Known limitations¶
SnowMapper is wrong in known, measured ways. This page lists them so you can judge whether a result is fit for your purpose. The numbers come from specific comparisons against observations. They show the size of each effect, not an error bound for your area.
In short
- Most reliable: where snow is, when it arrives and melts, how it varies with elevation and aspect, and anomalies against the model's own climatology.
- Less reliable: absolute snow amounts, valley temperatures, wind, and anything on glaciers.
Precipitation¶
The amount of precipitation is the largest source of error. Each terrain unit receives the precipitation of the coarse weather-model cell it sits in. The downscaling corrects whether it falls as rain or snow, but not how much falls. Global models smooth the mountains, so the amount can be far off:
| Where | What was found | Size |
|---|---|---|
| Dry inner-alpine valley (Zermatt, 1600 m) | wet-side precipitation spread into a rain-shadow valley | ~2.5× too wet |
| High Himalaya (Khumbu, above 4000 m) | precipitation does not fall off with elevation as observed | ~2.5× too wet at high elevation |
| Swiss Alps (Weissfluhjoch, 25 winters) | big winters underestimated | peak SWE varies half as much as observed |
| Nepal (~490 gauges, 2022–2026) | ERA5 wetter than gauges | ~1.3× |
What to do: treat absolute snow amounts as the least certain output. Prefer patterns, timing and anomalies.
Temperature¶
The temperature follows the free atmosphere, not the layer of air near the ground.
- Daily range too small at high elevation: at 5000 m in the Khumbu the modelled day–night range is about a third of the observed.
- Valley cold pools missing: valley floors can be several degrees too warm on clear winter nights. In Central Asian valleys the bias was about +2 K, against near zero above 2500 m.
- Summer days on glaciers can be too cold by a few degrees.
What to do: be careful with frost, daily extremes and melt on valley floors and glaciers.
Wind¶
Wind is an hourly mean over a terrain unit. It runs 20–30 % below station measurements and never reaches gust-type values. Compare wind with its own climatology, not with absolute thresholds.
Timing of showers¶
Global models place convective (showery) precipitation too early in the day. Over the South Asian monsoon the daily peak comes 6–12 h early. This can bias the rain/snow split hour by hour. Daily totals are less affected.
Glaciers¶
FSM1 has no glacier ice. Above the equilibrium line, snow accumulates every year and never melts out; snow depths of several metres on high peaks are an artefact. These units are excluded from anomalies and climatology, but they still appear in the SWE and depth maps. Where the glacier extension (GFSM) is used, ice melt is represented. Melt under warm air on melting ice is driven mostly by radiation, and the turbulent heat flux is underestimated.
Snowpack¶
- Melt-out at mid elevations can be 2–4 weeks late, as found in the Swiss Alps at 1500–2300 m. The pack there is also too deep.
- Density does not respond to the weight of the snow above, so deep packs are too deep for their SWE. SWE itself is not affected.
- Bare ground is too cold in the model, so soil and surface temperatures on snow-free ground should not be used for frozen-ground or permafrost questions.
Terrain units, not points¶
Every value is the average of a terrain unit. A station, village or ski run can differ from its unit's mean elevation and exposure. Differences at single points are expected.
Satellite and data assimilation¶
MODIS sees snow cover, not depth. Where the model is corrected with MODIS, the timing of snow cover improves more than the amount. Forests and long cloudy spells degrade the satellite signal.
Warnings¶
Warning thresholds are provisional unless a board says otherwise, and were calibrated on one season of one region. See Warnings tab.