Skip to content

Known EPA SWMM 5.3 issues

This page tracks defects and their modeling impacts in the audited EPA SWMM 5.3 source and shows whether swmmrs has corrected them. Some issues originated before 5.3 but remain in the 5.3 bug_fixes branch. Each heading identifies its origin: a 5.3 regression was introduced or exposed by a 5.3-line change, a 5.3 implementation bug is a defect in behavior newly added in 5.3, while a prior SWMM 5 bug predates that line.

The audit covers bug_fixes through commit b8863c2. Source links below target the latest bug_fixes branch so each claim can be checked against the current 5.3 code. Completed corrections also appear under the applicable release in the solver compatibility changelog.

Marker Status Meaning
✅ Resolved in swmmrs swmmrs has a correction and focused regression coverage.
⚠️ Open in swmmrs swmmrs still mirrors or only partially corrects the affected EPA behavior.

✅ Resolved in swmmrs

Positive buildup remains eligible for washoff (5.3 regression)

EPA bug EPA source Modeling impact swmmrs resolution
Commit b91f310 changes the no-buildup guard in landuse_getWashoffQual from buildup == 0.0 to buildup >= 0.0. landuse.c L632–634 Because buildup mass is normally nonnegative, the changed guard returns before calculating exponential, rating-curve, or EMC washoff. Pollutant accumulates on the land surface, but runoff pollutant loads and downstream concentrations are understated. Retains the EPA 5.2.4 zero-only guard.
EPA bug EPA source Modeling impact swmmrs resolution
findSFLinkQual assigns the upstream node index to j and then reads Link[j].apiExtQualMassFlux and updates Link[j].totalLoad, although the intended object is the current link i. qualrout.c L447–469 API forcing and pollutant load can be applied to the wrong link. A node index larger than the link array can also cause an invalid memory access instead of producing results. Uses the current link_index consistently for mass flux and total load.
EPA bug EPA source Modeling impact swmmrs resolution
Ordinary and steady-flow link-quality paths can divide signed-zero external flux by zero available water. qualrout.c ordinary path L383–393; steady-flow path L447–469 The default zero flux can produce non-finite link concentration, pollutant load, and continuity values on a dry, zero-inflow link even though the model has no effective external forcing. Keeps the calculations inline and changes the steady-flow else to else if c_out != 0.0, matching the ordinary path so either signed zero skips division, load, and mass-balance updates.
EPA bug EPA source Modeling impact swmmrs resolution
Link-quality paths divide nonzero API mass flux by old volume + inflow × timestep even when the conduit ends dry and the available-water denominator is zero. qualrout.c L340–469 Positive forcing can produce infinite concentration and non-finite pollutant loads or continuity totals. Negative forcing has no aqueous mass to withdraw. Pumps, regulators, outlets, and dummy conduits also bypass the conduit reactor in which the forcing calculation is defined. Restricts nonzero link mass flux to non-dummy conduits. For a conduit classified dry at the end of the quality step, positive forcing is recorded as external inflow, link load, and non-remobilizing final dry storage without changing concentration; negative forcing is a no-op. Wet-conduit arithmetic and signed-zero behavior are unchanged. A complete zero candidate remains valid for clearing unsupported links.

Positive external pollutant forcing into dry nodes (5.3 implementation bug)

EPA bug EPA source Modeling impact swmmrs resolution
Routing adds each positive API node pollutant mass flux to both the node's temporary quality-inflow accumulator and external-inflow continuity. With negligible hydraulic inflow, a dry non-storage node clears that accumulator, while a dry storage node's mixing path ignores it and adds only concentration × new volume to final storage. A completely dry node therefore retains none of the supplied mass. routing.c L498–513; qualrout.c L239–339 Positive external mass appears in inflow totals but not in node concentration, outflow, reaction loss, or final storage. The model silently loses the source and reports a quality-continuity error. When any node type ends a quality-routing step dry with negligible hydraulic inflow, deposits the positive API mass-flux rate into non-remobilizing final dry storage and leaves concentration zero. Routing has already recorded the same rate as external inflow, so the correction adds only the continuity sink. Negative and signed-zero forcing are no-ops on dry nodes; wet-node mixing is unchanged.

Dry no-inflow node quality is cleared (5.3 regression)

EPA bug EPA source Modeling impact swmmrs resolution
For a non-storage node with negligible inflow, EPA restores old concentration only when depth is positive and no longer explicitly clears the dry branch. qualrout.c L258–262 A pollutant mass rate left in newQual can survive as though it were a concentration, then appear as a false concentration or pollutant pulse when the node rewets. Restores the EPA 5.2.4 dry branch: wet nodes retain old_qual, while dry nodes clear the temporary mass-inflow accumulator without changing continuity totals.

Covered rain barrels return rainfall from their local footprint (5.3 implementation bug)

EPA bug EPA source Modeling impact swmmrs resolution
Inside the per-unit loop, every covered barrel adds rainfall times the aggregate subcatch->lidArea instead of that unit's area. lid.c L1678–1679 Multiple covered-barrel records or mixed LID groups can return too much rainfall, overstating flow to the pervious area and distorting subcatchment runoff and water balance. Uses the loop-local replicated lid_area, preserving each unit's coverage and replication.

Partial-area LIDs exclude full-subcatchment runon (5.3 regression)

EPA bug EPA source Modeling impact swmmrs resolution
Commit 9d3bb8d changes the full-area LID check to subcatch->area >= subcatch->lidArea, a relation true for nearly every valid partial-area LID. lid.c L1687–1691 Full-subcatchment upstream runon enters the partial-area LID while the non-LID area also accounts for it. This double-counts runon and distorts LID inflow, storage, infiltration, treatment, and runoff. Uses the defensive full-area relation subcatch.area <= subcatch.lid_area. Validation rejects material overallocation and snaps accepted allocations above 99.9% to full coverage.

Averaged maximum depth uses reportable-node indexing (5.3 regression)

EPA bug EPA source Modeling impact swmmrs resolution
EPA stores averaged results only for reportable nodes using a compact cursor, then indexes that array with the model-wide node index when computing maximum depth. output.c accumulation L829–862; maximum update L918–923 Selecting only some nodes for reporting can show another node's maximum depth or read beyond the averaged-results array. The hydraulic trajectory itself is unchanged. Keeps the reportable-result cursor separate from the model node index.

Averaged maximum depth retains report units in metric projects (5.3 regression)

EPA bug EPA source Modeling impact swmmrs resolution
EPA applies UCF(LENGTH) to averaged depth even though node_getResults already accumulated report-unit values. output.c accumulation L851–860; first conversion node.c L483; second conversion L918–923 Metric projects apply a second length conversion to the averaged maximum-depth statistic. Binary averaged depths and the hydraulic trajectory are unchanged, but the summary statistic is wrong. Uses the averaged report-unit NodeDepth directly when updating the statistic.

Scheduled hotstarts use the absolute routing clock (5.3 implementation bug)

EPA bug EPA source Modeling impact swmmrs resolution
The parser stores a requested calendar date and time in decimal days, while the save loop compares that absolute value directly with elapsed routing time in milliseconds. iface.c L82–105; hotstart.c L99–116 A scheduled hotstart can be written near the beginning of a run instead of at the requested instant. A later run then resumes from the wrong hydraulic and water-quality state. Compares the absolute schedule with the absolute date/time returned by the routing clock. EPA 5.3's early emission is intentionally not mirrored.

⚠️ Open in swmmrs

Land-use coverage can exceed 100% (#230) (prior SWMM 5 bug)

EPA bug EPA source Modeling impact swmmrs status
The engine stores every [COVERAGES] percentage independently and subcatch_validate never checks their aggregate. The EPA desktop interface rejects the condition, but engine and API callers can still create it. subcatch.c reader L287–313; validation L352–406; surfqual.c area use L105–118 Two land uses at 100% each spend the subcatchment area twice, silently doubling buildup and potentially washoff loading. Totals below 100% remain valid. ⚠️ Not corrected. Input parsing has no aggregate check; programmatic preparation bounds each individual fraction to [0, 1] but does not bound their sum.

Subcatchment pollutographs use concentration interpolation (#229) (prior SWMM 5 bug)

EPA bug EPA source Modeling impact swmmrs status
subcatch_getResults linearly interpolates old and new concentrations, while routing interpolates old and new pollutant loads (runoff × concentration) and divides by interpolated flow. subcatch.c reporting L819–884; surfqual.c load interpolation L347–351; routing.c inflow L619–633; qualrout.c concentration L239–243 At report times inside a wet runoff step, the reported subcatchment concentration can differ sharply from the concentration routed to a directly connected outfall. Flow and mass continuity can remain correct. Aligning REPORT_STEP with WET_STEP, or shortening the wet step, only masks the discrepancy. ⚠️ Not corrected; swmmrs mirrors the two interpolation paths. A fix must also define how LID drain flow and load participate in the reported concentration.

Long station names displace NCEI rainfall columns (#224) (prior SWMM 5 bug)

EPA bug EPA source Modeling impact swmmrs status
For the attached COOP:/HPCP export, findNWSOnlineFormat takes ValueOffset from the short header label but derives DataOffset from the actual data row. A station name wider than its nominal field shifts the row's rainfall value without shifting the saved header offset. rain.c offset discovery L584–634; row parsing L762–817 Date and time can parse from the row-relative offset while precipitation is read from the wrong character position, causing records to be discarded or the file to fail ingestion. Although the issue calls the download GHCN-Daily, the attached HPCP data selects the NWS Online rainfall parser, not the GHCND climate parser. ⚠️ Not corrected; the Rust port retains the split header-derived value_offset and row-derived data_offset. Its growable strings remove the C line-buffer limit but not the displaced-column defect.

API-provided buildup and washoff accounting (5.3 implementation bug)

EPA bug EPA source Modeling impact swmmrs status
The EPA 5.3 source header records a TODO to track API-provided buildup and washoff separately in the mass balance. surfqual.c L18–25 Continuity accounting for callers using these API paths has not been fully validated. ⚠️ Requires targeted mass-balance validation before it can be treated as resolved.

Updating issue status

When an issue is corrected in swmmrs:

  1. add focused regression coverage;
  2. record the correction in the translated source file's update history;
  3. move the issue from ⚠️ Open to ✅ Resolved on this page;
  4. add the correction to the applicable swmmrs release in the solver compatibility changelog; and
  5. validate representative results against an explicitly identified EPA source revision.