Idaho SNOTEL network · 87 active stations

SNOTEL Station Representativeness in Idaho

How well does each snow station stand for the terrain around it, and how much does its snow tell us about the water downstream? A physiographic index compared with repeated airborne LiDAR, a streamflow-informed companion, and where to add new stations.

Central Idaho relief with SNOTEL stations (stars), USGS gauges (circles) and station–gauge pairings

Background

Idaho's 87 active SNOTEL stations are the backbone of statewide snow-water-equivalent (SWE) monitoring. Their daily records feed the water-supply forecasts and reservoir models used by the U.S. Bureau of Reclamation, the NRCS Idaho Snow Survey, the Idaho Department of Water Resources and irrigation districts.

SNOTEL sites are typically placed in small, level forest openings sheltered from wind, where they give a stable index of snow accumulation for water-supply forecasting. Such a site measures one point, and a point measurement need not match the snowpack of the surrounding grid cell or basin (Molotch and Bales 2005). The sub-grid variability of SWE around SNOTEL stations is large and strongly site-dependent (Meromy et al. 2013).

The operational question is therefore which stations remain reliable at basin scale, and at what spatial scale.

The project's three interrelated goals are:

  1. Score every Idaho SNOTEL station on how closely its terrain and canopy resemble the surrounding landscape, at 500 m to 4 km, and check that score against airborne LiDAR snow depth.
  2. Score the same stations on how much unique skill their April-1 SWE adds to forecasts of April–July runoff at downstream USGS gauges.
  3. Combine the two scores into a station-utility map and a mapped priority for where new stations would add the most information.
Static-axis concept sketch
Fig 1.The static axis in concept: read terrain and canopy at the station pixel, compare them with the buffered landscape around it (500 m–4 km), and validate the resulting index against airborne LiDAR snow depth at Mores Creek Summit.

Two ways a station can be "representative"

The static (physiographic) axis is a look-alike test. It asks whether the terrain and canopy at the station resemble the wider area it must stand for. It needs no snow data, so it can be computed for any point in the state from public elevation and canopy layers, and then checked against independent LiDAR snow surveys.

The hydrological axis is an information-value test. It asks whether the station's snow actually predicts the runoff measured downstream. It is defined only where a station can be paired with a downstream gauge that has a long enough joint record.

The two axes turn out not to track each other, so neither can stand in for the other.

NASA SnowEx data at NSIDC

The physiographic index

Terrain metrics come from the 30 m 3DEP elevation model and canopy metrics from LANDFIRE Existing Vegetation Height, both read in buffers of 500 m, 1 km, 2 km and 4 km around every station. Eight components enter the index: seven risk terms — relative elevation, elevation standard deviation, the elevation 95–5 percentile range, slope, southness, relative canopy height and canopy standard deviation — and one benefit term, the fraction of the buffer lying within 100 m of the station’s own elevation. The design follows the physiographic representativeness index of Molotch and Bales (2006), which scored SNOTEL sites in the Rio Grande headwaters from terrain and remotely sensed snow-cover persistence; here the terrain metrics are joined by canopy instead.

Each component is standardised across the stations of the same buffer — a z-score, not a rescaling to 0–1 — and the components are combined in a weighted sum: relative elevation 0.25, elevation SD 0.20, elevation range 0.15, southness 0.15, slope 0.10, canopy height 0.10, canopy SD 0.05, minus 0.15 times the benefit term. That total is then min-maxed across the stations of the same buffer, and the index is one minus it. The index is therefore a rank within the network at one scale: 1 is the station that looks most like its surroundings at that buffer and 0 the least, by construction.

Two end-member stations

Drag the slider to compare the extremes. Island Park (index ≈ 1.0) sits on flat, homogeneous ground with little canopy contrast, so the station pixel looks like everything around it. Hyndman (index ≈ 0.0) sits in steep, dissected relief with strong elevation and aspect gradients inside every buffer.

Hyndman terrain and canopy panels
Island Park terrain and canopy panels
Island Park · index ≈ 1.0Hyndman · index ≈ 0.0
Fig 2.Terrain and canopy around two stations: elevation, relative elevation, slope, southness, canopy height (LANDFIRE EVH) and relative canopy, with the 500 m–4 km buffer rings (red star = station). Left of the divider: Island Park. Right: Hyndman.
Physiographic representativeness index at 4 km on Idaho terrain
Fig 3.Terrain-plus-canopy representativeness index at the 4 km buffer, on the ground that produced it. Dark blue = highly representative, pale = poorly representative. The index is z-scored and min-maxed within the buffer, so it is a rank inside the network rather than an absolute score, and exactly one station reaches each end by construction.

Statewide results

Representativeness depends on scale, as the spatial statistics of snow depth themselves do (Deems et al. 2006; Sexstone et al. 2022). Because the index is min-maxed within each buffer separately, its values are not comparable between scales as levels; what changes is the shape of the distribution. The network mean sits at about 0.56 at 500 m and about 0.42 at 4 km, so at the basin scale more stations fall in the lower part of the network’s own range and the leaders pull away from the rest. That is a redistribution of ranks, not a measured decline in representativeness.

At the 4 km basin scale the stations split into terciles of 29 high, 29 medium and 29 low. The low class is not scattered at random. It clusters in the Clearwater and north-panhandle headwaters and in the Sawtooth, Wood River and Big Lost headwaters.

The highest values sit on the homogeneous margins of the Snake River Plain. The top of the ranking is Mud Flat (1.00), Island Park (0.97) and Long Valley (0.94).


LiDAR validation at Mores Creek Summit

The static index uses no snow data, so it can be checked against an independent snow observation. Mores Creek Summit is an operating SNOTEL station flown repeatedly by NASA SnowEx airborne LiDAR, which gives metre-scale snow depth across the surrounding terrain on many dates. Airborne LiDAR is the usual reference for this kind of check (Grünewald and Lehning 2015; Herbert et al. 2024).

The common LiDAR analysis domain and the campaign-count field behind it
Fig 4.The unit every ratio in this section is computed over. Left: the common domain on the LiDAR DTM, with the buffer radii used in the scale analysis; everything under the grey veil is excluded from every mean. Right: how many of the 28 campaigns see each pixel. Only about half the flown area is valid in all of them, so the domain is a lace over the footprints rather than a study area anyone chose.

Twenty-eight flights, one fixed footprint

Twenty-nine airborne acquisitions span 17 February 2022 to 4 April 2026 and cover accumulation, mid-winter and melt across five winters. They come from two NSIDC collections: seventeen SNEX_MCS_Lidar campaigns and twelve NIVAL_MCS_Lidar campaigns flown in winter 2025-26, which roughly double the sampling of the early season.

Each flight covers a slightly different area, so the analysis uses a fixed common domain of pixels valid in every retained campaign. The 13 February 2024 flight is excluded because its anomalous footprint shrank the shared area. The twenty-eight retained campaigns leave a common domain of 68.6 million pixels at 0.5 m, 17.14 km². Adding the 2025-26 season costs only 3.8 % of the domain the sixteen retained SNEX campaigns share.

That domain is a lace rather than a patch: a pixel has to be valid in all twenty-eight campaigns, and the dropouts do not coincide, so only about half the flown area survives the intersection. Its shape is a property of the flight footprints, not a study area anyone chose.

LiDAR elevation, canopy, slope, southness, mean snow depth and persistence over the common domain
Fig 5.What the LiDAR measures on the common domain: elevation, canopy height, slope, southness, mean snow depth over the 28 campaigns and the fraction of campaigns with snow. The station (red star) carries its own value in each panel title. Canopy here is the 0.5 m LiDAR CHM, not the 30 m LANDFIRE product used state-wide.

Where the station sits

The station sits about 271 m below the mean elevation of the common domain. This structural offset sets the mean level of the ratio and is the main reason the station under-samples the surrounding snowpack at broad scales.

It does not explain the changes from one flight to the next, which are seasonal.

Campaign timing, ratio and air temperature against the SNOTEL record
Fig 6.Every campaign placed in the snow year, on one date axis. Top: the station's own daily snow depth and SWE, with the LiDAR station pixel and the LiDAR domain mean on each flight date; the grey stems between them are the quantity this section is about. Middle: the resulting ratio. Bottom: air temperature, which is what makes a campaign an onset or a melt one. Separate panels rather than twinned axes, so the crossings of metres and degrees cannot be read as meaning.
Representativeness ratio against buffer radius, one line per campaign
Fig 7.Representativeness ratio against buffer radius, one line per campaign, coloured by season. The heavy line is the median over campaigns. The departure from 1 is not monotone in the radius. At 50 m the mean ratio is 1.01 and over the whole domain 0.90; in between, from 250 m to 2 km, it rises to about 1.23 — the station holds roughly a quarter more snow than a buffer that is mostly the forest its clearing sits in. The near field and the far field ask different questions, and the middle is where the point and the area disagree most. Dotted grey: the two campaigns with too little snow for a ratio to mean anything — 13 November 2023, bare at the station, and 3 December 2025, with 8 cm — which are kept out of the median.

The observed representativeness ratio

For each campaign and buffer the benchmark is the ratio of the snow depth measured at the station pixel to the mean snow depth over the surrounding unit: depth(station pixel) / mean depth(buffer). The numerator is a single 0.5 m pixel, read at full resolution; the denominator is the mean over the common domain or over a disc of the given radius.

A ratio near 1 means the station holds as much snow as the ground it stands for. Above 1 the station sits in a locally deeper patch than its surroundings; below 1 the wider terrain holds more snow than the station does. Because no snow data enter the index, agreement between the two is an out-of-sample test, not a fit.

Representativeness is seasonal. The median ratio is 0.76 at onset, 0.91 in mid-winter and 0.92 during melt, so the station under-samples its domain most when the snowpack is building. Two onset campaigns are excluded because there is too little snow for a ratio to mean anything: 13 November 2023, when the station pixel is bare and the ratio is zero by construction, and 3 December 2025, with 8 cm at the station and 10 cm over the domain. Both are facts about the snowpack rather than about the siting, so they are drawn but kept out of the statistics. The lowest meaningful ratio is 0.67, on 28 December 2023. Winstral and Marks (2014) found the spatial pattern around an index site broadly stable from year to year over a long record, and Herbert et al. (2024) report that the direction of the station bias persists at many of the sites they reanalyse. This site is consistent with both: over the common domain the direction of the mismatch is predominantly the same — the station holds less snow than its domain — while its magnitude swings with the season.

What controls the snowpack

Mean snow depth rises with elevation in every campaign, and the gradient steepens toward melt. North-, east- and west-facing slopes hold clearly more snow than south-facing ones, where radiation speeds ablation. Canopy is not monotone, and reading it took fixing the classes first. With every flight binned by the same canopy model, snow is deepest not in the open but in the 2–5 m class — and thins both towards open ground and under tall canopy, in every campaign. Forest-edge effects are a possible reading — Currier and Lundquist (2018) measured deeper snow near forest edges across several western climates — but they defined an edge by distance to trees and by vegetation under 2 m, not by a 2–5 m height class, so this is our result and that is an interpretation of it, not a confirmation. The earlier panel could not show that, because each flight was binned by its own canopy model: on a fixed domain whose forest does not change, the campaign-mean canopy height ranges from 0.8 to 3.1 m and its spread from 2.5 to 5.8 m, and the two collections do not overlap at all. That variation is in the vegetation-height processing, not in the trees, which is why the driver audit still refuses canopy as a driver of the ratio even though the class gradient within a flight is real. Slope has only a weak effect.

Mean snow depth by elevation, aspect and canopy class, one line per campaign
Fig 8.Mean LiDAR snow depth over the common domain in fixed classes of elevation, aspect and canopy height; one line per campaign, melt dashed, and dotted grey for the two campaigns with too little snow for a ratio. Only the snow depth varies between lines. The classes come from the 17 February 2022 rasters and are held fixed, so a cell is in the same class on every flight and no campaign can have a class another lacks — which is what made the earlier version of this panel unreadable, since the two LiDAR collections process the canopy model differently.

What moves the ratio between campaigns is timing in the water year and how variable the snow depth itself is over the domain. The station's own terrain cannot: on a fixed domain it is the same on every flight.

Static proxy versus LiDAR observed ratio
Fig 9.The snow-free index (bottom axis) against the LiDAR-observed ratio, one point per campaign per buffer radius, coloured by season. The top axis gives the radius each index value corresponds to — at one station they are the same axis, which is the point of the figure. The relationship is not monotone: both ends sit near a ratio of 1 and the middle scales depart from it, and the rank correlation with |1 − ratio| is −0.17.

What one site can and cannot test

Plotting the index against the observed ratio looks like a validation and is not one. At a single station the index takes one value per buffer radius, so along this axis “index” and “radius” are the same variable and this is the scale plot with a relabelled x axis. The collinearity is the reason the weights cannot be calibrated from one site, and it is the reason the figure is kept: it makes the limit visible rather than hiding it behind an apparent agreement.

The shape is worth reading on its own terms, because it is not what a simple “higher index, better match” story would predict. The ratio is close to 1 and tightly clustered at the smallest radius (median 1.01 at 50 m, interquartile range 0.07) and again at the largest buffer (median 0.95 at 4 km, interquartile range 0.10; over the common domain, which is a different area from the 4 km disc, the median is 0.91 with an interquartile range of 0.09), and departs most in between: at 250 m to 2 km the station holds about 21–25 % more snow than its surroundings, with three times the spread. The rank correlation between the index and |1 − ratio| is −0.17 — no monotone relationship. Physically that is readable: at 50 m the buffer is the clearing itself; at a few hundred metres to a couple of kilometres it is mostly the forest the clearing sits in, which holds less snow; at 4 km it reaches the higher ground that holds more, and the two effects cancel.


How robust is the index?

The baseline weights are expert-informed, so a fair question is whether a different analyst, choosing the weights differently, would sort the stations differently. Two stress tests at the 4 km buffer say the weights are not what decides the ranking.

Fig 10.(a) Left, the fraction of 1000 Monte-Carlo weight draws (±25 % per weight) that keep each station’s tercile; right, each station’s baseline rank with its 90 % interval, coloured by what that interval actually does. (b) The same classes on the map, which answers a different question: whether the stations the weighting leaves uncertain are concentrated anywhere or scattered through the network.

Weights

Across 1000 weight sets, 61 of 87 stations keep their class in every draw and 73 keep it in at least 90 % of draws.

Ranks

Ranks are stable at the extremes: 21 stations are robust-top and 20 robust-bottom. Of the other 46, 37 keep their whole 90 % interval inside the middle half of the ranking: stably intermediate, which is a result and not an instability — Banner Summit, for instance, stays between ranks 35 and 41 of 87. Only 9 have an interval that actually crosses a quartile boundary, and only those are in doubt.

Dimensionality

The eight index inputs — seven risk terms and the benefit term — span five effective dimensions: PC1–PC5 explain 0.43, 0.19, 0.15, 0.10 and 0.09 of the variance, so five components are needed to reach 90 %. The only redundant pair is elevation SD and elevation range, which both measure terrain heterogeneity.


The hydrological axis

The static axis asks whether a station looks like its basin. The hydrological axis asks whether its snow predicts the water downstream. A station can look unlike its basin and still carry the signal that matters for runoff, and the reverse is also possible.

Concept: does SNOTEL SWE predict streamflow at a downstream gauge
Fig 11.The hydrological axis in concept: does the station's April-1 SWE predict April–July runoff at a paired downstream gauge?
SNOTEL to USGS gauge pairings over Idaho
Fig 12.The 100 scored station–gauge pairs across 51 gauges. Each line joins a station to a gauge downstream of it on the NHDPlus network, within 200 km, draining ≤ 5000 km² and with ≥ 25 water years of data; the station must fall inside the gauge's upstream basin. Line weight and colour both carry R², so the ranking survives in greyscale. Grey dots without a line are stations no gauge could be paired to.

Pairing stations with gauges

Pairing follows the river network, not straight-line distance. For each station, the USGS NLDI service locates its reach on NHDPlus and navigates downstream along the main stem for up to 200 km.

Candidate gauges are screened by drainage area (≤ 5000 km²) and by the number of water years that actually carry data, and a scored pair needs at least 20 joint water years between WY 1985 and 2024. The station must also fall inside the gauge’s upstream basin, which removes pairs that are downstream in the flow table but outside the contributing area on the ground. All but one station find a downstream gauge. After screening, 56 stations carry a scored pairing: 100 station–gauge pairs across 51 gauges.

The 31 unscored stations are a record-length and gauge-availability gap, not a verdict on those stations. Their amber and red stars cluster in the Clearwater/panhandle and along the southern margin, where local drainage is effectively ungauged.

Distance is a weak signal at best. Across the 100 scored pairs the rank correlation between station–gauge distance and R² is −0.09; the mean R² is 0.61 within 40 km against 0.43 beyond 80 km, but only six pairs sit beyond 80 km, so that contrast rests on very little. Position on the network matters more than separation.

Strongest and weakest station-gauge pairing
Fig 13.The strongest and the weakest scored pairing, drawn: Swede Peak against USGS 13147900 (R² = 0.84 over 40 water years) and, beside it, a pairing near the bottom of the ranking. April-1 SWE on the x axis, April–July runoff volume on the y axis, one point per water year.

Two hydrological scores

hydro-A is the mean R² of a station's SWE→runoff regressions across its gauges. It measures raw predictive strength.

hydro-B is the mean leave-one-station-out ΔR² in a multi-station regression: how much explained runoff is lost when the station is removed. It measures the unique skill a station adds beyond the rest of the network, and generalises the predictor-selection logic of NRCS water-supply forecasting (Garen 1992).

The strongest single pairing is Swede Peak against USGS 13147900 (R² = 0.84 over 40 water years), with Galena Summit and Atlanta Summit just behind. Ranked by hydro-A the leaders are Atlanta Summit (0.83), Swede Peak (0.83) and Banner Summit (0.81). The largest unique contributions come from different stations again — Deadwood Summit (ΔR² = 0.24), Sedgwick Peak (0.20) and Soldier R.S. (0.16) — and that the two lists do not overlap is the point.

hydro-A versus hydro-B scatter
Fig 14.Stand-alone strength (hydro-A) versus unique added skill (hydro-B) for the paired stations. The two are barely related (Spearman ρ = 0.14, n = 43, p = 0.37): much of a strong station's signal is already supplied by its neighbours.

Static index and hydrological skill, side by side

Mapped together, the two axes do not mirror each other. Stations that look most like their surroundings are not the ones whose snow best predicts runoff.

The physiographic index and hydro-A as two matched maps of Idaho
Fig 15.Left: how closely each station's terrain and canopy resemble the landscape it stands for, at 4 km. Right: how well its 1-April SWE predicts April–July runoff at the gauges below it. Both panels share the backdrop, the extent, the marker and the ramp, so any difference between them is in the data and not in the drawing. The 31 crosses are stations no downstream gauge could be paired to: unmeasured on that axis, which is a different statement from low.
The two axes of station utility, one panel each
Fig 16.The central result, one panel per axis because the two are measured on different samples. Left: hydro-A, mean R² of SWE against runoff (n = 56). Right: hydro-B, the leave-one-out ΔR² (n = 43). Each panel is split at the medians of the points it actually shows, and carries its own ρ, approximate confidence interval and p. Marker shape repeats the quadrant, so identity never rests on colour alone.

Combining the two axes

The static index and hydro-B correlate at Spearman ρ = 0.07 (n = 43, p = 0.66, approximate 95 % interval −0.24 to 0.36). In this sample there is no statistically significant monotone association between physiographic fidelity and marginal downstream information. That is a failure to detect a relationship, not a demonstration that none exists: the interval is wide enough to contain associations of moderate size, and it assumes the 43 stations are independent observations, which stations sharing gauges and nested basins are not. The nominal interval should therefore not be read as exact; quantifying the effect of that dependence would need a method that accounts for the shared gauges, and we have not done it. The operational consequence is firm all the same — nothing here licenses using one score as a stand-in for the other, so one index cannot rank the network.

Against hydro-A the correlation is negative and, on this sample, distinguishable from zero: ρ = −0.33 (n = 56, p = 0.012, interval −0.55 to −0.08). Stations that look most like their surroundings tend to have the lower mean R² against downstream runoff. Part of that is basin geometry rather than physiography — a station predicts runoff well when its basin is small and snow-dominated, which is a different property from being typical of the ground around it.

Each quadrant points to a different action:

Representative & informativeNetwork keystones.
13 stations
Unrepresentative but informativeKeep: they still predict downstream flow.
9 stations
Representative, lower marginal contributionLocally faithful; much of their signal is already supplied by others.
9 stations
Lower scores on both axesCandidates for review or supplemental monitoring.
12 stations

Where to add new stations

Judging the stations that exist and choosing sites for new ones are two different problems. A station that scores poorly as an index site is not, for that reason, a good place to put another one. A new station is most useful where two things hold at once: the existing network constrains the snow field poorly there, and snow there contributes strongly to downstream flow. This is the maximum-information, minimum-redundancy principle of monitoring-network design (Li et al. 2012; Mishra and Coulibaly 2009), applied here geostatistically rather than through entropy.

The score is a heuristic priority, built in the spirit of information-based network design rather than derived from it. It multiplies two terms: how much new snow information a site would carry, measured as the posterior variance of a Gaussian process at that cell (Zimmerman 2006; Krause et al. 2008), and how strongly snow there loads onto runoff, measured as the interpolated hydro_B of the nearest stations that carry it. Under a linear-Gaussian model a gain of that shape factorises into a product, which is why the two multiply and neither needs a weight; the analogy is close enough to motivate the form and not close enough to make the product an estimate of forecast improvement. The kriged loading in particular is an interpolation of a measured quantity, not the coefficient that a formal value-of-information calculation would require. Raleigh et al. (2025) support the premise that where you observe matters more than how much; they do not validate this score. One declared parameter, β, sets how far the product is traded against terrain the network under-samples.

One elevation rule, stated as eligibility. A candidate cell must reach 1000 m; below that it is not a candidate. Above the floor nothing re-weights by elevation — a cell at 1100 m and a cell at 2600 m are judged on coverage, loading and novelty alone. It is a floor and not a ramp on purpose: a ramp would keep penalising by elevation inside the band it is meant to admit, and the rain–snow transition is precisely where a warming winter changes the hydrology first (Klos et al. 2014; Godsey et al. 2018, on eleven years of transition-zone observations at Johnston Draw in Reynolds Creek, Idaho). In a warm winter a high station can still record a healthy peak SWE while much of the basin below it is taking rain.

1000 m is a judgement, and here is the reason for it. With no floor at all the shortlist puts four sites between 445 and 955 m, in canyon and irrigated-valley country: the lower Clearwater near Lewiston, the Weiser valley, the Magic Valley. Their downstream loading is 0.18–0.30, among the lowest such a list produces, so they are chosen because no station is near them rather than because their snow drives any river. With respect to the objective this project chose — monitoring a seasonal snowpack — those cells are artefacts of a coverage term computed over ground the network was never built to watch; they are not artefacts merely because they are low. The floor removes that ground and costs 15 % of the state. It is an exploratory eligibility choice and not a mapped boundary of the rain–snow transition: that boundary varies with local climate, with the season and with the individual storm (Klos et al. 2014; Godsey et al. 2018), and nothing here locates it.

U · coverage gap posterior variance, 50 km L · downstream loading kriged hydro-B, 43 stations N · terrain novelty distance in predictor space voi = U · L a product, no weights ÷ its 99th percentile priority β·voi + (1−β)·N β = 0.5; floor 1000 m greedy pick 12 sites, 10 km apart re-score after each the field is updated before the next pick
Fig 17.How the siting score is built. The coverage gap and the downstream loading multiply into a value of information. That product is divided by its 99th percentile over the state so it can be compared with terrain novelty, and β sets the balance between the two. Nothing else re-weights the score. A flat eligibility floor at 1000 m decides which cells are candidates at all. Sites are chosen one at a time, and the field is re-scored after each pick, so the second site is not simply the second-best cell of the first hotspot.

Three decisions the map depends on

The correlation length is fitted from the stations. The coverage gap is a posterior variance, and its one parameter is the distance over which snowpack anomalies stay alike. We estimate it from the station records: for every pair, the correlation of their 1-April SWE against their separation. Each year’s state-wide signal is removed first, and that step matters. Without it the pair correlations stay near 0.35 even at 700 km, which only reflects that a wet winter is wet across the whole state, and the fit returns 202 km. With the year signal removed the correlation crosses zero near 150 km and the fitted length is 50 km (1711 pairs, 59 stations).

The network does not stop at the state line. A posterior variance is always largest at the edge of the data it is given. Computed from Idaho stations alone, it reported a gap two kilometres from the Utah border, with a Utah SNOTEL twenty kilometres beyond it. The gap is therefore conditioned on 308 stations: the 87 in Idaho plus 221 within 150 km of the state line in Montana, Wyoming, Utah, Nevada, Oregon and Washington. Candidate cells still have to lie inside Idaho.

The grid has to be square. 3DEP is asked for 500 m but receives a bounding box in degrees, and for Idaho it returns a projected raster whose pixels are 371 m by 446 m. Slope computed on that grid, or on a reprojection of it to degrees, is wrong in one direction. Everything here is computed on a true square 500 m grid in EPSG:5070, 866,738 cells inside the state. Longitude and latitude are derived from that grid for the mask and the maps, and are never fed back into a calculation.

Maps of Idaho: coverage gap, downstream loading, terrain novelty, and the elevation gate that is no longer applied
Fig 18.The three fields the score is made of, each rescaled to [0, 1] over Idaho, and — fourth panel — the floor on the candidate set: a cell must reach 1000 m to be eligible, and 15 % of Idaho does not. The floor is an eligibility rule, not a weight. Filled dots are Idaho SNOTEL, open dots the out-of-state stations the coverage gap is also conditioned on. The gap and the novelty fall to zero at an existing station by construction, which is the check that each field behaves as intended.
Value of information and priority maps of Idaho with the twelve proposed sites
Fig 19.Left: the part of the score that follows from the data alone. Right: the full score, with the shortlist numbered in the order the greedy search chose it. The difference between the two panels is what the terrain-novelty term adds. The right panel is zero below the 1000 m candidate floor; above it nothing re-weights by elevation, so a cell in the rain–snow transition band competes on coverage, loading and novelty like any other.

What it returns

Of the 866,738 Idaho cells that carry an elevation, 737,016 reach the 1000 m floor and compete. The shortlist keeps its mountain core and adds a transition-zone band the old score could not reach. Five of the twelve sites are in the central Salmon River Mountains and the Frank Church country, between 44.6° and 45.2°N; two more are in the Lost River and Lemhi ranges and one in the southeastern highlands near Portneuf. The leading site is at 45.18°N, 115.25°W, 1785 m, where the coverage gap is 0.40 and the downstream loading 0.69. The other four are on the margins: 1162 m above the lower Clearwater, 1020 m on the north side of the Magic Valley, and two in the Owyhee uplands at 1237 and 1505 m.

By elevation band the twelve split six above 2000 m, three between 1500 and 2000 m and three between 1000 and 1500 m; the lowest sits at 1020 m and the mean at 1793 m. Run without any floor, the same score keeps ten of these twelve and puts four valley-bottom cells below 1000 m in place of the other two, so the floor is doing exactly the one job it is there for and nothing else.

β moves part of the shortlist, and that should be said plainly. Matched one to one, nine of twelve sites at β = 0.25 and ten of twelve at β = 0.75 fall within 25 km of a β = 0.50 site; the median matched separation is 0 km in both cases, so the sites that agree are the same cells, but two to three move a long way. By half-degree cell the overlap with β = 0.50 is Jaccard 0.64 and 0.46. The mountain core survives all three values of β; the picks between 1000 and 1500 m are the ones that move.

What it does not say

Terrain novelty does little to order the shortlist. Eight of the twelve sites sit at the top of its scale. Only 1.0 % of Idaho cells reach that level, so those eight really do lie in the most unusual one per cent of terrain; but with most candidates tied at 1.00, the term mainly separates eligible cells from ineligible ones and leaves the ordering to coverage × loading. The term is also terrain only. The two canopy predictors of the representativeness index have no grid-wide counterpart, so a cell whose vegetation is unlike anything the network samples is not flagged for that reason.

Half of Idaho lies beyond the loading measurements. hydro_B exists at 43 of the 87 stations. 51 % of grid cells lie further than one correlation length from any of them, and there the loading field is the network mean rather than a measurement, so a high score in those cells rests on coverage and novelty alone. The interpolation also returns negative values over 13.4 % of the state, although no station measures one. A loading is a weight on information and cannot be negative, so those cells are set to zero before the field is scaled, and the count is reported rather than absorbed into the rescaling.

A low site needs instrumentation that resolves the precipitation phase. Three picks sit between 1000 and 1500 m, where the snow pillow will read zero for part of the winter — and zero on its own does not distinguish rain from no precipitation from snow that has already melted. A standard SNOTEL already records SWE, precipitation and air temperature together, which resolves much of that; what these sites would need is that the three are read jointly, and ideally a sensor that identifies the phase directly.

Calling them transition-zone sites would be a claim we have not made. They are low relative to the rest of the shortlist, which is a statement about elevation, not about snow regime. Placing a cell in the rain–snow transition needs a climatic classification — snow fraction of precipitation, or an observed phase record — and this analysis contains none. The same caution applies to calling the valley-bottom cells the floor removes artefacts: they are artefacts with respect to the objective of monitoring a seasonal snowpack, which is the objective we chose, not because they are low.

The 1000 m floor is a judgement, and it is the only one of its kind in the score. It is an exploratory eligibility choice, not a mapped boundary: it excludes ground where a snow pillow would read zero for most of the winter, and it is not derived from any rain–snow classification. It is not fitted to anything and its sensitivity is not reported here. Moving it would move the low end of the shortlist; the mountain core would not notice.

No feasibility filter. Access, land status, wilderness designation and power are not considered, and the mountain picks fall in country where the first three would be the binding constraint. The map says where a station would be most informative; whether one can be built there is a separate question.


Recommendations

Representativeness is a question of station, scale and season, not a fixed label. The network-mean index falls from about 0.56 at 500 m to about 0.42 at 4 km, and the observed ratio departs most from 1 at onset and during melt.

Use the physiographic index as a first-order screen, not an automatic verdict. It is robust to the weights. Its best use is to rank stations and flag candidates for LiDAR or field follow-up. It has been compared with airborne LiDAR at one site, which is a check and not a validation of the state-wide ranking, so it is a screen rather than a measurement.

Judge stations on both axes. We detect no monotone relationship between physiographic fidelity and downstream information value, over an interval wide enough to leave a moderate association possible, so a station that looks unrepresentative may still be one of the most useful for forecasting. Stations that score low on both deserve review first.

Focus new monitoring on the central Salmon River Mountains and the Frank Church–Gospel Hump country, where the coverage gap and the downstream loading are high together, and secondarily on the Lost River and Lemhi ranges. Both are terrain the present network constrains poorly and whose runoff it predicts least well; neither recommendation survives a feasibility review unexamined.

Treat the rain–snow transition band as part of the network, not as ground outside it. Six of the twelve priority sites fall below 2000 m and three below 1500 m. These are the elevations a warming winter changes first, and the ones a score built to find persistent snow will never propose. They need a site that records precipitation and air temperature alongside SWE, because there the absence of snow is the measurement.

Next steps for the index itself: explicit wind and forest-gap process terms, a snow-state rather than calendar seasonal variable, naturalized flows for regulated reaches, and a canopy term in the siting novelty score, plus a feasibility and land-status filter for proposed sites.

Citations

The seventeen works cited in the text above, covering the point-to-area problem, physiographic controls on snow, scale dependence, LiDAR as a reference, and forecast value and monitoring-network design.

Currier, W. R., and Lundquist, J. D. (2018). Snow depth variability at the forest edge in multiple climates in the western United States. Water Resour. Res., 54, 8756–8773. doi:10.1029/2018WR022553

Deems, J. S., Fassnacht, S. R., and Elder, K. J. (2006). Fractal distribution of snow depth from lidar data. J. Hydrometeorol., 7, 285–297. doi:10.1175/JHM487.1

Garen, D. C. (1992). Improved techniques in regression-based streamflow volume forecasting. J. Water Resour. Plan. Manage., 118, 654–670.

Godsey, S. E., Marks, D., Kormos, P. R., Seyfried, M. S., Enslin, C. L., Winstral, A. H., McNamara, J. P., and Link, T. E. (2018). Eleven years of mountain weather, snow, soil moisture and streamflow data from the rain–snow transition zone – the Johnston Draw catchment, Reynolds Creek Experimental Watershed and Critical Zone Observatory, USA. Earth Syst. Sci. Data, 10, 1207–1216. doi:10.5194/essd-10-1207-2018

Grünewald, T., and Lehning, M. (2015). Are flat-field snow depth measurements representative? A comparison of selected index sites with areal snow depth measurements at the small catchment scale. Hydrol. Process., 29, 1717–1728. doi:10.1002/hyp.10295

Herbert, J. N., Raleigh, M. S., and Small, E. E. (2024). Reanalyzing the spatial representativeness of snow depth at automated monitoring stations using airborne lidar data. The Cryosphere, 18, 3495–3512. doi:10.5194/tc-18-3495-2024

Klos, P. Z., Link, T. E., and Abatzoglou, J. T. (2014). Extent of the rain–snow transition zone in the western U.S. under historic and projected climate. Geophys. Res. Lett., 41, 4560–4568. doi:10.1002/2014GL060500

Krause, A., Singh, A., and Guestrin, C. (2008). Near-optimal sensor placements in Gaussian processes: theory, efficient algorithms and empirical studies. J. Mach. Learn. Res., 9, 235–284.

Li, C., Singh, V. P., and Mishra, A. K. (2012). Entropy theory-based criterion for hydrometric network evaluation and design: maximum information, minimum redundancy. Water Resour. Res., 48, W05521. doi:10.1029/2011WR011251

Meromy, L., Molotch, N. P., Link, T. E., Fassnacht, S. R., and Rice, R. (2013). Subgrid variability of snow water equivalent at operational snow stations in the western USA. Hydrol. Process., 27, 2383–2400. doi:10.1002/hyp.9355

Mishra, A. K., and Coulibaly, P. (2009). Developments in hydrometric network design: a review. Rev. Geophys., 47, RG2001. doi:10.1029/2007RG000243

Molotch, N. P., and Bales, R. C. (2005). Scaling snow observations from the point to the grid element: implications for observation network design. Water Resour. Res., 41, W11421. doi:10.1029/2005WR004229

Molotch, N. P., and Bales, R. C. (2006). SNOTEL representativeness in the Rio Grande headwaters on the basis of physiographics and remotely sensed snow cover persistence. Hydrol. Process., 20, 723–739. doi:10.1002/hyp.6128

Raleigh, M. S., Small, E. E., Bair, E. H., Wobus, C., and Rittger, K. (2025). Snow monitoring at strategic locations improves water supply forecasting more than basin-wide mapping. Commun. Earth Environ., 6, 665. doi:10.1038/s43247-025-02660-z

Sexstone, G. A., Fassnacht, S. R., López-Moreno, J. I., and Hiemstra, C. A. (2022). Subgrid snow depth coefficient of variation spanning alpine to sub-alpine mountainous terrain. Cuad. Investig. Geogr., 48, 79–96. publisher page

Winstral, A., and Marks, D. (2014). Long-term snow distribution observations in a mountain catchment: assessing variability, time stability, and the representativeness of an index site. Water Resour. Res., 50, 293–305.

Zimmerman, D. L. (2006). Optimal network design for spatial prediction, covariance parameter estimation, and empirical prediction. Environmetrics, 17, 635–652. doi:10.1002/env.769

Credits

This story summarises the physiographic index, its comparison with airborne LiDAR at Mores Creek Summit and the hydrological axis, as rebuilt in October and November 2026. The multi-site transfer test is being redone and is not reported here.

Author
Irene Cionni, Boise State University (LEAF research group)
Figures
Project notebooks and technical report (July–November 2026)
LiDAR, 2022–2025
Ciafone, S., O’Neel, S., Adebisi, N., Zikan, K., Enterkine, J., Van Der Weide, T., Wadhwani, Y., Wilder, B., Hoppinen, Z., Filiano, D., and Marshall, H. (2024). SnowEx Mores Creek Summit (MCS) Airborne LiDAR Survey, Version 1 [SD, DTM and CHM products; 16 campaigns, Feb 2022 – May 2025]. NASA NSIDC DAAC. doi:10.5067/NE8CLN3ES5XQ
LiDAR, 2025–2026
Van Der Weide, T., Ciafone, S., Wadhwani, Y., Alabi, I., O’Neel, S., and Marshall, H. (2026). NIVAL Mores Creek Summit Snow Depth, Canopy Height, and DTM from Airborne LiDAR, Version 1 [12 campaigns, Dec 2025 – Apr 2026]. NASA NSIDC DAAC. doi:10.5067/DPFDH2M49DQG
Station & flow data
NRCS Air and Water Database (SNOTEL), station inventory retrieved 2026; USGS NWIS daily values, WY 1985–2024; USGS NHDPlus and NLDI for the station–gauge navigation
Terrain & canopy
USGS 3DEP 30 m elevation (and the 500 m resampling used for the siting grid); LANDFIRE Existing Vegetation Height [LANDFIRE version to be stated]