Rain gauges against an elevation-aware climate surface: how much a plain interpolation misses in the HJ Andrews mountains

Against the PRISM grid for 1971-2000, the 13 long-record HJ Andrews gauges agree to 0.5% on average and a distance-only interpolation of them misses only 9 mm (0.4%) over the forest, though 181 mm (7.1%) above 1,300 m. The first PRISM grid, for 1980-89, puts 9.2% more rain on the forest and rises 1.10 mm per metre of elevation against 0.61: most of the 20% storage-gauge shortfall an earlier version of this page reported came from that grid.

How we did Matches except Close 0 of 6 exact, 3 within tolerance, 3 explained all studies
HJ Andrews Fast Facts: average annual rainfall "7 feet (2.1 meters)", against the forest means · tolerance 0.19 m: 7 feet to the nearest foot is 1.98 to 2.29 m; the current forest means sit just above it, at 7.5 to 7.6 feet
  • PRISM 1971-2000 (MS027 ms02727), forest mean2.10 m → 2.30 m +0.2 m (+9.4%)explained why
  • PRISM 1991-2020 normals, 800 m, forest mean2.10 m → 2.31 m +0.2 m (+10.1%)explained why
  • IDW of the 13 long-record gauges, 1971-2000, forest mean2.10 m → 2.29 m +0.2 m (+9.0%)within tolerance why
  • PRISM 1980-89 (ms02714, the first grid, since updated), forest mean2.10 m → 2.51 m +0.4 m (+19.5%)explained why
PRIMET climate diagram (Bierlmaier and McKee 1989): 2,302 mm a year at 426 m, 1972-84 · tolerance 5%: our PRIMET gauge record starts in 1979, so its periods share at most 1980-84 with the diagram
  • PRIMET gauge, water years 1980-892,302 mm → 2,235 mm −67.3 mm (−2.9%)within tolerance
  • PRIMET gauge, its 21 complete water years of 1971-20002,302 mm → 2,247 mm −54.5 mm (−2.4%)within tolerance
+0.5%the 13 long-record gauges against the PRISM 1971–2000 cell holding each (mean of gauge minus PRISM)
[93]
−181 mmIDW of the gauges minus PRISM 1971–2000 above 1,300 m (−7.1%); −9 mm over the whole forest
[82]
1.10 vs 0.61mm of rain per metre of elevation over the forest: the first grid (1980–89) against 1971–2000
[153, 156, 159, 162, 165, 168]
183 of 183steps replay identically in a fresh session

Numbers in brackets, like [14], are step numbers in this study's run.json: the NativeGIS call that returned the figure. NativeGIS 0.3.0, recorded 2026-10-31 15:23 UTC.

Map of the HJ Andrews forest over a hillshade, coloured by IDW of the 13 long-record gauges minus PRISM 1971-2000 on a red-blue ramp centred on zero: red (IDW lower) on the high eastern ridges and around the wettest upland gauge, light blue (IDW higher) across the lower western valleys, with the gauges labelled
IDW of the 13 gauge means (water years 1971-2000) minus PRISM 1971-2000, on PRISM's 50 m cells, over the forest boundary and a hillshade of the 3DEP DEM. Red: the gauges alone give less. Rendered by gis_render in the recorded run, on a diverging ramp centred on zero.

Click or tap a map to open it full screen, then zoom and pan.

What was published

The problem: rain gauges stand mostly in valleys and on roads, while mountain precipitation rises with elevation, so how much does a simple interpolation of the gauges miss? HJ Andrews publishes both sides. MS027 holds Chris Daly's PRISM grids for the forest: the first, for 1980-89 at 100 m (ms02714), and newer ones for 1971-2000 at 50 m (ms02715 to ms02727; the MS027 page says the grids "were updated for a greater time span" and a larger extent). PRISM fits precipitation to elevation and slope orientation from station data and a DEM. MS004 holds the daily precipitation of every gauge with its coordinates. Two published figures anchor the result: Fast Facts gives the average annual rainfall as "7 feet (2.1 meters)", without a place or period, and the PRIMET climate diagram (Bierlmaier and McKee 1989, figure 2) gives 2,302 mm a year at 426 m for 1972-84.

Published data sets and statements by the HJ Andrews Experimental Forest (not a journal paper)Andrews Forest LTER and C. Daly (2015), PRISM precipitation grids for the HJ Andrews Experimental Forest (MS027, doi:10.6073/pasta/772a54d148ad9d1e291e4a951fd43cc2): mean annual grids ms02727 (1971-2000) and ms02714 (1980-89); PRISM Climate Group, Oregon State University, 1991-2020 normals, 800 m; HJ Andrews daily precipitation and gauge inventory (MS004: MS00401, MS00403, MS00404); HJ Andrews "Fast Facts"; Bierlmaier, F.A. and McKee, A. (1989), Climatic summaries and documentation for the primary meteorological station, H.J. Andrews Experimental Forest, 1972 to 1984, USDA Forest Service PNW-GTR-242.

MS027 PRISM grids (EDI) · PRISM 1991-2020 normals (PRISM Climate Group) · Fast Facts (HJ Andrews) · Bierlmaier and McKee 1989, PNW-GTR-242 (PDF) · HJ Andrews data catalogue

Published method. PRISM (Parameter-elevation Regressions on Independent Slopes Model) is a regression of station precipitation on elevation, fitted locally and weighted by distance, slope orientation and terrain. The MS027 metadata does not list which stations entered either fit, but the HJ Andrews meteorological network is cited as a source, so the gauges here are not independent of PRISM.

The data and where it came from

All of it is open: downloaded without an account or a form.

DataWhereLicence
PRISM mean annual precipitation 1971-2000, 50 m grid in millimetres x 100, ArcInfo E00 (MS027, ms02727.e00), NAD83 / UTM zone 10NHJ Andrews LTER, zip, no form or account
https://andrewsforest.oregonstate.edu/sites/default/files/lter/data/studies/spatialdata/ms02727.zip
HJ Andrews data policy (CC BY, acknowledgement requested)
PRISM mean annual precipitation 1980-89, 100 m grid in mm, ArcInfo E00 (MS027, ms02714.e00), NAD83 / UTM zone 10NHJ Andrews LTER, zip, no form or account
https://andrewsforest.oregonstate.edu/sites/default/files/lter/data/studies/spatialdata/ms02714.zip
HJ Andrews data policy (CC BY, acknowledgement requested)
PRISM 1991-2020 normals, annual precipitation, 800 m (30 arc-second) Cloud Optimized GeoTIFF in mm, NAD83 geographicPRISM Climate Group, Oregon State University, plain HTTP download, no form or account
https://data.prism.oregonstate.edu/normals/us/800m/ppt/monthly/prism_ppt_us_30s_2020_avg_30y.zip
PRISM terms of use (free with attribution)
Rain gauge inventory with coordinates and gauge type (MS00401)HJ Andrews LTER, static CSV, no form or account
https://andrewsforest.oregonstate.edu/sites/default/files/lter/fsdbdata/cache/MS00401_v1.csv
HJ Andrews data policy (CC BY, acknowledgement requested)
Daily precipitation, meteorological stations (MS00403) and rain gauge network (MS00404)HJ Andrews LTER, static CSV, no form or account
https://andrewsforest.oregonstate.edu/sites/default/files/lter/fsdbdata/cache/MS00404_v3.csv
HJ Andrews data policy (CC BY, acknowledgement requested)
HJ Andrews Experimental Forest boundary (GI006, boundary83.shp)HJ Andrews LTER, zip, no form or account
https://andrewsforest.oregonstate.edu/sites/default/files/lter/data/studies/spatialdata/gi00601.zip
HJ Andrews data policy (CC BY, acknowledgement requested)
USGS 3DEP elevation on the 1980-89 PRISM grid's 100 m cells (32-bit float GeoTIFF, EPSG:26910)USGS National Map 3DEP elevation image service (exportImage), no form or account
https://elevation.nationalmap.gov/arcgis/rest/services/3DEPElevation/ImageServer
US federal data, public domain

The input files the run read, pinned by checksum

File (relative to the data folder)Bytessha256
samples/hja-open/ms02727/x/ms02727.e005,230,399472e35507996077f…
samples/hja-open/ms02714/x/ms02714.e00457,91570a1a8eb0ed9dc3f…
samples/prism-normals/x/prism_ppt_us_30s_2020_avg_30y.tif64,773,157ffd3c9c775ab8525…
samples/hja-open/gi00601/x/boundary83.shp10,5400774de0a9ebd9693…
samples/dem3dep/hja_prism_grid_26910_100m.tif132,26679495ff02b14ea43…
derived/hja_gauge_water_years_1971_2000.csv29,9091be8e26289a97e47…
samples/hja-open/ms004/MS00401_v1.csv14,826116793f9be8e5bea…
samples/hja-open/ms004/MS00403_v9.csv8,150,3645e33768379f3c98f…
samples/hja-open/ms004/MS00404_v3.csv20,082,4026a943c75c7651032…

Full checksums are in run.json. Replay refuses an input whose checksum differs.

Method

  1. Water years, outside NativeGIS: node tools/repro/hja-precip-gauge-means.js <data folder> reads the two daily files and writes derived/hja_gauge_water_years_1971_2000.csv, one row per gauge and water year (1 October to 30 September) from 1971 to 2000, with the days that carry a value and their sum (days HJ Andrews flags as estimated count). It is a small deterministic script because NativeGIS expressions have no date functions to turn a date into a water year. The file is a download on this page; run the script after the fetch commands below, before replaying. Everything after this step runs in NativeGIS.
  2. Completeness rule, the same per year for both periods: a water year counts when at least 90% of its days carry a value. For 1971-2000 a gauge is kept with at least 15 such years (half the period); for 1980-89 with all 10. A gauge's mean is the total over its complete years divided by their days, times their mean length, which is the total divided by the years when no day is missing. In NativeGIS: field calculator for the share of days, a query for the rule, summarize by gauge, and gis_process join onto the gauge points of the MS00401 inventory, which has coordinates. 46 gauges report in 1971-2000 and 13 are kept; 29 report in 1980-89 and 20 (at 19 sites) are kept, every one with all 3,653 days.
  3. Check on the script: NativeGIS opens both daily files as attribute tables (csv table), keeps each period with a date filter and sums each gauge (summarize); every gauge's day count and sum equal the water-year table's, in both periods.
  4. Open the PRISM grids (gis_open reads the E00 exports; CRS from their PRJ). ms02727 stores millimetres x 100 and is divided by 100. ms02714's outer frame of 540 cells holds -108, a fill value its header does not declare as nodata, so negative cells are set to nodata, leaving 17,760 cells. The 800 m normals are opened as they are, and the forest boundary is reprojected to their NAD83 geographic coordinates.
  5. For each period, reproject the kept gauges to NAD83 / UTM zone 10N and interpolate their means by inverse distance weighting (power 2, 12 nearest) onto that period's PRISM cells, with leave-one-out cross-validation (interpolate crossValidate); subtract PRISM.
  6. Zonal means of each surface and difference over the forest boundary, and over four elevation bands of a 3DEP DEM exported on the 1980-89 grid's cells (reclassify, polygonize, dissolve, intersect with the boundary, zonal).
  7. Grid against grid: resample the 1971-2000 grid onto the 100 m cells (bilinear) and subtract it from the 1980-89 grid. The rise with elevation is the least-squares slope of precipitation on elevation over the forest's 100 m cells, from zonal means of elevation, precipitation and their centred products.
  8. At the gauges: sample PRISM, IDW and elevation in the cell holding each gauge; gauge minus PRISM overall and by gauge type (field calculator, summarize); the 1980-89 gauges also against the 1971-2000 grid.
  9. Render PRISM 1971-2000 and its IDW on one colour scale, and the two differences on diverging ramps centred on zero (gis_render redblue), over a hillshade of the 100 m DEM with the gauges labelled.

The run made 183 NativeGIS calls: gis_open × 9, gis_process × 77, gis_query × 53, gis_close × 36, gis_render × 8.

The numbers

Mean annual precipitation over the forest

Over the Experimental Forest (cells whose centre is inside)CellsMean (mm/yr)Lowest cellHighest cell
PRISM 1971–2000 (Daly, MS027 ms02727, 50 m)25,4792,2971,8972,944
IDW of the 13 long-record gauges, water years 1971–200025,4792,2891,8392,619
IDW minus PRISM 1971–200025,479−9−542+217
PRISM 1991–2020 normals (800 m)1022,3112,0182,709
PRISM 1980–89 (MS027 ms02714, 100 m), the first grid6,3702,5091,9603,519
IDW of the 20 gauges, water years 1980–896,3702,2701,8522,642
IDW minus PRISM 1980–896,370−240−1,201+305
Fast Facts, "average annual rainfall"–2,100––

Zonal statistics over the GI006 boundary [66, 67, 70, 73, 173, 102, 105, 108]; for the 800 m normals the boundary is reprojected to the grid's NAD83 geographic coordinates. Units: millimetres per year (ms02727 stores mm × 100 and is divided by 100 in the run). n = cells, as listed.

By elevation

Elevation band (3DEP, 100 m cells)Mean elevation (m)PRISM 1971–2000IDW 1971–2000IDW minus PRISMPRISM 1980–89IDW minus PRISM 1980–89PRISM 1980–89 minus 1971–2000
below 700 m5992,1142,209+95 (+4.5%)2,153+22 (+1.0%)+39
700–1,000 m8472,1722,267+95 (+4.4%)2,328−97 (−4.2%)+156
1,000–1,300 m1,1422,4292,319−110 (−4.5%)2,697−367 (−13.6%)+266
1,300 m and above1,3992,5592,378−181 (−7.1%)3,012−657 (−21.8%)+459

mm/yr. The 3DEP DEM on the 1980–89 grid's cells, reclassified into bands, polygonized, dissolved and cut to the forest boundary [55, 58]; band means by zonal statistics [60, 76, 79, 82, 111, 117, 146] (1971–2000 on its own 50 m cells, 4,532 / 9,303 / 8,078 / 3,566; 1980–89 on 100 m cells, 1,134 / 2,329 / 2,013 / 894). The last column puts the 1971–2000 grid on the 100 m cells (bilinear resample) [137]. Least-squares rise of precipitation with elevation over the forest's 100 m cells: 1.10 mm per metre for 1980–89 against 0.61 for 1971–2000, from zonal means of elevation, precipitation and their centred products [153, 156, 159, 162, 165, 168].

At the long-record gauges, 1971–2000

GaugeTypeElevation (m)Complete water yearsGauge mean (mm/yr)PRISM 1971–2000 cellGauge minus PRISMLeave-one-out IDW
PRIMET PPTPRI01recording437212,247.52,152+4.5%2,150
GSWS09 PPTGS901standard442222,067.62,062+0.3%1,916
GSWS10 PPTGS001standard475242,246.02,261−0.6%2,095
CS2MET PPTCS201standard481302,311.82,206+4.8%2,141
WS09RG PPTWS901storage637231,832.01,891−3.1%2,113
WS10RG PPTWS001storage639222,002.42,030−1.4%2,235
GSMACK PPTGSM01standard768162,440.92,371+2.9%2,419
GSMACK PPTGSM02recording783222,418.92,371+2.0%2,441
WS3JRD PPTJRD02storage841222,148.42,123+1.2%2,171
H15MET PPTH1501standard902302,259.12,172+4.0%1,941
H15RDG PPTRDG02storage1,104221,878.72,216−15.2%2,269
RDSEND PPTRDS01storage1,330212,618.52,545+2.9%2,335
RS13RG PPTO1301storageoff the DEM193,003.12,896+3.7%2,183

Gauge means from the water-year table, the completeness rule and the join onto the inventory points [14, 87]; PRISM and elevation are the cell holding each gauge. Mean gauge minus PRISM +14 mm (+0.5%), mean absolute 81 mm, root mean square 115 mm, n = 13 [93]; by type [95]: storage −2.0% (n = 6), standard +2.3% (n = 5), recording +3.2% (n = 2). Last column: each gauge predicted by IDW from the other 12 [64]; leave-one-out bias −82 mm, mean absolute error 228 mm, root mean square 307 mm (n = 13) [63].

The 1980–89 gauges against both grids

GaugeTypeElevation (m)Gauge, WY1980–89 (mm/yr)PRISM 1980–89 cellGauge minus PRISM 1980–89PRISM 1971–2000 cellGauge minus PRISM 1971–2000
PRIMET PPTPRI01recording4372,234.72,037+9.7%2,152+3.9%
GSWS09 PPTGS901standard4422,116.81,882+12.5%2,062+2.7%
GSWS10 PPTGS001standard4752,316.02,287+1.3%2,261+2.5%
CS2MET PPTCS201standard4812,265.22,098+8.0%2,206+2.7%
MIRKWD PPTMRK01standard5852,165.92,145+1.0%2,149+0.8%
WS09RG PPTWS901storage6371,887.32,138−11.7%1,891−0.2%
WS10RG PPTWS001storage6391,949.62,322−16.0%2,030−4.0%
SLTRWD PPTSWD01standard7412,203.02,180+1.1%2,199+0.2%
GSMACK PPTGSM01standard7682,574.52,368+8.7%2,371+8.6%
MIDWAY PPTMID01standard7752,047.92,057−0.4%2,053−0.2%
GSMACK PPTGSM02recording7832,399.52,354+1.9%2,371+1.2%
EARTHF PPTEAR01standard8152,193.32,179+0.7%2,186+0.3%
WS3JRD PPTJRD02storage8412,151.02,639−18.5%2,123+1.3%
H15MET PPTH1501standard9022,228.72,253−1.1%2,172+2.6%
H15PND PPTPND01standard9162,286.82,259+1.2%2,155+6.1%
TRAILS PPTTRL01standard1,0582,642.82,691−1.8%2,549+3.7%
RS18RG PPTR1801standard1,0942,350.42,347+0.1%2,313+1.6%
H15RDG PPTRDG02storage1,1041,835.82,409−23.8%2,216−17.2%
RDSEND PPTRDS01storage1,3302,575.03,322−22.5%2,545+1.2%
CARPMT PPTCAR01storage1,4231,906.72,621−27.3%2,398−20.5%

The 20 gauges with every water year 1980–89 complete [22, 122, 134]. Against the 1980–89 grid [130]: storage −20.0% (n = 6, −11.7% to −27.3%), standard +2.6% (n = 12), recording +5.8% (n = 2); overall −113 mm, root mean square 317 mm [128]. Against the 1971–2000 grid [136]: storage −6.6%, standard +2.6%, recording +2.5%. The periods differ, so a few percent either way is climate, not the grid. Leave-one-out IDW of these gauges: root mean square 203 mm, bias +27 mm (n = 20) [98].

The surfaces

Map of PRISM mean annual precipitation 1971-2000 over the HJ Andrews forest: wettest along the eastern ridges near 2,900 mm, driest in the western valley near 1,900 mm, with the 13 long-record gauges labelled
PRISM 1971–2000 (MS027 ms02727), on the same colour scale as the IDW map. Rendered by gis_render in the recorded run.
Map of the IDW surface of the 13 gauge means over the same area: a smooth field between about 1,850 and 2,600 mm with a high spot at the wettest upland gauge
IDW of the 13 gauge means, 1971–2000, same scale. It cannot rise above its wettest gauge.
Map of the first PRISM grid (1980-89) minus the 1971-2000 grid on 100 m cells: near zero in the western valley, deepening blue to the eastern and southern ridges where the first grid is up to about 900 mm wetter, with the 20 gauges of 1980-89 labelled
The first grid (1980–89) minus the 1971–2000 grid, on a ramp centred on zero [138]. Blue: the first grid is wetter. The difference grows with elevation.

What the numbers say, and what they do not

Replay it yourself

The run file records every call, its arguments, the checksums of the files it read and the numbers it returned. Replaying it repeats every call in a fresh NativeGIS session and checks that each answer comes out the same.

Verified reproducible on 2026-10-31: 183/183 steps identical. The run file was replayed in a fresh NativeGIS session (0.4.0 (source 3692d6e)) against a folder holding nothing but the public downloads below, fetched with these exact commands.
  1. Make an empty data folder and fetch the inputs into it exactly as below (bash; the folder layout is the one the run file expects):
    cd /path/to/an/empty/folder
    mkdir -p samples/hja-open/ms02727 && curl -fL -o samples/hja-open/ms02727/ms02727.zip https://andrewsforest.oregonstate.edu/sites/default/files/lter/data/studies/spatialdata/ms02727.zip && unzip -o samples/hja-open/ms02727/ms02727.zip -d samples/hja-open/ms02727/x
    mkdir -p samples/hja-open/ms02714 && curl -fL -o samples/hja-open/ms02714/ms02714.zip https://andrewsforest.oregonstate.edu/sites/default/files/lter/data/studies/spatialdata/ms02714.zip && unzip -o samples/hja-open/ms02714/ms02714.zip -d samples/hja-open/ms02714/x
    mkdir -p samples/prism-normals && curl -fL -o samples/prism-normals/prism_ppt_us_30s_2020_avg_30y.zip https://data.prism.oregonstate.edu/normals/us/800m/ppt/monthly/prism_ppt_us_30s_2020_avg_30y.zip && unzip -o samples/prism-normals/prism_ppt_us_30s_2020_avg_30y.zip -d samples/prism-normals/x
    mkdir -p samples/hja-open/ms004 && curl -fL -o samples/hja-open/ms004/MS00401_v1.csv https://andrewsforest.oregonstate.edu/sites/default/files/lter/fsdbdata/cache/MS00401_v1.csv
    mkdir -p samples/hja-open/ms004 && curl -fL -o samples/hja-open/ms004/MS00403_v9.csv https://andrewsforest.oregonstate.edu/sites/default/files/lter/fsdbdata/cache/MS00403_v9.csv
    mkdir -p samples/hja-open/ms004 && curl -fL -o samples/hja-open/ms004/MS00404_v3.csv https://andrewsforest.oregonstate.edu/sites/default/files/lter/fsdbdata/cache/MS00404_v3.csv
    mkdir -p samples/hja-open/gi00601 && curl -fL -o samples/hja-open/gi00601/gi00601.zip https://andrewsforest.oregonstate.edu/sites/default/files/lter/data/studies/spatialdata/gi00601.zip && unzip -o samples/hja-open/gi00601/gi00601.zip -d samples/hja-open/gi00601/x
    mkdir -p samples/dem3dep && curl -fL -o samples/dem3dep/hja_prism_grid_26910_100m.tif 'https://elevation.nationalmap.gov/arcgis/rest/services/3DEPElevation/ImageServer/exportImage?bbox=557887.02595787,4893065.8803265,572887.02595787,4905265.8803265&bboxSR=26910&imageSR=26910&size=150,122&format=tiff&pixelType=F32&noData=-9999&interpolation=RSP_BilinearInterpolation&compression=None&f=image'
  2. Download run.json (342,297 bytes, sha256 7dbce51cddd2e893be7784850f65380e9d265ec18af5cf48438efd22f36eaabb).
  3. Ask your assistant to replay it with NativeGIS, or call the tool directly:
gis_replay {"action": "replay",
            "file": "<absolute path to the run.json you downloaded>",
            "inputRoot": "<your data folder>",
            "outputRoot": "<a new, empty folder>"}

Paths in the run file are relative to the data folder (inputRoot), and outputs go under outputRoot, so nothing depends on where the original run lived.

Files

Limits

NativeGIS reproductions · every NativeGIS number on this page is read from the recorded run file it links to.