Measured precip stops after ~17 years (6210 daily records) — WGN fallback on 61.0.1 / 61.0.2 / 61.0.2.61 (Editor 3.2.4)

7 views

Skip to first unread message

yakob mohammed

unread,

Sep 8, 2026, 2:40:28 PM (3 days ago) Sep 8

to SWAT+

1. Primary Issue Location

Component: SWAT+ Executable (cli_pmeas.f90 / cli_precip_control.f90)
Area of impact: Climate input / basin water balance output
Editor: SWAT+ Editor v3.2.4, QSWAT+ v3.2.3
Project: Salima10 watershed, Malawi (25 CHIRPS stations)

2. Description

Daily measured precipitation is used correctly for the first ~15–17 simulation years, then the model switches to weather-generator fallback (~13 mm annual precip, then ~0 mm). This is not real drought — CHIRPS .pcp files contain valid data through 2025, with correct nbyr headers and record counts.

Failure year depends on simulation start, which points to an internal climate read/indexing limit, not bad file content.

\Executables tested \

\ Evidence — basin annual precip (mm)\

excuted versions.png

Climate files verified OK

Source: QSWAT+ CHIRPS Malawi, 25 stations (ML_000001.pcp … ML_000025.pcp)
ML_000001.pcp: nbyr=26, 9497 daily records, years 2000–2025
Runtime split file: nbyr=23, 8400 records, years 2003–2025
Record 6210 = last day of 2016 (end of 17th year from 2000) — matches test1 failure at 2017
Record 8401 ≈ end of 2022 — matches cal1 failure at 2023 (5844 skip records + 2557 loaded ≈ 8401 total read budget)

diagnostics.out (all scenarios)

ML_000001.pcp file not found (pgage) ML_000001.tmp file not found (tgage) file not found (basins_carbon.tes)

Suspected mechanism

In cli_pmeas.f90, array is allocated as ts(366, nbyr) from header, but effective load appears capped at ~6210 daily records (~17 years from file start) or ~8401 total records read (including skip-to-sim-start). When cli_precip_control detects out-of-bounds, precip is set to −98 and WGN fills in.

Python replication of read + bounds logic on the same files shows all 26 years should work — discrepancy is in the compiled executable, not file content.

Related group threads: CLI_PMEAS TSS overflow (nbyr mismatch), CLI_PMEAS TS overflow — our case is the silent bounds-fail variant, not a crash.

What is NOT the cause (already ruled out)

Wrong nbyr header (matches record count)
Missing .pcp/.tmp in TxtInOut (files present)
SQLite / Editor import errors (all 25 stations linked)
Dev-build regression (official 61.0.2 identical)
CHIRPS data quality (annual totals valid through 2025)

Attachments (can provide on request)

Minimal TxtInOut zip: test1 scenario (one station ML_000001.pcp sufficient)
basin_wb_yr.csv, diagnostics.out, time.sim, pcp.cli, weather-sta.cli

Question for the team

Is there a hardcoded daily-record or year limit in cli_pmeas / climate loading (legacy SWAT2012 ~8401-day cap)? Can it be allocated from actual loaded record count rather than header nbyr only?