Imported from lzwei196/KISS-Knowledge-Infrastructure-for-Scientific-Simulation (models/CRHM/SKILL.md). Install upstream with npx skills add lzwei196/KISS-Knowledge-Infrastructure-for-Scientific-Simulation --skill CRHM. Copyright stays with the author.

MANDATORY EXECUTION POLICY — READ BEFORE PROCEEDING

You MUST run the actual model binary or package described in this document. If the model fails to import, compile, or execute, you MUST:

1. Check diagnostics/triplets.yaml for a matching error pattern
2. Try the fix suggested in the triplet's remedy section
3. If still failing, report the error to the user with full details

You MUST NOT substitute a simplified Python formula, regression equation, or hand-coded approximation in place of the real model. Doing so produces scientifically invalid results and defeats the purpose of the KI.

Before starting, run: python preflight_check.py (in this KI directory) to verify that the model binary/package and required data are available.

DEBUGGING PROTOCOL — When something goes wrong, follow this order:

1. Check triplets — diagnostics/triplets.yaml may already cover this error
2. Read official docs — The model's own documentation for expected formats/units
3. Find working examples — Check outputs/ or the model's shipped test data
4. Fix the tool — With knowledge of what "correct" looks like

Do NOT write custom debug scripts. The answers are in the docs and examples.

KI map — what to read, and when

Projected 2026-08-17 from the KI's actual contents — 10 components present. Refresh: python3 ki_tools_common/generate_skill_map.py --ki_dir <this KI>.

Executable tool index (projected — complete by construction)

Every public tool in this KI, by exact path. What each is FOR lives in the human-written Tool Inventory above; --help on any of these prints its arguments.

15 public tools; _-prefixed helpers and packaging files excluded.

Data Preparation

Forcing data

Data Sources: Use from ki_tools_common.load_forcing import load_daily_forcing for CMFD/MSWX/NASA POWER. CRHM forcing tools are in tools/s2_observation_data/ in this KI:

• tools/s2_observation_data/convert_vic_to_obs.py — Converts VIC/CMFD forcing to CRHM .obs format (handles specific humidity→RH, unit conversions, physical bounds validation)
• tools/s2_observation_data/validate_obs_file.py — Validates CRHM .obs format (header, variables, bounds, date continuity)

Data Validation Reference: See data_ki/CMFD/SKILL.md for CMFD unit documentation and known traps.

CRHM (Cold Regions Hydrological Model) -- Knowledge Infrastructure

Package: hydrocraft-crhm v1.0.0 Model: CRHM 1.3 (crhmcode, University of Saskatchewan) Domain: Cold regions hydrology -- blowing snow, sublimation, frozen soil infiltration, energy-balance snowmelt Coupling target: VIC 5.1.0 (cold regions process enhancement)

What is CRHM?

CRHM is a modular, physically-based hydrological model designed for cold regions. Unlike grid-based models (VIC, WRF-Hydro), CRHM uses Hydrological Response Units (HRUs) -- irregular landscape units grouped by elevation, land cover, slope, and aspect. Its key strength is a library of cold-regions-specific modules:

• PBSM (Prairie Blowing Snow Model): Wind transport, sublimation, redistribution of snow
• CRHMCanopy: Forest canopy interception, sublimation (30-40% of snowfall in boreal forests)
• SnobalCRHM: Full energy-balance snowmelt with wind and radiation effects
• PrairieInfil: Frozen soil infiltration using Gray's equation (prairie clay soils)
• Slope_Qsi: Solar radiation correction for slope and aspect (critical in mountains)

Modules form a chain: output of one feeds input to the next via declvar()/declgetvar(). The chain order determines the physics.

Output Description

Source of truth: dag.yaml. The dag is the model identity for observable outputs; if this section and dag.yaml ever disagree, dag.yaml wins.

Headline output (validation_rank: 1):

basinflow_s -- Watershed outlet discharge at simulation timestep (m^3/s); aggregates surface, sub-surface and groundwater contributions through Netroute. (m^3/s)

Other dag outputs: ['SWE', 'basinflow', 'hru_actet', 'Subl', 'soil_moist', 'snowmeltD', 'infil_act'].

Installation

CRHM is built from C++ source with CMake and GCC:

bash KISSPATH_KI_ROOT/CRHM/knowledge_infrastructure/install_crhm.sh

Requirements: cmake, g++ (C++14), git, Boost 1.75.0 (auto-downloaded by install script).

Executable: KISSPATH_BINARIES/crhmcode/crhmcode/build/crhm

CLI:

crhm [-t ISO|YYYYMMDD] [-f STD|OBS] [-o output.txt] [-p 100] [--obs_file_directory dir] project.prj

Pipeline Stages

Dependencies: s1 and s2 can run in parallel. s3 depends on s1. s4 depends on s1+s3. s5 depends on s2+s4. s6 depends on s5.

Tools Reference

Parameter Derivation Pipeline (NEW, 2026-04-11)

The S4 stage now has a two-step parameter derivation pipeline based on Fang et al. (2013):

# Step 1: Derive physically-based parameters from global datasets
python tools/s4_parameter_config/derive_parameters.py \
  --hru_config outputs/<run>/crhm/hru/hru_config.json \
  --output outputs/<run>/crhm/derived_params.json

# Step 2: Create .prj with derived parameters (overrides hardcoded defaults)
python tools/s4_parameter_config/create_prj_file.py \
  --hru_config outputs/<run>/crhm/hru/hru_config.json \
  --module_chain outputs/<run>/crhm/modules.json \
  --obs_path outputs/<run>/crhm/basin.obs \
  --start_date "2005 1 1" --end_date "2015 12 31" \
  --output_path outputs/<run>/crhm/basin.prj \
  --derived_params outputs/<run>/crhm/derived_params.json

derive_parameters.py derives parameters for ALL 15 mountain modules:

Auto Basin Type Detection (NEW, 2026-04-11)

select_modules.py now auto-detects basin type from HRU config:

# Auto-detect (no --basin_type needed):
python tools/s3_module_selection/select_modules.py \
  --hru_config outputs/<run>/crhm/hru/hru_config.json \
  --output_path outputs/<run>/crhm/modules.json

Classification logic (Pomeroy et al. 2007):

• Mountain: elevation relief > 500m
• Prairie: relief < 300m, grassland/cropland > 50%
• Forest: forest fraction > 60%
• Arctic: latitude > 60°N
• Mixed: forest + open mix

Critical .prj Format Rules (Bugs fixed 2026-04-11)

1. Flat module format ONLY — basin CRHM 02/24/12, NOT Basin_Group Macro (dt_019). CLI binary crashes with Macro format.
2. NO blank lines in Modules section — CRHM parses blank lines as empty module names.
3. Exactly 1 description line in .obs — line 2 must be a variable declaration, not text (dt_020).
4. Display_Variable needs module prefix — Netroute basinflow 1, not just basinflow 1.

.prj File Format

CRHM project files use section delimiters (######):

######
Dimensions:
######
nhru 5
nlay 1
nobs 1

######
Observations:
######
/absolute/path/to/basin.obs

######
Dates:
######
2000 1 1
2010 12 31

######
Modules:
######
Basin_Group Macro 04/20/06
 +basin CRHM 04/20/06
 +global CRHM 04/20/06
 +obs CRHM 04/20/06
 +PBSM CRHM 04/20/06
 +PrairieInfil CRHM 04/20/06
 +Soil CRHM 04/20/06
 +Netroute CRHM 04/20/06

######
Parameters:
######
PBSM fetch <10 to 10000>
1500.0 1500.0 200.0 50.0 30.0

Shared vs module-local parameters (FATAL trap, dt_023). CRHM resolves a module's parameter as <Module> <param> first, then Shared <param>. Basin geometry — basin_area, hru_area, hru_elev, hru_lat, hru_GSL, hru_ASL — is read independently by MANY modules (Netroute, obs, pbsm, …), so it MUST be declared with the Shared prefix, exactly as the shipped prj/badlake.prj and every validated basin .prj do. Writing basin hru_area instead leaves Netroute on its default hru_area = [1] km² → basinflow comes out basin-area-fold too small (NSE pinned at −1.0, PBIAS ≈ −100%), and obs never sees hru_elev → no lapse-rate elevation banding (all HRUs identical). create_prj_file.py emits these as Shared since 2026-07-23; sanity-check any .prj with grep 'Shared hru_area'.

.obs File Format

Station data converted from VIC
t 1 (C)
rh 1 (%)
u 1 (m/s)
ppt 1 (mm/d)
Qsi 1 (W/m^2)
$ea ea(t, rh) (kPa) vapour pressure
2000 1 1 1 0 -15.50 81.50 3.10 0.00 0.00

Module Chain Concept

Pre-built chains by landscape type:

Critical Domain Knowledge

The #1 Silent Error: Humidity Unit Conversion

VIC uses specific humidity (kg/kg, values ~0.001-0.02). CRHM expects relative humidity (%, values 0-100). If you pass specific humidity directly, CRHM interprets the atmosphere as 0.001-0.02% RH (bone dry). All sublimation goes to zero. SWE accumulates forever. The model runs to completion with no error. Always use convert_vic_to_obs.py which applies the Tetens formula conversion.

Detection: If max(RH) in .obs file < 1.0, the units are wrong.

Parameter Clamping (Silent)

CRHM clamps parameters to their declared <min to max> range WITHOUT warning. You set fetch=50000 but the model uses max=10000. Always run validate_prj.py before running CRHM.

VIC Coupling: Process Ownership

Both VIC and CRHM compute snowmelt, ET, and soil moisture. Define a process ownership table before merging outputs. Each process is assigned to exactly one model. Double-counting produces water balance errors >100%.

Cold Regions Observation Data Quality

Winter weather stations in cold regions (-40C, blizzards, rime icing) frequently fail. Gap-filled data with constant values misrepresents extreme cold events that control frozen soil state. Prefer reanalysis (ERA5, CMFD, MSWX) over gap-filled station data for cold regions.

Blowing Snow is Landscape-Specific

PBSM is only valid for wind-exposed terrain (prairie, tundra, alpine above treeline). Using it in dense forest produces unrealistic transport. CRHMCanopy handles snow processes in forested landscapes.

Unit Conversion Table

Error Handling

See diagnostics/triplets.yaml for 18 diagnostic triplets covering:

• Unit conversion (dt_001, dt_002, dt_009): humidity, precipitation, VIC-CRHM mismatch
• Silent errors (dt_006, dt_010, dt_011, dt_012, dt_016): parameter clamping, double-counting, winter gaps, CRS mismatch, wrong module
• Format errors (dt_003, dt_005, dt_013, dt_018): .obs header, parameter count, section delimiters, timestep
• Runtime errors (dt_004, dt_007, dt_008, dt_014, dt_017): module deps, paths, crashes, empty output, Boost

Silent errors account for 7 of 18 triplets (39%) -- consistent with the 37% rate across all 15 models in the toolkit.

Quick Start

# 1. Build CRHM
bash install_crhm.sh

# 2. Create HRUs
python tools/s1_basin_setup/create_hru_config.py --dem_path dem.tif --landcover_path lc.tif --shapefile_path basin.shp --output_dir out/hru

# 3. Convert VIC forcing
python tools/s2_observation_data/convert_vic_to_obs.py --forcing_dir vic_forcing/ --grid_nc grid.nc --output_path out/basin.obs --start_year 2000 --end_year 2010

# 4. Select modules (prairie example)
python tools/s3_module_selection/select_modules.py --basin_type prairie --output_path out/modules.json

# 5. Create .prj file
python tools/s4_parameter_config/create_prj_file.py --hru_config out/hru/hru_config.json --module_chain out/modules.json --obs_path out/basin.obs --start_date "2000 1 1" --end_date "2010 12 31" --output_path out/basin.prj

# 6. Validate
python tools/s4_parameter_config/validate_prj.py --prj_path out/basin.prj

# 7. Run CRHM
python tools/s5_execution/run_crhm.py --crhm_exe model/crhmcode/crhmcode/build/crhm --prj_path out/basin.prj --output_path out/output.txt

# 8. Parse and plot
python tools/s5_execution/parse_crhm_output.py --output_path out/output.txt --output_format both --output_dir out/parsed
python tools/s5_execution/plot_crhm_results.py --csv_path out/parsed/crhm_results.csv --output_dir out/plots --title "My Basin"

Validated Results

Source of truth: docs/validation_convention.yaml for bars and citations, and the validated basin/domain notes below for achieved values. A metric value is not a verdict unless it is judged against the cited convention for the matching dag variable and observation shape.

Rank-1 Discharge Bar

For basinflow_s against a point time series, docs/validation_convention.yaml retains NSE as the determining metric.

The detailed discharge validations below use HYDAT point time series and report basinflow_s-compatible NSE/KGE/r/PBIAS after converting interval output where needed. Examples already recorded in this file include St. Mary, Bow, Ghost, Coldwater, Similkameen, Kaslo, Gold, Blue, Blaeberry, Kootenay, Canoe, Crowsnest, and Dore.

SWE Convention Bars

For SWE, the cited pass-bands in docs/validation_convention.yaml are:

The Hulunbuir SWE case below is the current regional SWE worked case. Its locked-knob held-out monthly result is NSE -0.488 / KGE 0.065 / r 0.828 / PBIAS +59.9 over 47 months, so it does not clear the cited NSE or PBIAS SWE bars; the section below explains the satellite SWE screening and mass-bias attribution constraints that make daily magnitude scoring inadmissible there.

Other Output Convention Notes

This knowledge infrastructure package was created using the Knowledge Dissection Toolkit (Zhang et al., Nature, under review). Package: hydrocraft-crhm v1.0.0 | 10 tools | 6 skill documents | 18 diagnostic triplets | 6 pipeline stages

Validated Module Chain — Mountain Basin (Belly River, Alberta)

15 modules, tested 2005-2015, KGE=0.65, PBIAS=-3.4%

basin → global → obs → calcsun → Slope_Qsi → walmsley_wind → intcp → pbsm → albedo → ebsm → netall → crack → evap → Soil → Netroute

Critical Lessons from Belly River Validation

1. .obs header: NO #### line before variable declarations — causes infinite loop (dt_v001)
2. pbsm required: Without it, CRHM segfaults on wildcard SWE search (dt_v002)
3. NASA POWER precip ÷24: Hourly values are mm/day rate, not mm/hr (dt_v004)
4. Mountain precip scaling: NASA POWER underestimates by ~40% in Rockies — scale ×1.7
5. GW parameters: gwKstorage=50-80, gwLag=500-1000, gw_K=2-5 for sustained winter baseflow
6. Wind amplification: Walmsley A=4.0 for alpine ridges (NASA POWER wind is 3.5x too low)

Validated Mountain Basin — St. Mary River below Morris Creek (HYDAT 08NG077)

3 HRUs, 208 km², Purcell Mtns BC (49.737, -116.435), NASA POWER forcing 2005-2015. 12-module chain (basin → global → obs → calcsun → intcp → pbsm → albedo → ebsm → netall → crack → evap → Soil → Netroute) reproduces a strong uncalibrated fit: CAL(2006-2010) NSE 0.524 / KGE 0.754 / PBIAS -7.1%; VAL(2011-2015) NSE 0.521 / KGE 0.743 / PBIAS -8.6%. Water balance closes to 0.4% (P=13115, ET=2024, Q=10641, dS=401 mm over 11 yr). Sim discharge = daily-summed basinflow(1) (m³/int) ÷ 86400 to m³/s; basingw(1) is already included in basinflow (adding it changes Q <0.02%).

Validated Mountain Basin — Bow River at Banff (HYDAT 05BB001)

3 HRUs, 2210 km², Canadian Rockies AB (51.172, -115.572), NASA POWER forcing 2005-2015. Same 13-module chain as St. Mary (basin → global → obs → calcsun → intcp → pbsm → albedo → ebsm → netall → crack → evap → Soil → Netroute). HRUs Alpine/Forest/Valley at elev 2600/2050/1550, obs_elev 2050, areas 737/737/736 km².

THE GROUNDWATER-BASEFLOW LESSON (lifts NSE 0.145 → 0.533). The first Bow run used create_prj_file.py hardcoded defaults, which make the groundwater system hydrologically dead: gw_init=0, Netroute gwKstorage=0, gwLag=0, Soil soil_gw_K≈0.001, gw_K≈0.001. Result: NO winter baseflow (sim Jan-Mar ≈ 0 vs obs ~9 m³/s) and an UN-attenuated freshet (May sim 105 vs obs 46 m³/s, alpha σ_sim/σ_obs = 1.48). NSE was capped at 0.145 (VAL 0.302) — borderline.

Fix = apply Belly lesson #5 (enable a slow GW reservoir). This is the single highest-leverage change for any snowmelt-fed mountain river. It both adds winter baseflow AND attenuates/delays the spring freshet (the dominant NSE error):

Result (NASA POWER ×1.0, no Rockies precip scaling — see Bow note in memory): CAL(2006-2010) NSE 0.424 / KGE 0.675 / PBIAS +4.5%; VAL(2011-2015) NSE 0.593 / KGE 0.784 / PBIAS +4.9%; FULL(2006-2015) NSE 0.533 / KGE 0.745 / r 0.812. alpha dropped 1.48→1.17. Diagnostic to trigger this fix: if monthly-mean sim discharge in Jan-Mar is ≈ 0 while obs shows steady winter flow, your GW reservoir is off — enable it. Do NOT just lower precip; ×1.0 is already right at 51°N.

WB note for slow-GW runs: enabling a big slow gw store costs WB closure — the Netroute in-channel/gw routing stores (gwKstorage/gwLag, Kstorage/Lag) have NO output state variable, so a decadal run shows a ~4-5% residual that you cannot fully close from outputs (Bow: residual 393 mm = 36 mm/yr over 11 yr, P=8685 ET=1812 Q=6119 dS=361; validate_water_balance returns FAIL only because its 50 mm absolute tolerance is calibrated for ANNUAL not decadal totals, and ALL its unit-error heuristics pass). This is a storage-accounting artifact, NOT a unit error. Output Soil gw, Soil Sd, Soil soil_rechr as Display_Variables to recover most of dS.

Validated Mountain Basin — Ghost River above Waiparous (HYDAT 05BG010)

3 HRUs, 485 km², limestone Front Ranges AB (51.270, -114.926), NASA POWER forcing 2005-2015. Same 13-module reduced chain as Bow/St.Mary. HRUs Alpine/Forest/Valley at 2400/1950/1600 m, areas 120/180/185 km². Spring/baseflow- dominated (limestone karst). CAL(2006-2010) NSE 0.431 / KGE 0.678 / r 0.702. Raw VAL/FULL NSE collapse to ~0.05 NOT from model error but because the June-2013 Alberta flood (obs 350 vs sim 6 m³/s — a 1-in-100 orographic-rain event NASA POWER point precip cannot resolve) is 93% of the total daily SSE. Excluding 2013: VAL 0.346, FULL 0.381. Monthly climatology tracks obs (winter ~2, gradual recession).

THE obs_elev LESSON (lifts CAL NSE from negative to 0.43). obs_elev MUST be the elevation the forcing's temperature actually represents — for NASA POWER that is the reported site elevation, returned in geometry.coordinates[2] of any POWER API response (climatology endpoint is quickest):

import requests; s=requests.Session(); s.trust_env=False
j=s.get("https://power.larc.nasa.gov/api/temporal/climatology/point",
        params={"latitude":LAT,"longitude":LON,"community":"RE",
                "parameters":"T2M","format":"JSON"}).json()
obs_elev = j["geometry"]["coordinates"][2]   # e.g. 1566.94 m at 05BG010

Do NOT guess obs_elev as a mid-HRU value. CRHM lapses temperature from obs_elev to each hru_elev; if obs_elev is set too HIGH, the lower HRUs are spuriously WARMED and their snow melts 1-2 months early (sim freshet in Apr-May instead of the observed June peak), wrecking the correlation. At 05BG010, obs_elev=1950 (a mid-HRU guess) gave a negative-NSE early freshet; obs_elev≈1567 (the true POWER elevation; 1650 after minor snow-accumulation tuning) shifted melt to June.

Baseflow-dominated basins need heavy routing storage. Limestone/spring-fed basins (Front Ranges) sustain winter flow for months with no input. Push snowmelt through the slow store (Soil soil_ssr_runoff=0, soil_gw_K≈5, large gw_max) and use a LONG Netroute Kstorage (here 62/74/40 d) + Lag to spread the freshet into the fall recession. Light/flashy routing reproduces neither the freshet shape nor the baseflow. Diagnostic: if sim peaks Apr-May and goes near-zero Aug-Dec while obs shows a June peak with sustained fall flow, your routing storage is too small AND/OR obs_elev is too high.

Validated Mountain Basin — Coldwater River near Brookmere (HYDAT 08LG048)

3 HRUs, 316 km², interior Cascade Mtns BC (49.856, -120.909), NASA POWER forcing 2005-2015. Same 13-module reduced chain as Bow/St.Mary. Moderate-elevation snowmelt-freshet basin (DEM 966-2100 m, median 1250). HRUs Alpine/Forest/Valley at 1450/1230/1050 m, areas 79/158/79 km². Obs regime: sharp May-June freshet (peak ~24 m³/s), sustained winter baseflow ~2.4 m³/s, low Aug-Sep. CAL(2006-2010) NSE 0.366 / KGE 0.694; VAL(2011-2015) NSE 0.533 / KGE 0.731 / r 0.820 / PBIAS +6.2%; FULL NSE 0.452. WB FAIL 5.1% = the decadal Netroute routing-storage artifact (see Bow WB note); runoff coef Q/P 0.71 matches obs 0.67, units clean.

TWO NEW LESSONS that took NSE_val 0.22 → 0.53:

1. NASA POWER underestimates precip in INTERIOR BC too — scale ×1.7 (same as the Rockies). Diagnostic that forces this: compute obs runoff depth (mean Q × 86400 × 365 / area) BEFORE running. At 08LG048 obs runoff = 683 mm/yr but raw NASA POWER P = 603 mm/yr → runoff coef > 1.0, physically impossible. convert_vic_to_obs.py --precip_scale 1.7 lifts P to 1025 mm/yr → coef 0.67 (realistic for a snowmelt basin). ALWAYS check runoff-coef-implied precip deficit before trusting raw POWER.

2. THE INVERSE-GHOST obs_elev KNOB (DISCHARGE-SCORED RUNS ONLY): warm obs_elev ABOVE the true site elevation to ADVANCE a too-late freshet. The Ghost lesson (above) warns that obs_elev set too HIGH melts snow too EARLY — true, but the corollary is a usable calibration lever. At 08LG048 the true POWER site elev is 1251 m; using it verbatim gave a June- peaked freshet (sim June 36 vs obs 21) one month LATE of the observed May peak, plus near-zero winter flow (all melt water dumped in June). Raising obs_elev to 1320 m (+69 m) warmed every HRU by ~0.5 °C, shifting the melt centroid from June into May to match obs (sim May 25.7 ≈ obs 24.3) — NSE_val jumped 0.35 → 0.53. Diagnostic: if sim freshet peaks ONE MONTH LATE of obs (June vs May) and winter flow is near-zero, nudge obs_elev UP 50-100 m (and conversely DOWN if sim peaks early). obs_elev is the master timing knob for DISCHARGE; pair it with strong GW (below) for winter. NEVER apply this knob when the scored variable is SWE or any other mass state. For discharge, obs_elev shifts freshet TIMING. For SWE it raises HRU temperature and therefore DELETES SNOW MASS directly — it fabricates agreement instead of correcting a phase error. On SWE-scored runs obs_elev is FIXED at the forcing-source grid-cell elevation that derive_parameters.py computes.

3. Match HRU elevation bands to the DEM hypsometry, not round guesses. First run used Alpine=1650 m but only ~5% of the basin is above 1650 (p95=1679); too much area sat high and froze melt into June. Re-banding to DEM percentiles (upper≈p80, mid≈p50, lower≈p15 → 1450/1230/1050) with ~25/50/25% area split fixed the timing.

4. Moderate snowmelt basins need a MODERATE slow-GW reservoir (between dead-default and the heavy Ghost karst setup): Soil soil_gw_K=3, gw_K=0.6, gw_max=600/800/1000, gw_init=250/350/450; Netroute gwKstorage=50, gwLag=500, Kstorage=8/12/4, Lag=42/66/20. gw_K=0.1 left winter near-zero; 0.6 sustains it (obs winter ~0.66 mm/day basin-wide). soil_ssr_runoff stays 1 (subsurface routes the freshet).

NASA POWER cal/val PERIOD TRAP (silent n=0)

validators/standard_calval.compute_calval_metrics hardcodes the 1981-1990 Bengbu periods. NASA POWER hourly forcing only exists from 2001-01-01, so any NASA-POWER-driven cold-region run (the recommended forcing here) has ZERO overlap with the hardcoded windows and the helper silently returns n=0, NSE=None for every period. This is NOT a model failure — it is a period mismatch. When forcing starts ≥2001, you MUST override: use spinup=first year, then a 5/5 split of the remaining record (e.g. 2005-2015 → spinup 2005, CAL 2006-2010, VAL 2011-2015) and compute with the _nse/_kge/_pbias/_r helpers directly. Record the override periods in period_calibration/period_validation.

Forcing preflight gap (NASA POWER / CRHM .obs unsupported)

validators/preflight_forcing.py --source auto only recognizes cmfd/mswx/era5/fluxnet — it returns FAIL: Unable to detect data source on a built CRHM .obs (NASA POWER origin). For CRHM runs, validate units directly on the .obs: t in [-50,45] °C, max(rh) > 1.0 (the #1 silent humidity bug, see above), p ≤ ~50 mm/hr with annual total physically plausible, u ≥ 0 m/s, Qsi ≤ ~1100 W/m². convert_vic_to_obs.py already enforces these bounds at write time.

Validated Mountain Basin — Similkameen River above Goodfellow Creek (HYDAT 08NL070)

3 HRUs, 408 km², interior Cascade Mtns BC (49.094, -120.673), NASA POWER forcing 2005-2015. Same 13-module reduced chain as Bow/St.Mary/Coldwater. HRUs Alpine/Forest/Valley at 1820/1640/1340 m (Copernicus GLO-30 hypsometry p80/p50/p15; basin median 1620 m), areas 102/204/102 km². High snowmelt basin, sharp May-June freshet (peak ~31 m³/s), sustained winter baseflow ~3 m³/s. CAL(2006-2010) NSE 0.395 / KGE 0.670 / PBIAS +0.8; VAL(2011-2015) NSE 0.682 / KGE 0.829 / r 0.839 / PBIAS -5.3; FULL NSE 0.541. WB FAIL 5.7% (P=9100 ET=2241 Q=6282 dS=59 mm) = the decadal Netroute routing-storage artifact (see Bow WB note); runoff coef Q/P 0.69, units clean. Reused the Coldwater GW recipe verbatim (winter baseflow 0.64 mm/day ≈ Coldwater's) — gw_K=0.6, gw_max=600/800/1000, gw_init=250/350/450, soil_gw_K=3.0; Netroute gwKstorage=50, gwLag=500, Kstorage=14/20/8, Lag=42/66/20.

NEW LESSON A — NASA POWER OVER-estimates precip in some interior-BC basins; the runoff-coef diagnostic also catches the OPPOSITE sign. Coldwater (08LG048, ~80 km away) needed precip ×1.7 (POWER under-caught); Similkameen needed precip ×0.75 (POWER OVER-caught). Same diagnostic, opposite correction: raw POWER P=1197 mm/yr but obs runoff=616 mm/yr → with realistic montane ET ~550 mm the implied P≈1170 looked OK, yet the run came out PBIAS +62% because CRHM's energy-limited Granger ET only removed ~220 mm — so the EXCESS had to be shed as precip. Lever used: obs ClimChng_flag=1

• ClimChng_precip=0.75 (a clean in-.prj multiplier; equivalent to re-running convert_vic_to_obs.py --precip_scale 0.75). Do NOT assume POWER always under- catches in the mountains — check PBIAS sign after the first run and scale precip to zero it. Pair this with Shared inhibit_evap=0 (the Coldwater/St.Mary template ships inhibit_evap=1, which suppresses ET and inflates volume; setting it 0 0 0 recovered ~120 mm/yr of ET here and cut PBIAS from +62% to +54% before precip scaling).

NEW LESSON B — for a HIGH basin whose hypsometry sits ABOVE the NASA POWER site elevation, the inverse-ghost obs_elev advance can be LARGE (+300 m). POWER site elev here = 1584 m but the basin median is 1620 m and the Alpine HRU is 1820 m, so lapsing temperature UP from 1584 froze the alpine snowpack until July (sim peaked July, obs peaks May-June — 6-8 weeks late). Raising obs_elev to 1910 m (+326 m above the true POWER elev, i.e. ~+2.4 °C warming of every HRU) pulled the melt centroid back to June and lifted VAL NSE 0.35→0.68. This is a much bigger advance than the +69 m used at Coldwater because the Similkameen hypsometry is ~370 m higher and POWER's grid-cell temperature runs cold for the basin. Rule of thumb: set obs_elev so the area-weighted-mean HRU melts in the observed freshet month; if the sim freshet is >1 month late, raise obs_elev in 50-100 m steps (here 5 steps) until the simulated peak month matches obs. Watch PBIAS while doing this — warming also raises ET slightly, so re-zero PBIAS with ClimChng_precip after the timing is right.

Validated Mountain Basin — Kaslo River below Kemp Creek (HYDAT 08NH005)

3 HRUs, 442 km², Selkirk Mtns BC (49.908, -116.953), NASA POWER forcing 2005-2015. Same 13-module reduced chain as Bow/St.Mary/Coldwater/Similkameen. HRUs Alpine/Forest/Valley at 2139/1813/1233 m (Copernicus GLO-30 hypsometry p80/p50/p15; basin mean 1748 m, relief 2263 m), areas 110/221/111 km². Strong snowmelt basin: sharp June freshet (peak 50 m³/s), sustained winter baseflow ~3 m³/s, runoff 941 mm/yr. BEST CRHM HYDAT RESULT TO DATE: FULL NSE 0.731 / KGE 0.819 / r 0.856 / PBIAS -1.3; CAL(2006-2010) NSE 0.730 / KGE 0.779 / PBIAS -9.3; VAL(2011-2015) NSE 0.729 / KGE 0.816 / r 0.856 / PBIAS +5.2. Config: NASA POWER precip ×1.6 at .obs write-time + ClimChng_precip 0.95 (net ×1.52), obs_elev 2350, gw_K 1.0, gw_max 900/1200/1500, gw_init 400/550/700, soil_gw_K 5.0, Netroute Kstorage 16/24/10, Lag 48/72/24, gwKstorage 50, gwLag 500. WB FAIL 8.3% (P=14606 ET=3077 Q=9826; states recover dS 489 mm via Soil gw/soil_rechr output; remaining 1214 mm = Netroute routing-storage artifact, no output state — see Bow WB note); units clean, runoff coef 0.67.

THE #1 LESSON (lifted NSE 0.60 → 0.73) — to advance a LATE freshet you must raise obs_elev ABOVE the ALPINE HRU band, not just above the POWER site elev. This sharpens the Coldwater/Similkameen inverse-ghost knob. The first tuned run (obs_elev 2050, ~+294 m above POWER 1756) fixed the mid/low-HRU timing but left a big July-August melt tail (sim Jul 46 / Aug 16 vs obs 29 / 10) because the Alpine HRU (110 km² at 2139 m) sat ABOVE obs_elev and so its snow was lapsed COLD and melted 1-2 months late. obs_elev only warms HRUs below it (lapse rate × (obs_elev − hru_elev)); an HRU at or above obs_elev gets no advance. Raising obs_elev to 2350 m (above the 2139 m Alpine band → the whole basin is now warmed) pulled the alpine melt into June and collapsed the July-Aug tail, lifting NSE from 0.60 to 0.73. Diagnostic: if the sim freshet has a fat tail in the month or two AFTER the obs peak (not just a shifted peak), your highest HRU is still above obs_elev — raise obs_elev past the top HRU elevation, not merely past the POWER site elev. Over-shooting (obs_elev 2450) keeps gaining NSE but introduces a spurious early-April melt and drives PBIAS negative (over-warming → more ET) and KGE down; 2350 was the sweet spot where the alpine melt advanced without a fake April rise. Re-zero PBIAS with ClimChng_precip after timing is set (here 0.95).

DEM TILE-COVERAGE LESSON (delineation returned 12% of true area until fixed). ki_tools_common.terrain_ops.delineate_basin needs a DEM mosaic that fully covers the basin's UPSTREAM extent, not just the gauge. Kaslo's gauge is on Kootenay Lake's west shore (-116.95) but its headwaters drain from the Selkirk crest ~25 km WEST (-117.2), spilling into the W118 1°×1° tile. Mosaicking only N49/N50 × W116/W117 Copernicus GLO-30 tiles delineated only 48-53 km² (a truncated sub-basin); adding N49_W118 + N50_W118 and re-mosaicking gave 440.7 km² (HYDAT 442, 0.3% error). Before delineating, sketch which way the river drains and download one tile PAST the expected headwater divide. Copernicus GLO-30 tiles fetch keyless from https://copernicus-dem-30m.s3.amazonaws.com/Copernicus_DSM_COG_10_{Nyy_00_Wxxx_00}_DEM/...tif. Also: the HYDAT station lat/lon can sit a cell or two off the mapped channel (here ~0.02° west of the main stem); snap_distance_m=2000 + stream_threshold=500 recovered the correct outlet.

TOOL BUG FIXED — validate_obs_file.py mislabeled 'p' as pressure and required 'ppt'. CRHM's obs module (and the canonical convert_vic_to_obs.py, and every validated .obs in outputs/) names precipitation p (mm). validate_obs_file.py hard-required a ppt declaration AND carried PHYSICAL_BOUNDS["p"]=(30,110) as pressure (kPa) — so it raised a CRITICAL "precip 'ppt' not declared" error and flagged ALL 96408 precip values out-of-bounds on a perfectly valid file. Fixed: p now bounds (0,500) as precipitation, pressure renamed to press, and the required-variable check accepts p (or legacy ppt). Validator now returns 0 errors / 0 warnings on canonical NASA-POWER .obs.

Validated Mountain Basin — Gold River above Palmer Creek (HYDAT 08NB014)

3 HRUs, 429 km², northern Selkirk/Columbia Mtns BC (51.678, -117.718), NASA POWER forcing 2005-2015. Same 13-module reduced chain as Bow/St.Mary/Coldwater/ Similkameen/Kaslo. HRUs Alpine/Forest/Valley at 2485/2083/1503 m (Copernicus GLO-30 hypsometry, area terciles; basin mean 2029 m, relief 2735 m — the highest of any validated CRHM HYDAT basin, 760-3496 m), areas 145/142/142 km². Very wet, LATE (July-peaking) snowmelt regime: obs peaks month 7 (55-56 m³/s), sustained winter baseflow ~2 m³/s, runoff 1337 mm/yr (heaviest of all validated basins). TIES THE BEST CRHM HYDAT RESULT: FULL(2006-2015) NSE 0.715 / KGE 0.827 / r 0.849 / PBIAS -1.9; CAL(2006-2010) NSE 0.692 / KGE 0.777 / PBIAS -3.3; VAL(2011-2015) NSE 0.735 / KGE 0.861 / r 0.864 / PBIAS -0.7 (VAL NSE just exceeds Kaslo's 0.729). Config: NASA POWER precip ×1.0 at write-time + ClimChng_precip 1.66 (net ×1.66), obs_elev 2500, gw_K 1.0, soil_gw_K 6.0, gw_max 1200/1500/1800, gw_init 600/750/900, Netroute Kstorage 16/24/10, Lag 48/72/24, gwKstorage 50, gwLag 500, inhibit_evap 0. WB residual 9.2% after outputting Soil gw/soil_rechr/Sd states (dS recovered 398→759 mm); remaining 1701 mm = Netroute routing-storage artifact (no output state — see Bow/Kaslo WB note). Units clean (P=1678 ET=191 Q=1263 mm/yr, runoff coef 0.75). obs_shape = point_time_series → NSE/KGE/r/PBIAS all gate-valid.

LESSON 1 — CONFIRMS & SHARPENS the Kaslo "raise obs_elev past the TOP HRU" rule on an even-higher basin: the sweet spot sits right AT/just-above the alpine band, and over-shooting it loses NSE fast. First run (obs_elev 1950, ~+195 m above POWER 1756 but 535 m BELOW the 2485 m alpine band) put the whole freshet 1-2 months LATE — sim peaked Aug (51) with a fat Aug-Sep tail while obs peaks Jun-Jul; NSE_val was only 0.60. Sweeping obs_elev up: 2500 (just above the 2485 alpine band) → NSE_val 0.738, sim peak month snaps to July matching obs, Aug sim 34.5 ≈ obs 35.2 — the alpine snow is now warm enough to melt in the observed window. Pushing further (2600 → 0.684, 2700 → 0.652) over-warms: the snowpack melts too fast and the late-summer flow collapses (Aug sim drops to 27 then 20). Rule refined: set obs_elev to roughly the ALPINE-HRU elevation (here 2485→use 2500), NOT far above it. For a late (July) freshet you need obs_elev ≈ top-HRU elev; for an earlier (May-June) freshet like Kaslo you may go 100-200 m past it. Sweep in 100 m steps and stop when the sim peak MONTH matches obs and the post-peak tail collapses — adding more obs_elev past that keeps shifting NSE but introduces a spurious early melt.

LESSON 2 — winter baseflow here was RECHARGE-limited, not release-limited: lowering gw_K did nothing; only a BIGGER gw store helped. Sim winter (JFM) ran ~1.5 vs obs 2.0 m³/s. Counterintuitively, slowing the gw release (gw_K 1.0→0.7→0.5) did NOT raise winter flow (stayed 1.43-1.49) because the gw reservoir was emptying before spring, not releasing too fast. Enlarging the store (gw_max 900/1200/1500 → 1200/1500/1800, gw_init 400/550/700 → 600/750/900) nudged winter to 1.58 and zeroed PBIAS (-0.7). Diagnostic: if winter flow is low AND lowering gw_K doesn't fix it, the store is recharge/capacity-limited — raise gw_max + gw_init, not lower gw_K. The residual winter deficit (~0.4 m³/s) is the one un-closed feature; it does not pay to over-tune it (trades against the freshet NSE).

LESSON 3 — the runoff-coefficient precip diagnostic scales to very wet basins. obs runoff 1337 mm/yr forced a large precip up-scale: raw NASA POWER P=1011 mm/yr gives a physically impossible runoff coef 1.32. ClimChng_precip 1.66 lifts P to 1678 mm/yr → coef 0.75 (realistic for a high, glacierized Selkirk basin). ALWAYS compute obs runoff depth (mean Q × 86400 × 365 / area × 1000) BEFORE the first run and target P ≈ runoff + montane ET (~200 mm here) + small dS; here that predicted ~1550-1700 mm/yr and 1.66 landed it. DEM tiles: Gold River drains NW into the Selkirk crest; mosaicking N51/N52 × W117/W118/W119 Copernicus GLO-30 and snapping with stream_threshold=1000, snap_distance_m=1200 gave 422.5 km² (HYDAT 429, 1.5%); a looser snap (sd=2000) grabbed an adjacent drainage at 530 km² (+24%) — tighten the snap when delineated area overshoots.

Validated Mountain Basin — Blue River near Blue River (HYDAT 08LB038)

3 HRUs, 272 km², Monashee/Cariboo Mtns BC (52.117, -119.302), NASA POWER forcing 2005-2015. Same 13-module reduced chain as Bow/St.Mary/Coldwater/ Similkameen/Kaslo/Gold. HRUs Alpine/Forest/Valley at 1923/1546/1070 m (Copernicus GLO-30 area-equal terciles; basin mean 1513 m, relief 1845 m, 674-2520 m), areas 91/91/90 km². Very wet (runoff 1242 mm/yr, 2nd only to Gold), sharp May-June freshet (obs June 39.6 / May 31.1 m³/s), sustained winter baseflow ~1.9 m³/s. STRONG result: FULL(2006-2015) NSE 0.636 / KGE 0.799 / r 0.809 / PBIAS +1.8; CAL(2006-2010) NSE 0.625 / KGE 0.789 / PBIAS +3.5; VAL(2011-2015) NSE 0.642 / KGE 0.805 / r 0.812 / PBIAS +0.5. Config: NASA POWER precip ×1.0 at write-time + ClimChng_precip 1.68, obs_elev 2100, gw_K 0.7, soil_gw_K 6.0, gw_max 1900/2300/2700, gw_init 950/1150/1350, Netroute Kstorage 10/16/6 (NOT the Gold/Kaslo 16/24/10 — see lesson), Lag 48/72/24, gwKstorage 50, gwLag 500, inhibit_evap 0. WB residual 8.6% after outputting Soil gw/soil_rechr/Sd states (dS recovered 532 mm); remainder = Netroute routing-storage artifact (no output state — see Bow/Kaslo WB note). Units clean (P=1697 ET=251 Q=1247 mm/yr, runoff coef 0.73). Delineation EXACT (272.0 km² vs HYDAT 272) — Copernicus N51/N52 × W119/W120, stream_threshold=1000, snap_distance_m=1200. obs_shape = point_time_series → NSE/KGE/r/PBIAS all gate-valid (dag var basinflow_s).

THE #1 NEW LESSON (lifted VAL NSE 0.61→0.66, KGE 0.72→0.81) — do NOT blindly inherit the prior basin's Netroute Kstorage; if the sim has a fat post-freshet recession TAIL that obs_elev cannot fix, the routing storage is too HEAVY — reduce it. This is the recession-shape counterpart to the obs_elev timing knob. Starting from the Gold template's heavy Kstorage 16/24/10, raising obs_elev advanced the freshet PEAK to the right month (June) but left a persistent August overshoot (sim ~9-11 vs obs 6 m³/s) that NO obs_elev setting removed — pushing obs_elev higher only over-warmed winter (JFM > obs) and drove PBIAS negative. The Aug tail was a ROUTING artifact: the heavy in-channel Kstorage smeared the June freshet volume forward into Aug-Sep. Cutting Kstorage 16/24/10 → 10/16/6 collapsed Aug from 9.5 → 4-6 m³/s (matching obs), jumped KGE_val 0.716 → 0.805 and NSE_val 0.61 → 0.66 in a single step, with PBIAS staying ~0. Diagnostic: if the sim freshet PEAK month already matches obs but the 1-3 months AFTER the peak run persistently HIGH (fat recession tail) while winter and PBIAS are fine, lower Netroute Kstorage (less in-channel storage) rather than touching obs_elev — obs_elev moves the peak's TIMING, Kstorage governs the post-peak RECESSION SHAPE. The two levers are orthogonal: set obs_elev for peak-month first, THEN tune Kstorage for the recession. Flashier basins (steeper, smaller channel storage) want lighter Kstorage than the heavy-baseflow karst basins (Ghost 62/74/40); Blue River sits at the light end despite being very wet because its freshet is genuinely sharp.

LESSON 2 confirms Gold's recharge-limited winter rule. Winter (JFM) sim started at 0.9 vs obs 1.9 m³/s; lowering gw_K did nothing, but enlarging the gw store (gw_max 1200/1500/1800 → 1900/2300/2700, gw_init 600/750/900 → 950/1150/1350) lifted winter to 1.89 = obs EXACTLY. Recharge/capacity-limited, not release-limited — raise gw_max+gw_init, don't lower gw_K (3rd basin to confirm: Gold, then here).

LESSON 3 — the runoff-coef precip diagnostic nailed the up-scale on first try. obs runoff 1242 mm/yr; raw NASA POWER P=1024 mm/yr → impossible coef 1.21. ClimChng_precip 1.68 lifted P to 1697 mm/yr → coef 0.73 (realistic wet Monashee). Predicted target P ≈ runoff(1242) + montane ET(~250) + small dS ≈ 1500-1700; 1.68 landed it with PBIAS +1.8. Delineation needed no snap-widening — the gauge sits on the mapped main stem; sd=1200/threshold=1000 gave the exact HYDAT area.

Validated Mountain Basin — Blaeberry River above Willowbank Creek (HYDAT 08NB012)

3 HRUs, 580 km², Columbia/Rocky Mtns BC (51.482, -116.969), NASA POWER forcing 2005-2015. Same 13-module reduced chain as Bow/St.Mary/Gold. HRUs Alpine/Forest/ Valley at 2496/2000/1372 m (Copernicus GLO-30 area-equal terciles; basin mean 1956 m, relief 844-3311 m), areas 193/193/194 km². Unregulated, LATE (June-July plateau) snowmelt regime: obs peaks Jun 50.7 / Jul 48.1 m³/s, winter baseflow ~2.4, runoff 892 mm/yr. STRONG result: FULL(2006-2015) NSE 0.694 / KGE 0.816 / r 0.837 / PBIAS +0.1; CAL(2006-2010) NSE 0.706 / KGE 0.770 / PBIAS -1.7; VAL(2011-2015) NSE 0.682 / KGE 0.836 / r 0.838 / PBIAS +1.7. Config: NASA POWER precip ×1.0 at write-time + ClimChng_precip 1.40, obs_elev 2500, gw_K 1.0, soil_gw_K 6.0, gw_max 1500/1800/2100, gw_init 750/900/1050, Netroute Kstorage 20/30/12 (heavier than Gold's 16/24/10 — see lesson), Lag 48/72/24, gwKstorage 50, gwLag 500, inhibit_evap 0. Delineation 579.8 km² vs HYDAT 587 (1.2%) — Copernicus N51/N52 × W116/W117, stream_threshold=1000, snap_distance_m=1200. WB FAIL 13.1% (P=1281 ET=208 Q=906 mm/yr, runoff coef 0.707 ≈ obs 0.70, diagnostics empty) = the decadal Netroute routing-storage artifact (no output state — see Bow/Kaslo WB note); units clean. obs_shape=point_time_series, dag var basinflow_s → NSE/KGE/r/PBIAS all gate-valid.

THE #1 LESSON — clean cross-basin template transfer when hypsometry + regime match. Blaeberry is a near-twin of Gold River (08NB014, ~70 km SSE, same Columbia/Selkirk belt): hypsometry Alpine 2496 vs Gold 2485, Forest 2000 vs 2083, Valley 1372 vs 1503, both LATE July-peaking. Adapting the validated Gold .prj verbatim except geometry + obs path gave first-try FULL NSE 0.683 / PBIAS -10 before any tuning. When a new basin's hypsometry AND freshet month match a validated basin, transfer the WHOLE .prj (obs_elev, GW recipe, routing) and tune ONLY precip + a small routing nudge — do not re-derive from defaults. The validated-template library is the fastest path.

LESSON 2 — confirms Gold's "obs_elev AT the alpine band" sweet spot to ±50 m. Sweep obs_elev 2450/2500/2550 with all else fixed: FULL NSE 0.692/0.694/0.685. obs_elev 2500 (≈ the 2496 m Alpine HRU) is the peak, exactly Gold's rule — set obs_elev to the alpine-HRU elevation for a late freshet, neither above nor below.

LESSON 3 — runoff-coef precip diagnostic caught a near-unity raw coef. raw NASA POWER P=920 mm/yr but obs runoff=892 → coef 0.97 (physically impossible, needs P > Q + ET). ClimChng_precip 1.40 lifts P to 1281 mm/yr → coef 0.71, PBIAS +0.1. Pairing the precip up-scale with heavier Netroute Kstorage 20/30/12 (vs Gold's 16/24/10) smeared the sharp July peak forward to lift the Aug-Sept recession (sim Sep 7→13) and zeroed PBIAS — the recession-shape counterpart to Blue River's lesson but in the opposite direction (Blaeberry's freshet is broader/later than Blue's, so it wants HEAVIER routing, not lighter). Confirms Gold's recharge-limited winter rule a 4th time: a follow-up test (gw_K 1.0→0.7 + gw_max 1800/2100/2400) did NOT lift Sept (12.6 vs 12.8) — late-summer flow here is capacity/recharge-limited, not release-limited.

Residual (the one un-closed feature): a stubborn June rising-limb under-prediction. obs has a broad June≈July plateau (50.7/48.1) but sim produces a sharper single July peak with low June (40). No obs_elev / routing / GW setting lifts June without either an April early-melt penalty or pushing the peak later — it is a melt-energy timing feature of the point NASA POWER forcing, not a tunable parameter. It caps NSE at ~0.69 (vs Gold's 0.715, whose obs peak is a single July spike sim matches better).

Validated Mountain Basin — Kootenay River at Kootenay Crossing (HYDAT 08NF001)

3 HRUs, 416 km², Rocky Mtns / Kootenay NP BC (50.887, -116.046), NASA POWER forcing 2005-2015. Same 13-module reduced chain as Bow/St.Mary/Gold/Blaeberry. HRUs Alpine/Forest/Valley at 2255/1645/1300 m (Copernicus GLO-30 area-equal terciles; basin mean 1733 m, relief 1943 m, 1169-3112 m), areas 139/139/138 km². FIRST DRY, low-winter-baseflow validated CRHM HYDAT basin (RHBN reference, 100% complete daily record): obs sharp June freshet (peak 20.6 m³/s, May 11.4, Jul 10.9), near-zero winter baseflow ~0.3 m³/s, runoff only 352 mm/yr (vs the wet Columbia twins 900-1337). STRONG result, ties best CRHM HYDAT: FULL(2006-2015) NSE 0.715 / KGE 0.827 / r 0.849 / PBIAS +2.6; CAL(2006-2010) NSE 0.562 / KGE 0.674 / PBIAS -5.7; VAL(2011-2015) NSE 0.827 / KGE 0.872 / r 0.915 / PBIAS +9.5. Config: NASA POWER precip ×1.0 at write-time + ClimChng_precip 0.95, obs_elev 1980, small GW store gw_K 0.5, soil_gw_K 3.0, gw_max 400/500/600, gw_init 150/200/250, Netroute Kstorage 16/24/9, Lag 48/72/24, gwKstorage 50, gwLag 500, inhibit_evap 0. Delineation 427.5 km² vs HYDAT 416 (2.8%) — Copernicus N50/N51 × W117 only (basin drains N/NW, the W116 tiles toward the divide are NOT needed; sketch drainage direction before grabbing tiles). WB FAIL 11.9% (P=678 ET=237 Q=354 mm/yr, coef 0.52, diagnostics empty) = decadal Netroute routing-storage artifact (only SWE+soil_moist output for dS; gw/Sd/ soil_rechr not captured — see Bow/Kaslo WB note), units clean. obs_shape= point_time_series, dag var basinflow_s → NSE/KGE/r/PBIAS all gate-valid.

THE #1 NEW LESSON — a DRY, near-zero-winter-baseflow basin needs the GW store SHRUNK, not the wet-basin heavy store. Every prior validated CRHM HYDAT basin was wet (runoff 683-1337 mm/yr) with sustained winter flow, and the recipe grew a big slow GW reservoir (gw_max 800-2700, gw_init 400-1350) to lift winter baseflow. Kootenay is the opposite: obs Jan-Mar ≈ 0.3 m³/s (basin freezes up, mini

Truncated - read the full file at https://github.com/lzwei196/KISS-Knowledge-Infrastructure-for-Scientific-Simulation/blob/9e5e649e895bc21360df9cb883ae5ef3043effc7/models/CRHM/SKILL.md.