in the hydrograph calculation is itself a 484 shape, integrating to an implied 483.05, and that module refuses another factor against the curvilinear shape rather than accepting one the volume rescale would cancel.","overridable":true},{"key":"hydrograph.time_step_warn_ratio","quantity":"Time step to time of concentration ratio at which the peak is flagged","value":0.133,"unit":"dimensionless","class":"C6","confidence":"Low","worth":"At exactly this ratio the computed peak discharge runs below the same case at a 0.005 hr step by 4.27 percent at Tc = 0.375 hr, 6.3 percent at Tc = 0.75 hr and 7.4 percent at Tc = 1.5 hr (measured on the NRCS 0.1 hr distribution tables). The volume is untouched at any step, so this bears on peak discharge and on nothing else.","displaced_by":"A rainfall table finer than the 0.1 hr interval the distributions hold, which is what actually limits the step, or a convergence study against a reference solution on the site in hand.","source":null,"reasoning":"REASONED, NOT MEASURED. About 0.133 Tc is the conventional NRCS unit hydrograph guidance and meeting it is not free: two effects are mixed in the figures above and only one of them is numerical. A longer step is a longer unit duration, which lengthens Tp through Tp = D/2 + 0.6 Tc and genuinely flattens the unit hydrograph, and on top of that the discrete convolution samples the peak less often.","overridable":false},{"key":"hydrograph.time_step_refuse_ratio","quantity":"Time step to time of concentration ratio at which the peak is refused","value":0.5,"unit":"dimensionless","class":"C6","confidence":"Low","worth":"At this ratio the computed peak is 7.7 percent low at Tc = 0.2 hr, 22.4 percent low at Tc = 0.5 hr and 23.8 percent low at Tc = 1.0 hr against a 0.005 hr step, and the error grows without bound as the step grows further. A peak a fifth low is not a conservative approximation, it is an undersized pipe.","displaced_by":"Not evidence-displaceable; it is a product policy line.","source":null,"reasoning":"REASONED, NOT MEASURED. The threshold is set where the error reaches about a fifth on the middle of the range above rather than at a round number, and a fifth is itself a judgment: nothing published says where the line is. The volume is unaffected at any step, so a volume based result is still reported when the peak is refused.","overridable":false},{"key":"routing.time_step_warn_ratio","quantity":"Routing step to practice response time ratio at which the peak is flagged","value":0.25,"unit":"dimensionless","class":"C6","confidence":"Low","worth":"Measured on two 400 cu ft trenches with response times of 0.823 and 3.080 hr, routing a triangular inflow sampled identically at every step so that only the routing changes: the peak discharge runs 0.65 and 3.00 percent below the converged value at this ratio, 2.17 and 6.10 percent below at 0.5, 5.54 and 18.22 percent below at 1.0 and 20.96 and 42.72 percent below at 2.0. The error is a clipped peak and it is one sided, so it is not conservative for a downstream pipe. The volumes are far less sensitive but they are not untouched, and the difference between those two statements matters: the volume BALANCE closes exactly at any step, by construction, while the volume SPLIT shifts a little because how much infiltrates depends on how long the water stands. Across a factor of forty in the step on the first of those trenches, the volume leaving the site moved by 1.5 percent and the infiltrated share by 10 percent of itself, against 2.3 percent on the peak over the same range.","displaced_by":"The step halving check the practice routing runs on the case in hand, which measures the sensitivity of this practice and this storm instead of predicting it from a ratio.","source":null,"reasoning":"REASONED, NOT MEASURED, and weaker than it looks. The response time is the practice's emptying time at the crest, and it is the right scale for the recession and the wrong one for the peak: the peak is governed by the local time constant dS/dz over dO/dz at the stage the routing actually reaches, which at an overflow weir or just above an orifice invert is orders of magnitude shorter than the emptying time and cannot be known before the routing is run. This ratio is therefore a cheap screen that catches the gross case, and the practice routing additionally routes the same storm at half the step and reports the difference in the peak, which is the check that catches what this one misses.","overridable":false},{"key":"routing.time_step_refuse_ratio","quantity":"Routing step to practice response time ratio at which routing is refused","value":1.0,"unit":"dimensionless","class":"C6","confidence":"Low","worth":"On the same two measurements, the peak discharge is 5.54 and 18.22 percent below the converged value at this ratio and 20.96 and 42.72 percent below at twice it. The spread between the two practices is as large as the error itself, which is the reason this is a screen and not a correction.","displaced_by":"Not evidence-displaceable; it is a product policy line.","source":null,"reasoning":"REASONED, NOT MEASURED. Set at the ratio where one step drains the whole practice from its crest, because a recession described by its two endpoints is not a routing, and because that is also where the measured peak error passes a twentieth on the better behaved of the two practices and a fifth on the other. Where the line falls inside that range is a judgment and nothing published says where it is. Refused rather than warned because the peak is what a downstream pipe is sized on.","overridable":false},{"key":"practice.riser_control_transition","quantity":"Rule for a riser inlet moving from weir control to orifice control","value":1.0,"unit":"policy flag","class":"C5","confidence":"Low","worth":"It is worth the difference between a structure that passes what it passes and one that passes 227 percent more. The two ratings for a riser inlet cross at h = Cd D sqrt(2 g) / (4 C), which is 0.408 riser diameters at the coefficients this registry ships, so an 18 in riser drowns at 0.61 ft of head: at 2 ft of head the weir equation alone gives 39.99 cfs against the 12.23 cfs a submerged 18 in inlet passes. Routed, on the reference extended detention basin under two acres of impervious cover in a 6 in Type II storm, modeling the 12 in riser as a weir over its whole range gives a peak discharge of 8.19 cfs against 4.78 cfs and a maximum stage of 2.862 ft against 3.114 ft: the peak is 71.1 percent high, which oversizes the pipe downstream, and the stage is 0.252 ft low, which undersizes the embankment. The error runs in the dangerous direction on the figure the embankment is built to and in the wasteful direction on the figure the pipe is built to, which is why it survives review.","displaced_by":"The rating curve for the structure as specified, which a manufacturer publishes for a proprietary riser and which a designer computes from the inlet geometry for a cast one. A physical model or a gauged structure displaces it outright.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS for the rule itself, which is the finding rather than a gap in the transcription: taking the smaller of the weir and the orifice rating is the universal convention and no publication this transcriber can name derives it or bounds its error. Both underlying equations are textbook and both are recorded here as unverified through practice.orifice_coefficient and practice.weir_coefficient. What the rule gets wrong is the transition itself, which is not a point: over roughly a diameter of head either side of the crossing the inlet runs in an unstable vortex regime and passes LESS than either curve, so the model overstates the discharge and understates the stage there by an amount nobody here has measured. The rule is kept because the alternatives are worse: adding the two ratings double counts one opening, and running the weir past the crossing is the 227 percent error above.","overridable":true},{"key":"practice.dead_storage_excluded","quantity":"Storage below a permanent pool is excluded from the capture volume","value":1.0,"unit":"policy flag","class":"C5","confidence":"Low","worth":"On the reference wet pond it is the difference between 87,319.3 cu ft and 36,176.8 cu ft. The dead storage is 58.6 percent of what the pond holds to its crest, so counting it would report 141.4 percent more capacity than the facility has, and a pond with only 41.4 percent of the live storage its requirement asks for would be reported as meeting it. The error is not a fixed fraction and it is not conservative in either direction: it scales with the pool depth, which is the one dimension a designer moves to buy residence time, so it is largest on the facilities that were designed most carefully. The same exclusion runs through the routing, which starts a wet facility at its pool rather than at its bottom: routing the reference pond from empty would put the whole 51,142.5 cu ft of pool on the inflow side of the volume balance and report the pond as capturing water that was in it before it rained.","displaced_by":"A jurisdiction that credits the permanent pool volume toward its water quality volume, which some do, in which case the jurisdiction's own accounting supersedes this one. The engine reports both figures separately so either rule can be applied to them, and it never adds them.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS for the accounting choice, and it is a choice rather than a computation: the two volumes are both computed exactly from the same stage-storage curve and what is in question is which of them the word capture refers to. Dead storage is full when the storm starts and full when it ends, it attenuates nothing and it takes no water out of the event, so calling it capacity describes the facility wrongly. What it does do is buy residence time, which is what the treatment in a wet facility actually rests on, and this engine has no term for residence time, for settling, for a length to width ratio or for short circuiting. So the exclusion is right about the volume and silent about the treatment, and the result says so rather than letting the silence read as a pass.","overridable":true},{"key":"practice.detention_loading_ratio_exempt","quantity":"Kinds with no filter bed are exempt from the loading ratio ceiling","value":1.0,"unit":"policy flag","class":"C5","confidence":"Low","worth":"Applying the ceiling to these kinds would refuse every one of them. Two acres of impervious cover onto the reference extended detention basin is 87,120 sq ft over a 4,022.3 sq ft bottom, which is 21.7 to 1, above the 20 to 1 ceiling, and that is a small drainage area for a basin: the ratio for a real one runs past 100 to 1 and nothing is wrong with it. The same two acres onto the reference wet pond is 9.1 to 1 and passes, which is the clearest evidence the ceiling is measuring the wrong thing here: it separates the two facilities by their plan area and not by anything either of them does. What is given up by exempting them is not small, because nothing replaces it: the checks that govern a detention facility are the release rate against the allowable, the residence time, the forebay volume and the embankment, and this engine computes none of the four.","displaced_by":"A jurisdictional loading limit written for detention rather than for filtration, which supersedes, or a release rate criterion, which is what actually sizes these facilities and which this engine would need an allowable discharge to check against.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. The loading ratio in this registry is a filter bed criterion and says so: practice.loading_ratio_warn reasons about clogging, scour across the media surface and pretreatment capacity, and all three are properties of a bed that water passes through. A detention vault, a dry extended detention basin, a wet pond, a wet swale and a constructed wetland have no bed to load. The exemption is granted to those five kinds by name and to nothing else: an infiltration basin and a drywell have no media either and keep the ceiling, because their bottom is the outlet and clogging it is exactly how they fail. The ratio is still computed and still reported for the exempt kinds, with a note saying it was not checked and what would have to be checked instead, because a ratio that disappears from the output is a ratio nobody notices was never applied.","overridable":true},{"key":"practice.added_kind_templates","quantity":"Layer depths that the kinds added after the first four start from","value":1.0,"unit":"template flag","class":"C5","confidence":"Low","worth":"The kind chosen moves the answer further than any coefficient in this module does. On the same 1,000 sq ft footprint the seven dry templates give effective storage depths over their own bottom of 15.1 in for micro-bioretention, 15.9 in for a tree box, 24.0 in for a drywell, 33.3 in for a sand filter, 45.7 in for an infiltration basin and 60.0 in for a detention vault, which is a factor of 4.0 from end to end, against the 20.3 in the bioretention template gives on the same ground. The extended detention basin template is the one to read twice: at 1,000 sq ft the 3H:1V side slopes leave 116.5 sq ft of bottom and the effective depth reads 155.8 in, which is not storage worth having but a footprint that has almost entirely become side slope, and below 648 sq ft the same template has no bottom at all. The three wet templates supply geometry and no permanent pool, so they have no effective depth until a pool is declared, which is the point of not defaulting one.","displaced_by":"The section on the drawing. These are starting dimensions for a facility described by its area alone, and every one of them is replaced by a designed profile before anything is stamped.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. A second entry rather than an edit to practice.kind_geometry_defaults, and the reason is that the first entry quotes the effective depth of each of its four templates and a test pins those four figures: folding seven more into it would either leave that worth stale or reword a number a report has already printed. The templates follow. Micro-bioretention: 6 in ponding over 24 in media over 6 in gravel on the registry side slopes, with an underdrain. Infiltration basin: 24 in ponding over prepared subgrade, with no media, no stone and no underdrain. Sand filter: 24 in ponding over 18 in sand over 12 in stone, vertical walled, with an underdrain. Tree box: 6 in ponding over 30 in media over 6 in stone, vertical walled, with an underdrain. Drywell: 60 in of stone in a square vertical walled pit. Detention vault: 60 in of open chamber, carried as ponding at a side slope of zero because it is chamber volume and not stone voids. Dry extended detention basin: 36 in of ponding on the registry side slopes, with no underdrain and no outlet. Wet pond: 72 in on the registry side slopes at a 3 to 1 aspect. Wet swale: 18 in at a 10 to 1 aspect. Constructed wetland: 36 in at a 2 to 1 aspect. They are the middle of what is commonly drawn and they are not a recommendation. Neither the detention vault nor the extended detention basin template carries an outlet structure, and both are refused as unsizeable without one, because the invert elevations in a riser are the release rates the jurisdiction asked for and there is no default release rate.","overridable":true},{"key":"practice.green_roof_refused","quantity":"A green roof is refused rather than modeled with an assumed ET rate","value":1.0,"unit":"policy flag","class":"C5","confidence":"Low","worth":"The whole of the credit, and the arithmetic is why the refusal stands. A 4 in green roof media at the 0.25 porosity this registry ships holds 1.00 in of water, which is 83.33 cu ft over 1,000 sq ft, against a water quality volume of exactly 83.33 cu ft over the same 1,000 sq ft at a 1 in treatment depth: a ratio of 1.000, so the roof either takes the entire design storm or takes none of it, and which of the two depends on nothing but whether it was dry when the storm arrived. That is set by the evapotranspiration rate, and over the plausible range of 0.10 to 0.25 in/day the roof empties in 10.0 days or 4.0 days, a factor of 2.5. A seasonal ET rate would therefore not be a refinement of the answer, it would BE the answer, and this product has no source for it, no interevent period and no season.","displaced_by":"A media water holding capacity measured on the specified mix together with a local evapotranspiration series and an interevent period, which is the computation a green roof is actually sized by and which is not the one this module performs. A jurisdictional credit for a stated media depth supersedes both.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS for a single design evapotranspiration rate, and that is the finding rather than a gap in the transcription: the rate depends on the climate, the month, the planting and the media, and the published values are series and not constants. Adding a constant rate to this module would be half a day's work and the drawdown would come out as the media storage divided by the rate, which is arithmetic dressed as hydrology. The refusal is recorded here rather than left as a sentence in the practice model so that it is ranked, priced and visible beside the values the engine does ship.","overridable":true},{"key":"practice.rainwater_demand_refused","quantity":"A cistern is refused rather than credited against an assumed demand","value":1.0,"unit":"policy flag","class":"C5","confidence":"Low","worth":"The entire credit, and it is the practice most often credited on paper and least often performing. A 1,500 gallon cistern is 200.5 cu ft, which is 160.4 percent of the 125.0 cu ft water quality volume of the 1,500 sq ft roof it drains at a 1 in treatment depth, so on paper the tank covers the requirement with 60 percent to spare. It covers it only if the tank is empty when the storm arrives, which is a statement about the demand and not about the tank. An irrigation demand is zero in the month the design storm arrives, so the credit is either 200.5 cu ft or 0.0 cu ft, and an annual average demand splits the difference and is wrong in both months.","displaced_by":"A metered demand series for the end use, month by month, together with a written commitment that the use continues: an indoor demand such as toilet flushing or cooling tower makeup is continuous and is a different practice from irrigation with the same tank on it. A jurisdictional credit that already prices demand continuity supersedes.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS for a demand this engine could assume. Two things are missing and neither is a coefficient. The demand is a monthly series and this module has no time axis longer than a storm, so there is nothing here to hang one on. And the water drawn from a cistern leaves the site as neither runoff nor recharge: it goes to a lawn, a toilet or a cooling tower, which is a fifth volume category beside the infiltration, underdrain, outlet and overflow this module carries, and folding it into the infiltrated volume would claim groundwater recharge for water that went down a drain. Adding the category is the work; inventing the demand is the thing that must not be done, and it is the reason the practice is refused rather than shipped.","overridable":true},{"key":"practice.jurisdictional_credit_refused","quantity":"Credit only practices are refused for want of a jurisdiction layer","value":1.0,"unit":"policy flag","class":"C5","confidence":"Low","worth":"The span between the whole credit and none of it, and this engine cannot narrow it by a cubic foot. Disconnecting a 1,500 sq ft roof from the storm system is worth between 0.0 and 125.0 cu ft of the 1,815.0 cu ft required over a half acre of impervious cover at a 1 in treatment depth, which is between 0.0 and 6.9 percent of the requirement, and which of the two it is depends on a flow path length, a receiving slope and a table in a manual. The same is true of a filter strip, a grass channel, sheet flow to a conservation area and a compost amendment. The worth of REFUSING them is that the alternative is a number: a routing invented for a practice with no storage would produce a volume that looks computed, carries this engine's provenance, and is a jurisdictional rule somebody guessed at.","displaced_by":"A jurisdiction layer, which this product does not have. Until there is one, the governing manual's own credit entered as a reduction in contributing impervious area, which keeps the arithmetic in the engine and the rule with the person who is allowed to apply it.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS in this engine, which is the whole point: the sources exist, in state and county manuals, and this product holds none of them and has no way to read one. The five kinds are carried in PracticeKind and refused at construction rather than left out of the vocabulary, and that is the judgment being recorded. A user who types grass channel should be told what a grass channel needs and this engine lacks, not that the words are not a practice, and an interface built against the enum then has the shape of the missing thing already drawn for the day the jurisdiction layer exists.","overridable":true},{"key":"segmentation.material_gap_fraction","quantity":"Unresolved soil share above which no headline number is produced","value":0.1,"unit":"fraction","class":"C6","confidence":"Low","worth":"A depth computed over the resolved share and applied to the whole site is low by the gap fraction before anyone asks what is in the gap, and by about twice that where the gap is the wetter group. Resolving the rest of the site to B and letting the gap really be D, which is the ordinary suburban pairing of graded fill against bottomland, is CN 69 against CN 84 and 0.6697 in against 1.5163 in at P = 3 in: a volume per unit of site area of 0.6027 against a true 0.7544, which is 20.1 percent low at a tenth unresolved and 10.6 percent low at a twentieth.","displaced_by":"Nothing displaces it with evidence, because it is a decision about when one number may stand for a site rather than a measurement of anything. A reviewer who wants five percent has a good argument, and the threshold is a parameter on segment_site as well as an entry here.","source":null,"reasoning":"REASONED, NOT MEASURED. A tenth is not a level below which the gap is harmless: at a twentieth the compounded error above is already over ten percent. It is the point at which warning stops being a proportionate response and refusing starts. Every gap is reported with its acreage and its map unit name whatever its size, so the threshold never decides what a user is told, only whether a single number gets to stand for the site.","overridable":true},{"key":"landcover.min_pixels","quantity":"Minimum pixel count for a reportable land cover share","value":30.0,"unit":"pixels","class":"C6","confidence":"Low","worth":"The sampling error alone. The standard error of a share estimated from n pixels is sqrt(p(1-p)/n), worst at p = 0.5, which at n = 30 is 9.13 percentage points on the impervious fraction. That moves a composite curve number by roughly 4 units and a design volume by more than a tenth. The boundary discretization is a second and independent error of about twice the pixel size over the site's side length. The consequence is the uncomfortable one and it is the point: a quarter acre lot is 1,011.7 sq m against a 900 sq m pixel, so it is 1.12 pixels and can never clear this bar.","displaced_by":"A higher resolution land cover product, or a site survey. Neither raises this number; both make it easy to clear, which is the difference between displacing a threshold and arguing with it.","source":null,"reasoning":"REASONED, NOT MEASURED. 30 is the round number where both readings above say the result is still worth reporting and is no longer worth reporting without a warning. The figure has lived in the land cover reader as MIN_PIXELS with this argument beside it and no registry entry; the entry is added here so the taxonomy id CN-XWK-006 has a threshold to point at. the land cover reader still holds the literal and still raises this as a caveat rather than as a refusal, which is a gap named in the list of refusals rather than closed here.","overridable":true},{"key":"segmentation.min_pixels","quantity":"Boundary size, in 30 m land cover pixels, below which derived land cover does not stand for the site","value":3.0,"unit":"pixels","class":"C6","confidence":"Low","worth":"Below three pixels the land cover of the whole site is read off one or two 30 m squares that mostly lie outside it: a 0.3 ac lot is 1.35 pixels. The cover shares can then be anything from all lawn to all pavement for the same lot, which on group B at P = 3 in is 0.670 in against 2.77 in.","displaced_by":"Surfaces drawn or typed by the engineer (roof, pavement, lawn), or a waiver with a reason.","source":null,"reasoning":"REASONED, NOT MEASURED. The owner's round one decision: warn under ten pixels (about 2.2 ac) and require hand entered surfaces under three, as a waivable refusal highlighted in the report. Refusal CN-SEG-006.","overridable":true},{"key":"segmentation.warn_pixels","quantity":"Boundary size, in 30 m land cover pixels, below which derived land cover is flagged as coarse for the site","value":10.0,"unit":"pixels","class":"C6","confidence":"Low","worth":"At ten pixels one pixel read as the wrong class moves the cover share by a tenth of the site.","displaced_by":"Surfaces drawn or typed by the engineer.","source":null,"reasoning":"REASONED, NOT MEASURED. The owner's round one decision; a warning only, never a refusal.","overridable":true},{"key":"site.area_closure_warn","quantity":"Difference between a condition's total area and the drainage area above which a run warns","value":0.005,"unit":"fraction","class":"C6","confidence":"Low","worth":"A runoff volume is depth times area, so the volume is off by the same share as the area.","displaced_by":"Areas that add up to the drawn boundary.","source":null,"reasoning":"REASONED, NOT MEASURED. Half a percent is about the rounding of areas typed to two decimals on a small site. The owner's round one decision.","overridable":true},{"key":"site.boundary_refuse_acres","quantity":"Boundary area above which a run refuses to report a volume and peak unless waived","value":16000.0,"unit":"acre","class":"C6","confidence":"Low","worth":"16,000 acres is 25 square miles. Above it a run is refused (CN-GEO-005, waivable, highlighted); a boundary above 2,000 acres is warned about; the hard save limit is 1,000,000 acres. All three are settings (Limits).","displaced_by":"A jurisdiction rule on the largest site the method may be used for.","source":null,"reasoning":"REASONED, NOT MEASURED. The upper range the review triage gave for TR-55's small-watershed methods, to be checked against TR-55 (1986) in the partner check.","overridable":false},{"key":"site.area_closure_refuse","quantity":"Difference between a condition's total area and the drainage area above which a run refuses to report a volume","value":0.02,"unit":"fraction","class":"C6","confidence":"Low","worth":"At two percent the volume and the peak are two percent off before anything else is wrong; a unit slip (square feet typed as acres) is a factor of 43,560.","displaced_by":"Areas that add up to the drawn boundary, or a waiver.","source":null,"reasoning":"REASONED, NOT MEASURED. The owner's round one decision: 0.5 percent warns, 2 percent is a waivable refusal (CN-SEG-007), both highlighted in the report.","overridable":true},{"key":"refusal.waiver_reason_base_words","quantity":"Words of reason demanded of a waiver","value":5.0,"unit":"words","class":"C6","confidence":"Low","worth":"It is the floor on every waiver in the product, and it is the same floor as every other reason a person types: 5 words, whatever the size of the departure. It separates a filed reason from a full stop; the size of the departure is shown beside the reason in the report rather than demanded as more words.","displaced_by":"A sample of waivers actually filed against this engine, scored by a licensed reviewer for whether the reason answers the refusal. The word count is a proxy for substance and the substance is what a reviewer reads, so a scored sample would replace it with a criterion rather than raise the number.","source":null,"reasoning":"REASONED, NOT MEASURED. NO EXTERNAL SOURCE EXISTS for how long a professional justification should be. THE OWNER'S DECISION (D-14, 24 September 2026): every reason box in the product takes 5 to 512 words, shown before typing, with no scaling by the size of the change. A large change is flagged in the report instead. The product's shared rule is cnapp.reasons (5 to 512 words); this entry holds the lower end of it for the engine.","overridable":true},{"key":"refusal.waiver_reason_words_per_doubling","quantity":"Additional words demanded per doubling of the departure","value":0.0,"unit":"words per doubling","class":"C6","confidence":"Low","worth":"Zero: the demand does not grow with the departure. It used to add 8 words per doubling, so a waiver asked for 8 to 35 words and a person learned the figure only after a refused attempt. The departure itself is printed beside the reason in the report.","displaced_by":"The owner's decision. A scored sample of filed waivers showing that argued waivers get longer as the departure grows would be the case for bringing scaling back.","source":null,"reasoning":"REASONED, NOT MEASURED. THE OWNER'S DECISION (D-14, 24 September 2026): every reason box in the product takes 5 to 512 words, shown before typing, with no scaling by the size of the change. A large change is flagged in the report instead.","overridable":true},{"key":"refusal.waiver_reason_cap_words","quantity":"Ceiling on the words of reason any waiver can be asked for","value":512.0,"unit":"words","class":"C6","confidence":"Low","worth":"The upper end of the shared rule, 5 to 512 words. With no scaling the demand never reaches it; the application refuses a reason longer than this before it reaches the engine.","displaced_by":"The owner's decision on reason length.","source":null,"reasoning":"REASONED, NOT MEASURED. THE OWNER'S DECISION (D-14, 24 September 2026): every reason box in the product takes 5 to 512 words, shown before typing, with no scaling by the size of the change. A large change is flagged in the report instead.","overridable":true},{"key":"refusal.attestation_min_words","quantity":"Words demanded of what an evidence waiver rests on","value":5.0,"unit":"words","class":"C6","confidence":"Low","worth":"It is the entire difference between an attestation and an assertion, and therefore between the evidence class and the policy class. Five words is enough for a document, a date and a who: 'boring log B-3, Geotech Inc, 2026-04-11' is six. It is not enough for 'I am confident', which is three and is the record this bar exists to keep out of the file.","displaced_by":"A typed attestation: a reference to a document the product holds, with its own provenance, rather than free text. That is the real fix and it is an application layer change, not a larger number here.","source":null,"reasoning":"REASONED, NOT MEASURED. THE OWNER'S DECISION (D-14, 24 September 2026): every reason box in the product takes 5 to 512 words, shown before typing, with no scaling by the size of the change. A large change is flagged in the report instead. The attestation box is a reason box, so it takes the same 5 to 512 words.","overridable":true},{"key":"practice.green_roof_media_capacity","quantity":"Water a green roof medium holds, per inch of medium","value":0.25,"unit":"in of water per in of media","class":"C5","confidence":"Low","worth":"The whole of the credit, because the credit is all or nothing and this number decides which. At 0.25 a 4 in medium holds 1.00 in, which is 83.3 cu ft over 1,000 sq ft against a water quality volume of 83.3 cu ft over the same roof at a 1 in treatment depth: a ratio of 1.000, so the roof either takes the design storm or does not. At 0.20 the same medium holds 0.80 in, 66.7 cu ft, 80.0 percent of the requirement, and the roof runs 20.0 percent of the storm off however dry it was. At 0.35 it holds 1.40 in, 116.7 cu ft, 140.0 percent, and clears the requirement with room. Across that span the credit moves by a factor of 1.75.","displaced_by":"A maximum media water retention figure measured on the specified mix at the specified depth, which green roof media suppliers publish for their own build-ups and which is the number a green roof is actually sized on. A jurisdictional retention credit for a stated media depth supersedes both.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS in this product for a green roof growing medium, and the value shipped is the bioretention media porosity reused. That reuse is the judgment being recorded and it is not a small one: an engineered bioretention soil and an extensive green roof medium are different materials with different grading, and the second is chosen for weight as much as for storage. The number is kept at 0.25 rather than adjusted because adjusting it would be inventing a second figure with no more behind it than the first, and because 0.25 is what practice.green_roof_refused already prices the refusal against.","overridable":true},{"key":"practice.cistern_demand_within_storm","quantity":"A month's demand is drawn evenly through the design storm rather than at the hour it is used","value":1.0,"unit":"policy flag","class":"C5","confidence":"Low","worth":"Very little, and the smallness is the finding rather than an excuse. On a 1,500 gallon cistern, which is 200.5 cu ft, at an indoor demand of 100 gallons a day, which is 13.4 cu ft a day, a 24 hour storm draws 13.4 cu ft: 6.7 percent of the tank and 10.7 percent of the 125.0 cu ft water quality volume of the 1,500 sq ft roof it drains. The two extreme alternatives, drawing the whole day's demand before the storm or none of it during, move the answer by at most that same 13.4 cu ft. What decides a cistern's credit is not this: it is the antecedent state, which spans the whole 200.5 cu ft and is required from the user rather than assumed here.","displaced_by":"An hourly demand profile for the end use together with the clock time of the design storm. The first exists for indoor uses, as published diurnal water use patterns; the second does not exist at all, because a design storm is a depth and a distribution and has no time of day in it. That is why this is recorded as a judgment rather than left as a gap to be filled.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS for the within storm timing of a harvested water demand, and there could not be one that this engine could use, because the design storm has no clock. The even spread is the only assumption that does not privilege an hour, and the entry exists so that the assumption is visible and priced beside the input that actually matters.","overridable":true},{"key":"practice.supplied_curve_extension","quantity":"A supplied stage-storage table is refused above its last row rather than extrapolated","value":1.0,"unit":"policy flag","class":"C5","confidence":"Low","worth":"Everything the table does not say, which on a real device is a third of it. A chamber system rated to 3.00 ft holding 1,200 cu ft over a last interval of 400 sq ft per foot, asked to hold a storm that reaches 4.00 ft, is credited with 400 cu ft more by carrying that last slope upward: 33.3 percent of its entire rated storage, invented by this engine and printed under the manufacturer's name. That is refused. The extension itself still exists inside the stage solve, because a routing that cannot place a cubic foot loses it and still reports a balance of 0.0 cu ft, and the refusal fires the moment the routed stage passes the top of the table, so no reported figure ever stands on it.","displaced_by":"A longer table. The manufacturer has one or the device has not been rated that high, and those are two different findings that the designer can tell apart in a phone call and this engine cannot tell apart at all.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS for what a device does above the last stage its maker rated, and that is the point rather than a gap: the tables ARE the source, and past their last row there is no source. Silently extrapolating is how a device gets credited for storage it does not have, and it is the failure mode of the widest door in this product.","overridable":true},{"key":"estimate.green_roof_crop_coefficient","quantity":"Ratio of an extensive green roof's evapotranspiration to the grass reference","value":0.6,"unit":"dimensionless","class":"C5","confidence":"Low","worth":"It is the whole of the green roof credit, because it multiplies the one figure the credit turns on. On the reference roof in July the media empties in 7.95 days at this value, 11.92 days at 0.40 and 4.77 days at 1.00, a factor of 2.50 end to end. Against the median 4 day dry spell the credit runs 28.0, 41.9 and 69.9 cu ft of the 83.33 cu ft required, which is the same factor of 2.50. Against the ninetieth percentile 11 day dry spell it runs 76.9, 83.3 and 83.3 cu ft, so the coefficient is the difference between a roof that misses the requirement by 7.7 percent and one that meets it, and the boolean flips between 0.40 and 0.60 with nothing visible happening in between.","displaced_by":"A manufacturer's assembly test on the specified build-up, which green roof media suppliers run and which measures the actual assembly rather than a ratio to a grass surface. Failing that, a published crop coefficient for an extensive green roof of the stated media depth and planting, read against the same ASCE short crop reference gridMET publishes. A jurisdictional retention credit for a stated media depth supersedes both.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS for this purpose, which is what the climate reader says in its own words before any number is fetched: a reference evapotranspiration is a property of the weather over a standardized well watered short grass surface, and a green roof is not that surface. The real ratio depends on the media depth, its water retention, the planting, the age of the planting and, dominantly, on how much water is in the media at the time, which falls through a dry spell so that the coefficient this entry treats as a constant is in fact largest on the first day and smallest on the last. 0.60 is the middle of the range published crop coefficients for extensive green roofs scatter across and it is a judgment, not a reading. The direction of its error is known: a constant coefficient overstates recovery late in a long dry spell, which is the spell a design is sized on.","overridable":true},{"key":"estimate.dry_spell_percentile","quantity":"Percentile of the interevent dry period distribution taken as the design antecedent dry period","value":90.0,"unit":"percentile","class":"C6","confidence":"Low","worth":"At the reference station the distribution is P10 1 day, P25 2, P50 4, P75 7, P90 11, P95 15, P99 22, mean 5.42 and longest 42, so the choice spans a factor of 22 across the published percentiles and a factor of 2.75 between the median and this entry. On the reference roof in July it is the credit: 41.9 cu ft at P50 and 83.3 cu ft at P90 of the 83.33 cu ft required, which is 98.8 percent more and is the difference between a roof that releases half the design storm and one that releases none of it. The margin is what the percentile really buys: at P90 the media empties with 3.05 days to spare, at P75 it is 0.95 days short and at P50 it is 3.95 days short.","displaced_by":"A continuous simulation over the local record, which answers the question this percentile is a proxy for: how often the roof is full when a storm arrives, rather than how dry a typical gap is. A jurisdiction that names an antecedent dry period supersedes outright. Failing both, a local gauge record in place of the station this one reads, which is 4.8 km away and is a different point.","source":null,"reasoning":"REASONED, NOT MEASURED, and NO EXTERNAL SOURCE EXISTS for the choice. A median dry spell is the typical case and a design is not sized on the typical case: half of all storms would arrive on a roof with less recovery than the design assumed. The ninetieth percentile is the conventional bar for this kind of antecedent assumption and it is chosen for that and for nothing stronger. It runs in the GENEROUS direction, which is the opposite of how a safety factor usually runs, and that is the thing to notice: a longer assumed dry period empties more of the media and credits the roof with more storage, so the conservative choice here is a LOW percentile and this entry does not take it. It does not, because the credit is what the roof does on the storm the design names rather than on every storm, and a P10 antecedent period would refuse every green roof in the country. The entry records which way it errs instead of implying it errs safely.","overridable":true},{"key":"estimate.irrigation_plant_factor","quantity":"Plant factor in a landscape water budget","value":0.5,"unit":"dimensionless","class":"C4","confidence":"Low","worth":"It scales the irrigation demand exactly linearly and the demand is the whole of a cistern's drawdown. On 5,000 sq ft of irrigated landscape at the reference site in July the demand is 58.2 cu ft per day at this value, 34.9 at 0.30 and 93.2 at 0.80. A 1,500 gallon tank is 200.5 cu ft, so it empties in 3.4 days at this value, 5.7 days at 0.30 and 2.2 days at 0.80, against a median 4 day dry spell: over that spell the tank is drawn 116.2 percent of its volume at 0.50 and 69.7 percent at 0.30, so it is empty when the storm arrives at the shipped value and 30.3 percent full at the low end, and the credit is the empty space. Taken with estimate.irrigation_efficiency at its own generous end the tank is 41.9 percent full.","displaced_by":"A plant factor from the WUCOLS listing or from the state model ordinance that governs the site, for the species actually specified on the landscape plan, which is a document the project already has. A metered irrigation record displaces both.","source":null,"reasoning":"NOT VERIFIED AGAINST A SOURCE. Plant factors are published, by species and by hydrozone, in the WUCOLS listing and in the landscape water budget appendices of several state model ordinances, and no page of any of them was read in setting this value. 0.50 is the figure commonly used for a mixed moderate water landscape and is recorded as unverified rather than given a citation. It is a single number standing for a plan that usually has three or four hydrozones on it, which is a larger error than the value's own uncertainty.","overridable":true},{"key":"estimate.irrigation_efficiency","quantity":"Irrigation application efficiency in a landscape water budget","value":0.75,"unit":"dimensionless","class":"C4","confidence":"Low","worth":"It divides the demand, so it moves it the other way and by less over its plausible range. On the same 5,000 sq ft in July the demand is 58.2 cu ft per day at this value, 72.8 at 0.60 and 48.5 at 0.90, a factor of 1.50 end to end against the plant factor's 2.67. Taken together the two span 29.1 to 116.5 cu ft per day, a factor of 4.00, which is the honest width of an estimated irrigation demand and is wider than either entry suggests alone.","displaced_by":"A distribution uniformity test on the system as installed, which is the measurement this number stands for and which irrigation auditors run as a matter of course. A drip system designed and certified to a stated efficiency displaces it directly.","source":null,"reasoning":"NOT VERIFIED AGAINST A SOURCE. Application efficiency is published in irrigation design texts and in the same state model ordinances, commonly 0.70 to 0.75 for spray and 0.81 or better for drip, and no page was read here. 0.75 is the spray figure in common use. It is optimistic in a known direction: it describes a system that was commissioned correctly and has not drifted, and a real efficiency falls with every broken head.","overridable":true},{"key":"estimate.irrigation_season","quantity":"Months in which an estimated irrigation demand is non-zero","value":7.0,"unit":"months","class":"C5","confidence":"Low","worth":"It is the whole of the cistern credit in the five months it zeroes. A landscape water budget computed from the reference evapotranspiration alone returns a positive demand in every month, because the reference rate is positive in every month: at the reference site on 5,000 sq ft it returns 14.8 cu ft per day in January. Carrying that figure would draw 59.2 cu ft over the median 4 day dry spell, which is 29.5 percent of a 1,500 gallon tank, and would report a tank as 70.5 percent full when it is in fact brim full because nobody irrigates a dormant landscape in Virginia in January. The registry entry practice.rainwater_demand_refused prices exactly this case and says the credit is either the whole tank or none of it.","displaced_by":"The irrigation controller's own schedule, which the operator has, or a metered monthly draw. A jurisdiction that publishes a landscape water budget with a stated season supersedes.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS for a national irrigation season, and the quantity is not a climate statistic: it is when somebody turns a controller on. April through October is the ordinary mid-Atlantic season and it is wrong for most of the country in one direction or the other, longer in the south and west and shorter in the north. It is shipped rather than left out because the alternative is a demand schedule with a positive January in it, which is the specific error the cistern refusal exists to prevent, and a demand that is wrong about a month is more visible than a demand that is wrong about a quantity. The season is a parameter on the estimator as well as an entry here.","overridable":true},{"key":"estimate.flow_length_from_boundary","quantity":"Hydraulic flow length taken as the longest chord of the drawn boundary","value":1.0,"unit":"policy flag","class":"C5","confidence":"Low","worth":"Less than it looks, and that is the finding. On the reference site the longest chord is 590.3 ft and the side of a square of the same area is 417.4 ft, a factor of 1.41, and the time of concentration over short grass is 0.1484 hr against 0.1355 hr, which is 9.6 percent, because the sheet flow term is capped at 100 ft and does not move at all. The peak discharge moves 1.8 percent, 7.634 against 7.768 cfs. The two judgments beside it in this module, the sheet flow cap and the surface, are worth 96 percent and a factor of 10.8 of the same quantity. A user who spends their attention on the flow length has spent it on the smallest of the three.","displaced_by":"The flow path drawn on the grading plan, measured from the hydraulically most distant point to the design point, which is what the method actually asks for and what no dataset contains. A LiDAR flow accumulation routing over the graded surface displaces it too and is not implemented.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS, because the quantity is not in any dataset: the elevation reader opens by saying so. The longest chord is a real geometric quantity, exactly computable, and it is an UPPER bound on any straight path across the shape, so the time of concentration it gives is an upper bound and the peak discharge is at the low end, which is not conservative for a pipe. It is preferred anyway, for two reasons. The hydraulic length is conventionally the longest flow path and not an average one, so the upper bound is the quantity being asked for. And the error it makes is the one that gets caught: an estimate that lands a site just above the applicability floor is invisible, and one that lands it below is refused out loud. Both bounds are reported on every estimate so the width is visible rather than implied.","overridable":true},{"key":"estimate.sheet_flow_length_cap_ft","quantity":"Length above which sheet flow is treated as concentrated","value":100.0,"unit":"ft","class":"C6","confidence":"Low","worth":"The largest of the three judgments in an estimated time of concentration on an ordinary site. On the reference site over short grass, the whole 590.3 ft flow path gives 0.1484 hr at this cap and 0.2908 hr at 300 ft, which is 95.9 percent more, and a peak discharge of 7.634 against 6.349 cfs, 16.8 percent lower. The cap bites on every site whose flow path exceeds it, which is every site, because sheet flow is far slower than shallow concentrated flow over the same ground: 100 ft of sheet flow over short grass at 5.29 percent takes 0.1117 hr and the remaining 490.3 ft as shallow concentrated flow takes 0.0367 hr.","displaced_by":"The governing jurisdiction's own sheet flow limit, which supersedes and which many manuals publish. Failing that, the point on the grading plan where the designer says flow concentrates, which is the real answer and is a drawn thing.","source":null,"reasoning":"REASONED, NOT MEASURED. TR-55 is recalled by this transcriber as printing a 300 ft maximum on the sheet flow equation in Chapter 3, and that recollection has NOT been checked against the document; NRCS guidance and most state manuals that came after it cap sheet flow at 100 ft, and the reason given is that sheet flow does not survive 300 ft on real ground. No Source is attached to this entry for that reason: shipping 100 ft under a TR-55 citation would cite a document for a number it does not print, and shipping 300 ft on an unchecked recollection would put the less conservative of the two behind every peak discharge in the product. 100 ft is the shorter, gives the shorter time of concentration and the higher peak, and is recorded as a product policy line rather than as a reading.","overridable":true},{"key":"estimate.flow_surface_from_land_cover","quantity":"Sheet flow roughness and shallow concentrated flow surface, inferred from the land cover class","value":1.0,"unit":"cover mapping","class":"C5","confidence":"Low","worth":"A factor of 10.8 on the time of concentration and the applicability of the method itself. On the reference site the same 590.3 ft flow path gives 0.0429 hr on smooth paved surfaces at n = 0.011, 0.1484 hr on short grass at n = 0.15, 0.1994 hr on dense grasses at n = 0.24 and 0.4630 hr in woods at n = 0.80. The first of those is below the 0.1 hr floor peak.min_tc_hours sets, so on a paved site the peak discharge is REFUSED and on a grassed one it is 7.634 cfs, and the land cover class is what decides which. Among the three that compute, the peak runs 7.634 to 5.015 cfs, a factor of 1.52.","displaced_by":"The surface named by the designer, which is the ordinary answer and takes one click, or a high resolution land cover product that separates mown turf from tree canopy from pavement. The sheet flow roughness table this reads is a published table and is not what is being defaulted here: the mapping from a class code to a row of it is.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS for the mapping, on exactly the argument the crosswalk.* entries make about cover types: nobody publishes a national crosswalk from a land cover class to a sheet flow roughness, and one NLCD class covers surfaces that sit in different rows of the table. Class 71 covers a mown lawn at n = 0.15 and an unmanaged old field at n = 0.24, and the developed classes cover pavement at 0.011 and the lawn beside it at 0.15 in the same pixel. The estimator maps each class to the row for its dominant surface and records this entry, and it takes the PERVIOUS row wherever the class is mixed, which is the longer time of concentration and the lower peak: that is the less conservative choice and it is made because the alternative, reading a developed class as pavement, refuses a peak discharge on most suburban sites through the floor above. Both are wrong and the estimator reports the span rather than hiding inside it.","overridable":true},{"key":"estimate.temporal_curve_selection","quantity":"Which of the NOAA Atlas 14 quartile cases and probability of occurrence lines a design storm shape is taken from","value":1.0,"unit":"curve selection","class":"C5","confidence":"Low","worth":"The largest single judgment in this module, and larger than the choice between NRCS types that the rainfall distribution tables refuses to make for the user. The Volume 2 general area 24 hour file that covers the reference site publishes 45 curves, four quartile cases and a combined case at nine cumulative probability of occurrence lines each, and their peak one hour fraction runs from 0.0542 on the combined case 60 percent line to 0.2766 on the first quartile 10 percent line, a factor of 5.10. The peak discharge from the 3.14 in storm over four acres at CN 80 and a time of concentration of 0.1484 hr runs 0.445 to 1.843 cfs across them, a factor of 4.14. Against NRCS Type II, which is what this product uses when it has no Atlas 14 curve, the same case gives 7.634 cfs: the span from the flattest Atlas 14 line to the NRCS curve is a factor of 17.14 on the peak discharge and 8.37 on the peak one hour fraction. The runoff VOLUME is 19,688 cu ft on every one of them, to the cubic foot, because volume does not depend on shape.","displaced_by":"The quartile case and probability line the governing jurisdiction names, which supersedes outright. Failing that, NOAA's own guidance on which case a design should use, which is published alongside the files and which this product has not read. A local gauge record of the storms that actually produced the design flows displaces the whole question.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS in this product for the selection, which is the finding rather than a gap: the curves are published and the rule for choosing among them is not held here. The estimator selects the FIRST QUARTILE 10 PERCENT line, which is the peakiest the file publishes, and the reason is the direction of the error rather than a belief about storms. Every other choice moves the peak discharge DOWN against the NRCS curve the product already ships, by up to a factor of 17, and an estimator that silently lowered a peak discharge by a factor of seventeen while upgrading its citation from a memory to a hashed file would be the exact failure this registry exists to prevent. Even the peakiest line is 39.0 percent below NRCS Type II on the peak one hour fraction and gives a peak discharge 4.14 times lower, and most of that gap is not a statement about storms either: Volumes 1 to 3 tabulate against twelfths of the duration, so a 24 hour curve has an ordinate every two hours and cannot express any intensity block shorter than that. The estimator therefore reports the curve as unusable for a peak discharge whose time of concentration is short against the tabulation interval, and usable for everything else, rather than preferring or refusing it wholesale.","overridable":true},{"key":"estimate.credit_geometry_bound","quantity":"Which geometric bound is offered for a jurisdictional credit measurement","value":1.0,"unit":"policy flag","class":"C5","confidence":"Low","worth":"The credit, which for a disconnection of a 1,500 sq ft roof is between 0.0 and 125.0 cu ft of a 1,815.0 cu ft requirement, as practice.jurisdictional_credit_refused already prices. What this entry decides is which way the estimate errs when it cannot know. On the reference square the longest chord is 590.3 ft and the narrowest crossing is 417.4 ft, so against a manual capping the contributing flow path at 500 ft the chord denies the credit and the width grants it, and the true flow path is between them and is in no dataset. The bound is chosen so that an estimate can only ever deny a credit the real measurement would grant, never grant one it would deny.","displaced_by":"The measurement off the plan, which is what a manual asks for and what a reviewer checks. Every one of these fields is a dimension somebody draws, and the estimate exists to fill the box before they do, not to stand in for them.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS, and the choice is a policy about the direction of error rather than a measurement. the local rules calculation already refuses a credit whose condition reads a field left None, so an unfilled field is a denial; filling it with the wrong bound is a GRANT, which is worse, because a denied credit is argued and a granted one is stamped. So a maximum condition gets the upper bound and a minimum condition gets the lower bound. Slope is the exception and is reported as the terrain measurement rather than bounded, because CreditArea carries one slope field and the two slope conditions pull opposite ways: the conservative figure for a maximum slope condition is the high percentile and for a minimum slope condition it is the low one, and no single number is conservative for both. The estimator supplies the median, names the P10 and P90 beside it, and says which condition each would decide. Two of the five measurements are not offered at all, because a receiving flow path length and a filter strip width are properties of a practice that has not been drawn yet, and a bounding dimension of the site is not either of them.","overridable":true},{"key":"classify.impervious_from_imagery","quantity":"That an impervious fraction read off aerial imagery is a classification and not a measurement","value":1.0,"unit":"policy flag","class":"C5","confidence":"Low","worth":"The whole of the difference between a number and a guess about a photograph. On the recorded Fairfax site the classification returns an impervious fraction of 0.6693 where the NLCD fractional impervious raster, which is a published measurement, returns 0.6553: 1.40 points apart, 0.41 curve number units on group B with the pervious remainder as open space in fair condition, and 1.81 percent of the runoff depth at P = 3 in. That agreement is the best case. The band this same photograph supports once the shadowed area is allowed to be anything runs 0.5746 to 0.7161, which is CN 85.66 to 89.77 and 1.6372 in against 1.9642 in at P = 3 in, 19.97 percent more on the smaller, and 0.2949 in against 0.4473 in at P = 1.2 in, 51.70 percent more. The classified figure is therefore one choice inside a band four curve number units wide, and it is DEFAULTED against this entry rather than DERIVED from the imagery it was computed from.","displaced_by":"A planimetered impervious area off a site survey or a stamped as-built, which supersedes outright. Failing that, the engineer relabelling segments against the photograph, which is recorded as an override with who and why. The published 30 m fractional impervious raster does not displace it and is not displaced by it: the two are independent readings and the product reports both with the difference between them.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. Nobody publishes the classification this module performs, and the imagery it reads is a photograph rather than a map: it records how much light of four wavelengths left the ground on one day, not what the ground is. The provenance rule follows from that and is the point of the entry. An impervious fraction of 0.43 read out of the NLCD fractional impervious raster is DERIVED and cites MRLC; an impervious fraction of 0.43 produced here is DEFAULTED and cites this entry; and the second must read differently on the face of a report, because a 0.6 m number carries an authority a 30 m number does not and that authority is exactly what must not be extended to a guess by accident. One more thing belongs on the face of this entry: the classification has been checked against an independent measurement on exactly ONE site, the recorded Fairfax square, which is a suburban office park flown in October. Every figure quoted anywhere in the classify.* entries is from that site. A method validated at n = 1 is not validated; it is one agreement, and it is reported as one.","overridable":true},{"key":"classify.class_set","quantity":"Which classes an imagery classification resolves the site into","value":5.0,"unit":"classes","class":"C5","confidence":"Low","worth":"What the set leaves unresolved, which is the split between mown grass and tree canopy inside the vegetated class. TR-55 puts woods in fair condition at 60 on group B and open space in fair condition at 69. On the recorded Fairfax site, where the classification is 28.39 percent vegetation and 0.6693 impervious, reading the whole vegetated remainder as woods rather than open space is composite CN 85.43 against 88.41, 2.98 units, and 1.6201 in against 1.8513 in at P = 3 in, which is 14.27 percent more on the smaller, and 0.2878 in against 0.3911 in at P = 1.2 in, 35.90 percent more.","displaced_by":"A canopy height model: 3DEP LiDAR first return minus bare earth, or a published canopy layer, either of which answers the height question the photograph cannot. A site visit, or the engineer relabelling a segment, displaces it for that segment.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS for the choice, which is a decision about what this product will claim to see. The five classes are the branches a curve number computation actually takes: CN 98 for sealed surface, the open space family for pervious vegetation, the newly graded row for bare ground, the refusal for open water, and an explicit undetermined class for shadow. A general land cover legend would add classes this method does not distinguish and would still not answer the one question it needs. Lawn against canopy is left unresolved rather than guessed because a photograph taken from overhead has no height in it, and the texture cues that would separate most of them fail on the two cases that matter, a single mature tree over a lawn and an unmown field.","overridable":true},{"key":"classify.vegetation_break","quantity":"Whether the NDVI break between vegetation and everything else is computed from the scene or fixed","value":1.0,"unit":"policy flag","class":"C5","confidence":"Low","worth":"23.88 points of impervious fraction on the recorded site, which is the largest single figure anywhere in this module. NAIP is delivered as uncalibrated 8 bit digital numbers, so the conventional fixed NDVI threshold of 0.2 is a statement about the scene it was tuned on. The recorded Fairfax flight is 2023-10-11, an October flight over a suburban office park whose street trees have turned: only 9.19 percent of the site clears NDVI 0.2, so the fixed threshold classifies the site 0.9081 impervious. The break computed from this scene's own histogram falls at NDVI 0.0487 and gives 0.6693, and the independent NLCD measurement over the same ground says 0.6553. Between the two: CN 95.34 against 88.41 on group B, 6.93 units, and 2.4835 in against 1.8513 in at P = 3 in, 34.15 percent more, and 0.7633 in against 0.3911 in at P = 1.2 in, 95.18 percent more.","displaced_by":"A radiometrically calibrated product, surface reflectance rather than digital numbers, which would make a fixed break meaningful and portable. Failing that, a scene with published vicarious calibration targets, which NAIP does not carry.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS for a portable threshold, because there is nothing to be portable across: the values are 8 bit counts with no radiometric calibration, and the same ground flown on a different day by a different vendor returns different numbers. The break is therefore computed from the scene in front of the classifier by Otsu's method (Otsu 1979), which is a published procedure rather than a number this product chose. Its own failure mode is that it assumes two modes, and it is not hypothetical: on a synthetic site that is entirely lawn the break lands at NDVI +0.48, inside the single vegetation mode, and more than half of a lawn classifies as impervious. Nothing in the image catches that. Otsu's own separability measure does not, being 0.757 on that scene against 0.724 on the real mixed site, so a threshold on it would reject the good site and accept the bad one, and none is implemented for that reason. What catches it is the independent measurement: the cross check against the published raster fires at five times its tolerance, which is the argument for that comparison being part of the method rather than a diagnostic.","overridable":true},{"key":"classify.cluster_count","quantity":"Number of k-means clusters an imagery classification fits","value":6.0,"unit":"clusters","class":"C6","confidence":"Low","worth":"Mostly whether shadow gets a cluster at all, which is worth more than where the cluster boundaries land. On the recorded Fairfax site k = 4 produces no cluster below the shadow brightness break: the answer is a confident 0.6967 impervious with no band at all, 0.79 curve number units above the k = 6 figure and, far worse, with nothing on it to say that a seventh of the site is ground the photograph does not show. Across k = 5 to 8, where shadow is resolved, the impervious fraction runs 0.6255 to 0.6693, 4.38 points, CN 87.14 to 88.41, 1.27 units, and 1.7500 in to 1.8513 in at P = 3 in, 5.79 percent more on the smaller. The eight-cluster result sets the lower bound; the former 0.6418 bound was the seven-cluster result.","displaced_by":"A model selection criterion computed on the scene rather than a fixed count: a gap statistic or a silhouette scan across k, which is a real improvement and is not implemented here. Until then a reviewer can pass a different k and see the whole span, which is what the figures above are.","source":null,"reasoning":"REASONED, NOT MEASURED. Six is the smallest count that, on the recorded site, separates the surfaces a curve number cares about and still leaves a cluster for shade: sunlit pavement splits into two brightness modes, vegetation into a lit and a shaded one, and shadow takes the sixth. Below five the shade merges into pavement and the classification stops being able to say that it does not know. Above eight the clusters start splitting on illumination within one surface, which adds classes with nothing behind them to label.","overridable":true},{"key":"classify.shadow_break","quantity":"Brightness below which a cluster is called shadow, as a fraction of the scene's own sunlit reference","value":0.5,"unit":"fraction of the 95th percentile visible brightness","class":"C5","confidence":"Low","worth":"The size of the undetermined area, which on the recorded Fairfax site is 14.15 percent and is the width of the whole uncertainty band: impervious 0.5746 if all of it is pervious against 0.7161 if all of it is impervious, CN 85.66 against 89.77 on group B, 4.10 units, and 1.6372 in against 1.9642 in at P = 3 in, 19.97 percent more on the smaller, rising to 51.70 percent at P = 1.2 in. Leaving shadow undetected does not remove that uncertainty, it spends it: the dark pixels then classify as pavement, which is the 0.7161 end reported as a fact.","displaced_by":"A sun angle and a building height model, which together predict where a shadow must fall and let the surface under it be inferred rather than guessed. The acquisition time is published per scene and the height model is not held by this product. A leaf off or high sun flight of the same ground displaces the question rather than answering it.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. A half is a physical argument rather than a calibration: a surface in open shade is lit by skylight alone, which in the visible is a fifth to a half of global irradiance, so a cluster whose centroid sits above half the scene's sunlit reference is not in shade whatever else it is. It is expressed as a fraction of the scene's own 95th percentile brightness rather than as a digital number, because a fixed DN on an uncalibrated 8 bit image is not portable between flights; the 95th percentile rather than the maximum because a maximum is a specular highlight off a windscreen and is 255 on every scene ever flown. The corroboration is reported and not acted on: on the recorded site the dark cluster sits 0.200 below the sunlit impervious clusters on (red - blue) / (red + blue), which is what blue skylight does to a shaded surface and is evidence, not proof, that it is shade rather than dark roofing.","overridable":true},{"key":"classify.shadow_allocation","quantity":"How area whose surface the photograph does not show is allocated in the single reported impervious fraction","value":1.0,"unit":"policy flag","class":"C5","confidence":"Low","worth":"The choice is worth the whole band the shadow break prices: on the recorded Fairfax site, allocating the 14.15 percent undetermined area entirely to pervious gives 0.5746 and entirely to impervious gives 0.7161, CN 85.66 against 89.77 on group B and 1.6372 in against 1.9642 in at P = 3 in, 19.97 percent more on the smaller. Pro rata gives 0.6693, which is 0.41 units above what the independent NLCD measurement of the same ground implies and 1.36 units below the all impervious end.","displaced_by":"The engineer relabelling the shadow segments against the photograph, which is the cheapest real evidence available and is recorded as an override. A height model or a second flight at a different sun angle displaces it outright.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. Pro rata assumes shadow falls on the site's surfaces in the proportion they occur, which is known to be wrong in a specific direction: shadows are cast by buildings and trees, so they fall disproportionately on the pavement beside a building and the grass under a tree rather than uniformly. It is chosen because the alternatives are worse in a way that hides: allocating shadow to impervious is the conservative choice and would silently return the top of the band as a point estimate, and allocating it to pervious understates runoff on exactly the dense sites this product is for. Whichever end a reviewer prefers is printed beside the figure every time, which is the actual defense.","overridable":true},{"key":"classify.bare_soil_redness","quantity":"Red to blue ratio above which an unvegetated cluster is called bare soil rather than pavement","value":0.1,"unit":"(red - blue) / (red + blue)","class":"C5","confidence":"Low","worth":"12 curve number units on group B over whatever area it decides: the newly graded row is 86 and connected impervious is 98. Over a tenth of a site that is 1.2 units of composite, and it runs both ways, because a cluster of bright weathered concrete that tripped the test would be read as pervious ground at 86 where 98 belongs. On the recorded Fairfax site the test did not fire at all: the sunlit impervious clusters sit at +0.019 and +0.037 and the shaded cluster at -0.189, none of them within 0.06 of the break, and the one cluster above it at +0.103 is vegetation and is claimed by the vegetation test first. So the break has never decided anything on recorded data and there is no evidence here that it works.","displaced_by":"A site visit or a construction sequence, either of which says whether there is graded ground on the site at all, and the engineer relabelling the segment. A calibrated surface reflectance product would let a published soil line be used instead of a ratio.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS, and unlike the vegetation break this one is FIXED rather than computed from the scene, which is a weakness stated rather than hidden. There is nothing in a scene to derive it from: a site with no bare soil has no second mode for a histogram method to find, so a scene derived break would manufacture one and classify pavement as graded ground. The value rests on the color of iron stained mineral soil against the near neutral color of concrete and asphalt, and it is an assertion about the eastern United States that has not been checked anywhere else. The classification tests vegetation before soil, because senescent autumn foliage is red shifted and would otherwise satisfy this test: on the recorded site that ordering keeps 18.10 percent of the area in vegetation rather than moving it to newly graded ground.","overridable":true},{"key":"classify.water_test","quantity":"Near infrared ceiling, as a fraction of the scene's sunlit reference, for calling a cluster open water","value":0.25,"unit":"fraction of the 95th percentile visible brightness","class":"C5","confidence":"Low","worth":"Whether a pond is refused or given a curve number, which crosswalk.water_as_storage already prices: on a site otherwise at CN 69 on group B, a pond over a tenth of the area read as CN 98 moves the answer from 0.670 to 0.880 in at P = 3 in and from 0.019 to 0.116 in at P = 1.2 in, and the error lands on precisely the sites that have a pond. The test did not fire on the recorded Fairfax site, which has no water on it: the closest cluster reaches NDWI +0.133 against the 0.15 break, and its near infrared is 123.6 against a ceiling of 53.25.","displaced_by":"The National Hydrography Dataset, or a jurisdictional pond inventory, either of which says where water is without asking a photograph. The engineer relabelling the segment displaces it for that segment.","source":null,"reasoning":"NO EXTERNAL SOURCE EXISTS. Water absorbs near infrared almost completely, so a cluster that is bright in green and dark in the near infrared is the one spectral signature in this module with a clean physical basis. It is still a weak test on an uncalibrated image, because deep shadow is also dark in the near infrared and blue shifted, and the two are separated here only by how far the near infrared falls. The direction of the error is the reason the test is worth having at that confidence: a cluster wrongly called water is carried out of the computation as a REFUSAL a user can see and argue with, rather than given a curve number that gets stamped. On the recorded site NDWI alone would have called 23.94 percent of a parking lot water, which is why the near infrared ceiling is part of the test rather than NDWI on its own.","overridable":true},{"key":"classify.min_segment_area_sqm","quantity":"Smallest classified region that becomes a segment of its own","value":25.0,"unit":"sq m","class":"C6","confidence":"Low","worth":"On the recorded Fairfax site the raw classification has 409 connected regions with a median size of 2 pixels, which is speckle rather than a map. At 25 sq m, which is 70 pixels at this 0.6 m resolution, 37 regions survive and 2,142 pixels, 4.78 percent of the site and 773.6 sq m, are absorbed into the neighbor they share the most boundary with. That absorption moves the impervious share of the determined area from 0.6693 to 0.6782, which is CN 88.41 to 88.67 on group B, 0.26 units, and 1.14 percent of the runoff depth at P = 3 in. The threshold is what a user can edit, not what the site is: the shares on the classification are computed before any of this.","displaced_by":"Nothing displaces it with evidence. It is a decision about what a person will draw as a separate management unit, and it is a parameter on segments_from_classification as well as an entry here. A reviewer who wants every speckle kept can pass zero.","source":null,"reasoning":"REASONED, NOT MEASURED. 25 sq m is about two parking spaces or a single car garage, which is the smallest patch of a site anybody drains separately or draws as its own polygon. It is stated as an AREA and not as a pixel count on purpose: four pixels is 1.4 sq m at the 0.6 m NAIP is usually flown at and 0.36 sq m over the 0.3 m imagery some states are flown at, so a threshold in pixels would mean four different things on four different sites.","overridable":true},{"key":"classify.simplify_tolerance","quantity":"Douglas-Peucker tolerance on a classified boundary, in pixels","value":0.5,"unit":"pixels","class":"C6","confidence":"Moderate","worth":"Vertices against area, and at half a pixel it is free. On the recorded Fairfax site the 37 segments come back with 4,281 vertices instead of 9,580 and not one segment's area changes by a square meter: at 0.30 m on a 0.6 m grid the only vertices Douglas-Peucker removes are the ones lying on a straight run, whose perpendicular distance from the chord is zero. The trade starts at a whole pixel, where the same site simplifies to 1,681 vertices and 76.4 sq m of area moves between neighboring segments, 0.47 percent of the sampled area, with 11.0 sq m lost off the outer staircase where there is no neighbor to move it to. The polygons tile the site at every tolerance: each shared chain is simplified once and reused by both regions, which is why the area that leaves one segment arrives in the next rather than becoming an overlap or a gap.","displaced_by":"Nothing displaces it with evidence. It is a decision about how many vertices a person should have to drag, and it is a parameter as well as an entry. Passing zero returns the exact pixel boundary and passing a whole pixel buys the reduction priced above.","source":null,"reasoning":"REASONED, NOT MEASURED. Half a pixel is the largest tolerance that cannot move a boundary at all on a cell grid, because every corner where a pixel boundary actually turns is at least a pixel's half diagonal, 0.42 m here, from the chord across it. So the default is the lossless one: it halves the vertex count by dropping points that carry no information and leaves the areas exactly as the pixels counted them, which matters because area is the deliverable and every practice volume on the site scales linearly with it. It is expressed in pixels rather than meters for the reason the minimum segment area is expressed in meters: the tolerance should follow the resolution of the thing being simplified, and a fixed 0.3 m would be a whole pixel on the 0.3 m imagery some states are flown at.","overridable":true},{"key":"classify.cross_check_tolerance","quantity":"Difference between the classified and the published impervious fraction above which the two are reported as disagreeing","value":0.1,"unit":"fraction","class":"C6","confidence":"Low","worth":"10 points of impervious fraction is 2.90 curve number units on group B with the pervious remainder as open space in fair condition, 13.43 percent of the runoff depth at P = 3 in and 33.10 percent at P = 1.2 in. That is what the product is prepared to call agreement. On the recorded Fairfax site the two differ by 0.0140, well inside it, and the classification and the published raster reach CN 88.41 and 88.00. The threshold cannot usefully be much tighter: the published figure over that site rests on 19 pixels of a 30 m raster, whose sampling standard error at the worst share is sqrt(0.25/19) = 0.1147, wider than the tolerance itself, so a difference below about 11 points cannot distinguish a classification error from noise in the thing it is being compared against. The result says so when that is the case.","displaced_by":"A planimetered impervious area from a survey, which ends the comparison by making one side a measurement of the site rather than of a neighborhood. A high resolution published land cover product over the same ground, such as the 1 m Chesapeake Conservancy coverage this engine already knows how to fetch, would be a third reading and a better second one.","source":null,"reasoning":"REASONED, NOT MEASURED. The two numbers being compared are not the same measurement taken twice: one is a classification of a 0.6 m photograph flown on a named day, the other a published 30 m product for a calendar year, and on a four acre site the second is a neighborhood average. So the threshold is not an error bar, it is the point at which the disagreement is too large to be explained by that difference in kind and one of them must be wrong about the site. Neither is preferred when it fires: a finer number that is a guess and a coarser number that is a measurement fail in different directions, and the product reports both with the signed difference rather than choosing.","overridable":true},{"key":"assistant.confidence_tier_binding","quantity":"Confidence terms from a higher fidelity tier's row that one response may contain","value":0.0,"unit":"terms","class":"C6","confidence":"Low","worth":"What keeps the assistant honest about coarse data. A result from the coarse global data, described as well constrained, reads to a user exactly like a result from the detailed United States surveys, and the two differ by the resolution of every contributing layer. Zero tolerance is what makes the vocabulary a property of the data rather than of the sentence: at any other threshold the model chooses, and the model's impression of the data is the thing the CurveNumber assistant design Section 6.6 exists to take out of the loop.","displaced_by":"A measured relationship between the qualitative term a reader sees and the interval they infer from it. If readers turn out to take 'reasonably constrained' as a wider range than the middle grade of data actually carries, the table is the thing to change rather than this threshold.","source":null,"reasoning":"REASONED, NOT MEASURED. NO EXTERNAL SOURCE EXISTS for how many words of overclaim a reader tolerates. Zero is chosen because the alternative is a budget, and a budget for overstating certainty is spent on the sentence the reader remembers.","overridable":true},{"key":"assistant.citation_coverage_target","quantity":"Share of C1 to C4 claims in a response that must carry a citation marker","value":0.99,"unit":"fraction","class":"C6","confidence":"Low","worth":"One uncited sentence in a hundred, measured over the evaluation set of the CurveNumber assistant design Section 9.1. The figure is a target rather than a gate: coverage below 100 percent is advisory and a regulatory claim with no marker at all is blocking whatever the average says, because the average of a response containing one fabricated requirement and ninety nine cited sentences is 0.99 and the response is still unusable.","displaced_by":"A scored sample in which a reviewer judges whether each marker supports the sentence it is attached to. Coverage counts markers and cannot read them, and a measured support rate would replace this number with a criterion rather than raise it.","source":null,"reasoning":"REASONED, NOT MEASURED. Set below 100 percent because a target nobody meets is a target nobody reads, and a response is allowed one sentence of connective prose that a lexical classifier calls a claim. The gate that actually protects a reader is the blocking one on a C3 sentence, and it has no budget at all.","overridable":true},{"key":"assistant.interpretation_label_threshold","quantity":"Inference sentences a response may carry without an Interpretation label","value":0.0,"unit":"sentences","class":"C6","confidence":"Low","worth":"The reader's ability to tell which sentences to disbelieve first. An unlabeled inference sitting between two cited computations inherits their authority, and the inference is the one sentence in the response that nothing checked. This is also the only defense the guard has against a correct number supporting a wrong claim, which no deterministic check catches.","displaced_by":"A presentation change rather than a different number: if inferences were rendered in their own visual register the label would be redundant. the CurveNumber assistant design Section 5.3 rule 2 asks for visual distinguishability and names the label as one way of getting it.","source":null,"reasoning":"REASONED, NOT MEASURED. NO EXTERNAL SOURCE EXISTS for an acceptable rate of unlabeled judgment. Zero is chosen because any other figure requires deciding which inferences do not matter, and the ones that do not matter are not the ones that get written.","overridable":true},{"key":"assistant.disclaimer_placements","quantity":"Placements in a conversation at which a disclaimer is required","value":3.0,"unit":"placements","class":"C6","confidence":"Low","worth":"The three conversational rows of the CurveNumber assistant design Section 7.3: the full statement once on the first substantive response, the short form on every response that sizes a practice, and the short form on every response containing an adequacy comparison. The other four rows of that table are report and interface placements and are not this layer's to enforce. the CurveNumber validation plan Section 6.4 states the reason the count is not smaller: a disclaimer that appears only after the reader has seen and acted on the number provides no protection and no honesty.","displaced_by":"Evidence about whether a repeated disclaimer is read or skipped. If repetition measurably reduces attention to it, the answer is a different presentation rather than fewer placements, and the non dismissible micro form beside every displayed curve number is the shape that argument would take.","source":null,"reasoning":"REASONED, NOT MEASURED. NO EXTERNAL SOURCE EXISTS for how often a professional boundary statement has to be repeated to survive paraphrase and export. Three is the count of rows in the source table that describe an assistant response, so the number is read off the specification rather than chosen freely, and moving it means editing that table.","overridable":true},{"key":"assistant.sizing_margin","quantity":"Fraction by which a provided volume must clear a required volume before the margin is quantified rather than named","value":0.1,"unit":"fraction","class":"C6","confidence":"Low","worth":"Decides whether an unprompted message on the coarse global data quantifies the margin. On that data the soil grid is 250 m and the cover is coarse, and a provided volume clearing a required volume by less than this is not a margin, it is the resolution of the inputs. the CurveNumber validation plan Section 4.1 puts the impervious fraction alone at plus or minus 10 percentage points, which is roughly 4 to 6 curve number units in the medium density residential range, and its Section 4.3 shows CN 75 plus or minus 5 spanning 0.084 to 0.286 in at P = 1.5 in against a central 0.167, which is minus 50 to plus 71 percent. A ten percent margin sits well inside that span, so quoting it as a margin would report the noise as headroom.","displaced_by":"Expert panel calibration, per the CurveNumber assistant design open question 4. A measured relationship between input tier and delivered volume error would replace this with a figure per tier rather than one figure.","source":null,"reasoning":"REASONED, NOT MEASURED. Proposed default per the CurveNumber assistant design open question 4, which names it as the 10 percent sizing margin. A tenth is the round number at which the margin and the input resolution are the same size; it is not calibrated against anything and it is where a reviewer should start arguing.","overridable":true},{"key":"assistant.impervious_tolerance","quantity":"Difference between a segment's mapped impervious fraction and the largest fraction consistent with its own curve number above which the two are reported as describing different surfaces","value":0.1,"unit":"fraction","class":"C6","confidence":"Low","worth":"Percentage points, not percent: the comparison is between two fractions. Decides when a segment's curve number and its mapped imperviousness are reported as describing different surfaces. Ten points is 2.90 curve number units on group B with the pervious remainder as open space in fair condition, since the impervious row is 98 and that cell is 69, and 2.90 units moves the runoff depth at P = 3 in from 0.6697 to 0.8032 in, 19.94 percent. It is the same figure and the same arithmetic as classify.cross_check_tolerance and deliberately so: that entry compares two published rasters, this one compares a curve number against a raster, and a product that called ten points a disagreement in one place and agreement in the other would be reporting the threshold rather than the site.","displaced_by":"Expert panel calibration, per the CurveNumber assistant design open question 4. Before that, a capability: back-solving the impervious fraction currently refuses because CurveNumber cannot invert a composite curve number, so the comparison is made against a one-directional bound rather than against a back solved fraction, and a true inversion would narrow what the tolerance is being applied to.","source":null,"reasoning":"REASONED, NOT MEASURED. Proposed default per the CurveNumber assistant design open question 4, which names it as the 10 percentage point imperviousness tolerance. The two numbers being compared are not one measurement taken twice: a curve number carries a cover type and a hydrologic condition as well as a surface, so a segment can differ from its mapped imperviousness for reasons that are not an error in either. The threshold is where that explanation stops covering the difference.","overridable":true},{"key":"assistant.landcover_staleness_years","quantity":"Age of a retrieval above which the layer it produced is reported as stale","value":5.0,"unit":"years","class":"C6","confidence":"Low","worth":"Which retrievals on a site are named as old enough that the ground may have moved under them. Applied to the RETRIEVAL DATE and not to the publisher's edition, because the CurveNumber validation plan Section 3.6 keeps stale, superseded and retracted apart and only the first is a fact this record carries: an edition is stale when the publisher has issued a newer one, which is a fact about the publisher, and a retracted edition is a notice this product has no feed for. The annual land cover product publishes one year per year, so five years of retrieval age is up to five published editions newer than the one the site stands on.","displaced_by":"Expert panel calibration, per the CurveNumber assistant design open question 4. Independently of that, a publisher edition feed would displace the whole shape of this test rather than its number: with one, staleness is checked against what the publisher has actually released, which is the definition the CurveNumber validation plan Section 3.6 gives, and retrieval age becomes a proxy nobody needs.","source":null,"reasoning":"REASONED, NOT MEASURED. Proposed default per the CurveNumber assistant design open question 4, which names it as the 5 year land cover staleness threshold. Doc 05 Section 8 row 4 states the example as land cover more than 5 years old WHERE THE SITE SHOWS IMAGERY CHANGE. This engine cannot see imagery change and does not pretend to, so the second clause is dropped and the message says what it is testing rather than implying the test it could not run.","overridable":true},{"key":"assistant.impervious_over_vegetation","quantity":"Impervious fraction declared over a polygon whose cover row is a vegetated one at or above which the two inputs are reported as incompatible","value":0.9,"unit":"fraction","class":"C6","confidence":"Low","worth":"the CurveNumber assistant design Section 8 row one's own figure: 90 percent impervious declared over a polygon mapped as forest. What it is worth is the whole segment. Woods in fair condition on group B is CN 60; the same polygon at 90 percent connected impervious composites to 94.2, and at P = 3 in the runoff depth goes from 0.3333 to 2.3697 in, a factor of 7.11. At P = 1.2 in the cover row produces no runoff at all and the declared imperviousness produces 0.6851 in, which is not a discrepancy in magnitude but a disagreement about whether the surface sheds water. One of the two inputs is wrong about this polygon and the product says so rather than compositing them.","displaced_by":"Expert panel calibration, per the CurveNumber assistant design open question 4. A measured distribution of impervious fraction within each land cover class would replace a single ceiling with a per class one, which is the right shape: 90 percent over forest and 90 percent over low intensity developed are not the same claim.","source":null,"reasoning":"REASONED, NOT MEASURED. Read off the specification's own worked example rather than chosen freely, which is why it is 0.90 and not a rounder or a tighter figure. It is a ceiling on a contradiction and not an error bar: a vegetated cover row and a near total impervious fraction do not describe one surface at any tolerance, and the threshold only decides how far the product lets the two drift before saying so.","overridable":true},{"key":"departure.band_normal","quantity":"Override departure at or below which the departure is within normal judgment and is recorded silently","value":3.0,"unit":"CN","class":"C6","confidence":"Low","worth":"The boundary between a departure the product records and one it raises. Three units is not small: from CN 69 it moves the runoff depth at P = 3 in from 0.6697 to 0.8081 in, 20.66 percent. What makes it silent is not that it is negligible but that a field judgment about hydrologic condition is worth 5 to 12 units on its own per the CurveNumber validation plan Section 4.1, so three units is inside the range a licensed engineer standing on the site is entitled to move without explaining themselves to a piece of software.","displaced_by":"Expert panel calibration, per the CurveNumber assistant design open question 4.","source":null,"reasoning":"REASONED, NOT MEASURED. Read off the table in the CurveNumber assistant design Section 3.4, whose first row is 0 to 3 CN units. The table is the specification and this entry is the place the number is answerable; moving it means editing that table.","overridable":false},{"key":"departure.band_notable","quantity":"Override departure at or below which the departure is notable and a justification is requested","value":8.0,"unit":"CN","class":"C6","confidence":"Low","worth":"Whether a justification is requested. Eight units from CN 69 moves the runoff depth at P = 3 in from 0.6697 to 1.0710 in, 59.93 percent, which is most of one hydrologic soil group step: the CurveNumber validation plan Section 4.1 puts one group step at 8 to 12 units for row crop and 5 to 9 for woods. A departure of this size is the size of having assumed a different soil, and the product asks why.","displaced_by":"Expert panel calibration, per the CurveNumber assistant design open question 4.","source":null,"reasoning":"REASONED, NOT MEASURED. Read off the table in the CurveNumber assistant design Section 3.4, whose second row is 4 to 8 CN units.","overridable":false},{"key":"departure.band_questionable","quantity":"Override departure at or below which the departure is questionable, and above which it is implausible","value":15.0,"unit":"CN","class":"C6","confidence":"Low","worth":"Two things at once, which is why it is one entry and not two. At or below it the departure is questionable: a justification is required before the value enters a report. Above it the departure is implausible and the product states the physical implication and asks whether the soil or the cover classification is instead wrong. Fifteen units from CN 69 moves the runoff depth at P = 3 in from 0.6697 to 1.5163 in, 126.42 percent, which is more than the whole antecedent runoff condition span the method admits between ARC II and ARC III. A departure that large is a claim about a different site.","displaced_by":"Expert panel calibration, per the CurveNumber assistant design open question 4.","source":null,"reasoning":"REASONED, NOT MEASURED. Read off the table in the CurveNumber assistant design Section 3.4, whose third row is 9 to 15 CN units and whose fourth is everything above 15. The fourth band has no threshold of its own, so this entry carries both boundaries and says so.","overridable":false},{"key":"observed.max_interpolated_gap_hours","quantity":"Longest gap in an observed series this engine will bridge by linear interpolation","value":1.0,"unit":"hr","class":"C6","confidence":"Low","worth":"Which missing intervals become interpolated samples carrying a flag, and which become a refusal under CN-OBS-003. It is worth the difference between a direct runoff volume with a straight line drawn through part of it and a refusal to report one. On a small flashy catchment an hour can span the whole rising limb, which is why the refusal CN-OBS-004 exists beside this entry and is structural: a gap inside a selected event is refused whatever its length, and this threshold governs only the gaps outside one.","displaced_by":"A published study of interpolation error in direct runoff volume as a function of gap length and catchment response time, which would make this a function of the record rather than one number. Failing that, a recession model fitted to the gauge's own record, which would let a gap on a recession be filled with something better than a straight line.","source":null,"reasoning":"REASONED, NOT MEASURED. An hour is the round number at which a straight line through a gap stops being obviously harmless on the between event parts of a record, where the discharge is baseflow and is changing slowly. It is not calibrated against anything and it is where a reviewer should start arguing.","overridable":true},{"key":"observed.max_qualified_fraction","quantity":"Fraction of an event's discharge samples that may carry a provisional or estimated quality flag before the event is refused for fitting","value":0.25,"unit":"fraction","class":"C6","confidence":"Low","worth":"Whether an event enters the fitting set. A provisional sample is a real measurement the agency has not yet reviewed and an estimated one is the agency's own model output, not a measurement at all, so the two are counted together here and reported separately on the record. What the threshold is worth is the difference between a fit that rests on reviewed gauge data and one that rests partly on somebody else's estimate, which is a distinction a reviewer of a stamped calculation is entitled to make for themselves.","displaced_by":"A comparison, on this gauge, of provisional against approved values for the same intervals after the agency's review. That comparison is publishable and this product has not made it.","source":null,"reasoning":"REASONED, NOT MEASURED. A quarter is the point at which the qualified samples stop being incidental. Nothing measures where that point is, and the honest alternatives were zero, which refuses most recent records because recent data is provisional by definition, and one, which is no threshold at all.","overridable":true},{"key":"calibration.min_site_fraction_of_gauge","quantity":"Smallest fraction of a gauge's contributing watershed a site may be before a curve number fitted at that gauge is refused as a description of the site","value":0.1,"unit":"fraction","class":"C6","confidence":"Low","worth":"The entire claim that a gauged fit describes a particular parcel. Below this the fitted figure is a property of the watershed, most of which is ground the site does not contain, and calling it the site's curve number is the substitution CN-GAU-003 refuses. On a two acre site inside a one square mile gauged watershed the fraction is 0.0031, so the ordinary commercial case is three orders of magnitude inside the refusal rather than near it.","displaced_by":"Nothing measured. What actually resolves the underlying question is a gauge on the site's own outfall, which makes the fraction 1.0 and makes this entry irrelevant, and that is stated in CN-GAU-003's resolves sentence rather than pretended away here.","source":null,"reasoning":"REASONED, NOT MEASURED. A tenth is a judgment about when a part stops describing a whole, and the true answer depends on how alike the site and the rest of the watershed are, which this engine cannot measure from a boundary and an area. The threshold is deliberately generous: it passes cases that are still not site calibration, and the refusal it guards is therefore a floor on the claim rather than a certificate of it.","overridable":true},{"key":"calibration.single_rain_gauge_area_sqmi","quantity":"Catchment area above which one rain gauge is refused as an areal rainfall record","value":10.0,"unit":"sq mi","class":"C6","confidence":"Low","worth":"Whether a fit may be attempted at all from one rain gauge. A curve number fitted to a point rainfall depth inherits the whole difference between that point and the mean depth over the catchment, and the fit absorbs the difference into the curve number, where it is indistinguishable from a property of the soil. The direction is not one sided, which is what makes it dangerous: a gauge that missed the cell reads low and the fitted curve number comes out high, and the reverse is equally common.","displaced_by":"A radar rainfall product or a gauge network with a published areal reduction, either of which replaces the point assumption with an areal estimate and makes this threshold unnecessary. This engine holds neither and holds no areal weighting method, which is CN-GAU-005.","source":null,"reasoning":"REASONED, NOT MEASURED. Ten square miles is a round number in the range where convective storm cells and catchments are the same size, which is the condition under which a point gauge stops representing an area. It is not read off a published areal reduction curve and it must not be quoted as though it were.","overridable":true},{"key":"calibration.max_parameter_correlation","quantity":"Magnitude of the correlation between two jointly fitted parameters above which the joint result is refused","value":0.95,"unit":"dimensionless","class":"C6","confidence":"Low","worth":"Whether a joint fit of the curve number and the initial abstraction ratio is reported at all. Above this the two parameters trade off along a ridge on which the objective is nearly flat, so the pair that comes back is a property of where the search started and of its tolerance rather than of the record, and the two numbers will be quoted separately by somebody. The volume the pair reproduces is still right; it is the split of it between the abstraction and the retention that is not determined.","displaced_by":"A larger and more varied event set, which is the only thing that separates these two parameters. Events spanning a wide range of rainfall depth, and in particular events near the abstraction threshold, carry nearly all of the information about lambda.","source":null,"reasoning":"REASONED, NOT MEASURED. The number is this product's opinion about where a correlation stops being high and starts being a ridge. There is no published threshold; 0.95 corresponds to the joint confidence region being about three times longer than it is wide, which is already well past the point at which either parameter can be quoted on its own.","overridable":true},{"key":"calibration.recession_return_fraction","quantity":"Fraction of the event's peak discharge above baseflow that the hydrograph must fall back to before the next event begins, for the direct runoff volume to be integrable","value":0.05,"unit":"fraction","class":"C6","confidence":"Low","worth":"Whether an event's direct runoff volume is reported or refused under CN-CAL-008. An event whose recession is cut off by the next storm has part of its volume in the next event's integration window and part of the next event's volume in its own, and the fitted curve number absorbs both errors. The error is one sided on the truncated event, which reads low, and one sided the other way on the event that follows.","displaced_by":"A recession model fitted to this gauge's own record, which would let the truncated limb be extrapolated rather than refused and would replace this fraction with the gauge's recession constant.","source":null,"reasoning":"REASONED, NOT MEASURED. Five percent of the event's own rise is the point at which what is left of the recession is small against the volume already integrated. It is a judgment about an acceptable truncation error and not a measurement of one.","overridable":true},{"key":"baseflow.lyne_hollick_alpha","quantity":"Filter parameter of the Lyne and Hollick recursive digital filter","value":0.925,"unit":"dimensionless","class":"C4","confidence":"Low","worth":"How much of an event hydrograph is called baseflow, and therefore the direct runoff depth every fitted curve number is fitted to. Raising alpha toward 1 makes the separated baseflow flatter and assigns more water to direct runoff, which raises the fitted curve number; lowering it does the reverse. The size of the choice is quantified on a record whose answer is known: the synthetic catchment in the worked calibration example is built at curve number 74 over a constant baseflow, the straight line convention returns 74.000 on its largest event and this filter returns 66.818, a difference of 7.18 curve number units. At a three inch storm that is 0.9082 in of runoff against 0.5776 in, 57.2 percent of the smaller figure and nearly this product's own notable override band of 8 CN. The filter's per event values also fall with storm size where the true ones do not, so the record is classified complacent and the fit is refused under CN-CAL-004 rather than reported wrong. Both figures are pinned in the calibration tests.","displaced_by":"Not displaceable by a measurement of a site, because nothing measures baseflow. What displaces it is a different convention, stated by the person running the calibration, whose consequence this engine then reports by refitting under both.","source":"Lyne & Hollick, Stochastic Time-Variable Rainfall-Runoff Modelling, Institution of Engineers Australia National Conference, Perth (1979) the one parameter recursive digital filter. The 0.925 parameter and the three pass convention are as reported by Nathan & McMahon, Water Resources Research 26(7):1465, 1990, rather than by the 1979 paper; pages not established by the transcriber","reasoning":"A convention, not a measurement. 0.925 is the value Nathan and McMahon report for the three pass application of the filter and is the figure quoted throughout the literature that followed. This transcriber has not established the page and says so in the source locator.","overridable":true},{"key":"baseflow.lyne_hollick_passes","quantity":"Number of filter passes in the Lyne and Hollick separation","value":3.0,"unit":"count","class":"C4","confidence":"Low","worth":"One pass forward leaves the separated baseflow with a phase shift, which is what the backward pass removes; three passes forward, backward, forward is the published convention and gives a smoother and lower baseflow than one. Each additional pass assigns more water to direct runoff and raises the fitted curve number, with the increments diminishing.","displaced_by":"The same answer as the filter parameter above: a different convention stated by the person, not evidence about a site.","source":"Lyne & Hollick, Stochastic Time-Variable Rainfall-Runoff Modelling, Institution of Engineers Australia National Conference, Perth (1979) the one parameter recursive digital filter. The 0.925 parameter and the three pass convention are as reported by Nathan & McMahon, Water Resources Research 26(7):1465, 1990, rather than by the 1979 paper; pages not established by the transcriber","reasoning":"A convention, not a measurement. An odd number of passes is required for the filter to end facing forward, and this engine refuses an even one rather than silently producing a phase shifted separation.","overridable":true},{"key":"baseflow.nday_coefficient","quantity":"Coefficient in the empirical rule for the time from peak to the end of direct runoff, N = c A ** e days with A in square miles","value":1.0,"unit":"day/(sq mi ** exponent)","class":"C4","confidence":"Low","worth":"Where the straight line separation meets the recession, and so how much of the falling limb is called direct runoff. The rule is a function of area alone, which is its weakness and the reason it is offered as one convention among several rather than as the answer: two catchments of the same area with different slopes and different geology do not have the same N, and this rule gives them the same N.","displaced_by":"An inflection point identified on this gauge's own recession, which is what the rule is a stand in for. A record long enough to fit a recession constant displaces the rule entirely.","source":"Linsley, Kohler & Paulhus, Hydrology for Engineers (3rd ed., 1982) the N = A ** 0.2 days rule for the end of direct runoff, A in square miles; page not established by the transcriber","reasoning":"A convention, not a measurement. The rule appears throughout the textbook hydrology literature in this form with the coefficient at one and the area in square miles; this transcriber has not established the primary page and says so in the source locator. It is held as two entries rather than one because a relation with half of it in prose is the arrangement that produced the abstraction basis defect, as cn.lambda005_coefficient records.","overridable":true},{"key":"baseflow.nday_exponent","quantity":"Exponent in the empirical rule for the time from peak to the end of direct runoff, N = c A ** e days with A in square miles","value":0.2,"unit":"dimensionless","class":"C4","confidence":"Low","worth":"The area scaling of the same rule. At the exponent of 0.2 a one square mile catchment gets one day and a hundred square mile catchment gets 2.51 days, which is the weak dependence the rule is known for. On the small catchments this product is usually pointed at, N is close to a day whatever the area, so the straight line separation on a small catchment is governed by this rule far more than by anything about the site.","displaced_by":"The same as the coefficient above: an inflection identified on the gauge's own recession.","source":"Linsley, Kohler & Paulhus, Hydrology for Engineers (3rd ed., 1982) the N = A ** 0.2 days rule for the end of direct runoff, A in square miles; page not established by the transcriber","reasoning":"A convention, not a measurement. See baseflow.nday_coefficient for why the relation is two entries.","overridable":true},{"key":"calibration.min_events","quantity":"Fewest events from which this engine will report a fitted curve number","value":5.0,"unit":"count","class":"C6","confidence":"Low","worth":"Everything, on the question the product is most likely to be asked to lie about. Below this the engine refuses under CN-CAL-001 and reports the per event curve numbers instead, which are arithmetic rather than a fit. A curve number fitted to one event is the event's own inversion with a confident name on it, and a bootstrap over four events has at most a few dozen distinct resamples, so the interval it returns is a property of the resampling and not of the record.","displaced_by":"Nothing this product can measure. The literature that fits curve numbers to gauged data works with tens of events per catchment and Hawkins (1993) is explicit that the ordered pairs method needs a long record, so this floor is far below good practice rather than at it, and it is a floor on reporting anything at all rather than a statement that five events are enough.","source":null,"reasoning":"REASONED, NOT MEASURED. Five is the smallest number at which a bootstrap has enough distinct resamples for a percentile interval to be more than a restatement of the sample, and it was chosen for that arithmetic property and for no hydrological one. The refusal it drives says so in those words.","overridable":true},{"key":"calibration.bootstrap_resamples","quantity":"Resamples drawn when forming the interval on a fitted curve number","value":2000.0,"unit":"count","class":"C6","confidence":"Moderate","worth":"The stability of the reported interval's endpoints and nothing else. It does not change the estimate and it does not make the interval mean more than it means. At 2000 resamples the endpoints of a 90 percent percentile interval are repeatable to a few hundredths of a curve number across seeds, which is well inside the width of any interval this engine will report.","displaced_by":"Nothing about a site. A larger number costs run time and buys endpoint stability that is already far finer than the interval is wide.","source":null,"reasoning":"REASONED, NOT MEASURED. Two thousand is the conventional order for a percentile interval and the confidence is MODERATE rather than LOW because the property it governs, Monte Carlo error in the endpoints, is one that can be and has been checked here by varying the seed.","overridable":true},{"key":"calibration.interval_probability","quantity":"Probability content of the reported interval on a fitted curve number","value":0.9,"unit":"fraction","class":"C6","confidence":"Low","worth":"The width of the number a reviewer is asked to look at. A 90 percent interval is roughly two thirds the width of a 95 percent one on the same resamples, and the choice between them is a presentation decision this product made rather than a property of the record. What it must not be read as is a probability that the catchment's curve number lies in the interval: the interval is over event sets like this one, which provenance.fitted requires every fitted value to state.","displaced_by":"A jurisdiction or a client that requires a stated interval at another probability. Nothing about a site displaces it.","source":null,"reasoning":"REASONED, NOT MEASURED. Ninety percent is the convention in engineering hydrology for a reported band, and the entry exists so that the number is looked up rather than typed in the fitting code where nothing could question it.","overridable":true},{"key":"calibration.min_event_rainfall_in","quantity":"Smallest event rainfall depth admitted to a fitting set","value":0.5,"unit":"in","class":"C6","confidence":"Low","worth":"Which events carry information about the curve number. Below the initial abstraction the method returns exactly zero for every curve number, so a small event constrains nothing and contributes a residual of zero to any fit that includes it, which flatters the diagnostics without improving the estimate. At the tabulated ratio, 0.5 in is the abstraction of a curve number of 80, so events below this threshold are uninformative over most of the range this product works in.","displaced_by":"Nothing measured. The bound that is actually meaningful is the abstraction of the curve number being fitted, which is not known until the fit is done, and this entry is the fixed stand in for that circularity.","source":null,"reasoning":"REASONED, NOT MEASURED. Half an inch is chosen as the abstraction of a curve number near the middle of the urban range. It is a judgment about where an event stops carrying information and it will be wrong in both directions on any particular catchment.","overridable":true},{"key":"calibration.interevent_dry_hours","quantity":"Dry period that separates one rainfall event from the next when events are selected from a continuous record","value":6.0,"unit":"hr","class":"C6","confidence":"Low","worth":"How the record is cut into events, which decides both the rainfall depth and the runoff volume of every point the fit sees. A shorter period splits one storm into two, giving a second event whose runoff belongs to the first and whose curve number comes out absurd; a longer one merges two storms and reports their combined depth against a hydrograph with two peaks. Both errors move the fitted curve number and neither announces itself.","displaced_by":"The catchment's own response time, which is the quantity the dry period is standing in for: a storm is separate from the last one when the catchment has finished responding to the last one. A time of concentration for the gauged watershed would replace this fixed figure with one derived from the catchment.","source":null,"reasoning":"REASONED, NOT MEASURED. Six hours is a common inter event period in the stormwater literature and is the same order as the response time of the watersheds this would be used on. It is not read off any publication this product holds.","overridable":true},{"key":"calibration.behaviour_asymptote_tolerance","quantity":"Curve number units by which the per event curve numbers must settle for the record to be classified as standard behavior","value":2.0,"unit":"CN","class":"C6","confidence":"Low","worth":"Whether a fitted curve number is reported at all. Hawkins (1993) separates catchments whose per event curve numbers decline with rainfall toward a constant, which is standard behavior and is the only case where a single curve number describes the catchment, from those that decline without settling, which is complacent behavior and where no curve number is approached. This entry is where this product draws the line between them, and a complacent record is refused under CN-CAL-004.","displaced_by":"A published criterion for the classification. Hawkins describes the three behaviors and this transcriber has not established a numerical test in the source, so the test here is this product's own and says so.","source":null,"reasoning":"REASONED, NOT MEASURED. Two curve number units is inside this product's own smallest override departure band, so a record whose largest events agree to within it is agreeing to within a difference the product elsewhere treats as immaterial. That is the argument for the number and it is an argument by analogy rather than a measurement.","overridable":true},{"key":"calibration.lambda_joint_min","quantity":"Lower bound on the initial abstraction ratio in a constrained joint fit","value":0.01,"unit":"dimensionless","class":"C6","confidence":"Low","worth":"Where the joint fit's search stops rather than what it finds. The bound matters because the objective is nearly flat along the ridge that trades lambda against the retention, so an unconstrained search wanders to whichever end it is pointed at and reports that end as a result. A fit that lands on a bound is reported as having landed on it rather than as a parameter estimate.","displaced_by":"Nothing about a site. The two published conventions are 0.2 and 0.05, and a ratio outside the interval these two bounds enclose is outside anything the curve number literature has fitted.","source":null,"reasoning":"REASONED, NOT MEASURED. One hundredth is an order of magnitude below the smaller published convention. The bound is a search constraint and not a statement that a catchment's ratio could be 0.01.","overridable":true},{"key":"calibration.lambda_joint_max","quantity":"Upper bound on the initial abstraction ratio in a constrained joint fit","value":0.3,"unit":"dimensionless","class":"C6","confidence":"Low","worth":"The other end of the same search interval. See calibration.lambda_joint_min: what both bounds are worth is that a flat objective is prevented from reporting an arbitrary extreme as an estimate, and that a fit which reaches either bound is disclosed as having reached it.","displaced_by":"Nothing about a site, on the same argument as the lower bound.","source":null,"reasoning":"REASONED, NOT MEASURED. Half again above the tabulated 0.2, which is the larger of the two published conventions. It is a search constraint.","overridable":true}]}