What is a snow depletion curve, and why does a forecast desk watch it?
A snow depletion curve is just the plot of how much of a basin is still snow-covered, day by day, from peak accumulation down to bare ground. Flat near the top of the season, then it bends over and drops, sometimes fast, sometimes in fits and starts depending on aspect and elevation band. Forecasters have been drawing these by hand off survey data and SNOTEL records for decades. The reason it still matters, even with SWE telemetry everywhere, is that depletion tells you something point measurements can't: how much of the contributing area is still holding snow versus how much has already shed it.
What the curve plots
The y-axis is fractional snow-covered area, usually 0 to 100 percent of the basin or sub-basin polygon. The x-axis is time, typically day of water year. Early season the line sits flat near 100. As melt starts on south-facing low-elevation slopes first, the line starts to sag. By the time you're well into the melt season the curve is dropping steeply, and the point where it crosses some threshold, often 10 or 50 percent, gets called the melt-out date for that basin.
That single number, the melt-out date, is what a lot of streamflow guidance leans on. ESP-style ensemble forecasts, and the simpler index methods some offices still run alongside them, both want to know when the snow stops being a source term in the water balance. A basin that depletes two weeks early changes the shape of the whole runoff hydrograph, not just the peak timing.
Why forecasters use it instead of a single SWE reading
A SNOTEL pillow gives you SWE at one point, usually mid-elevation, often in a clearing picked for instrument siting rather than basin representativeness. That's useful but it's a point. A basin can have 40 percent of its area already bare while the pillow is still reading snow, especially in basins with a wide elevation spread or a lot of south-aspect terrain mixed with north-aspect timber.
The depletion curve fixes that by working at the basin scale. It answers the question a desk actually has, which is not "how much snow is at this one site" but "how much of the drainage is still capable of producing snowmelt runoff." When the curve bends toward zero, that's a real signal that the basin's snowmelt contribution is winding down, independent of what any single instrument says.
Building the curve from a daily fractional snow cover series
The input is a fractional snow cover timeseries, daily snow/no-snow classification turned into a percent-covered number for each basin polygon, plotted over the season. The classification itself comes from optical or multispectral imagery wide enough to cover a basin in one pass and frequent enough that gaps from cloud cover don't blow holes in the curve. That's the practical bottleneck most desks run into: a single seasonal satellite pass or an infrequent field survey gives you two or three points on a curve that needs daily resolution to catch a fast melt-out.
This is the gap a daily classification feed is built to close. Snow Cover Monitoring tracks snow extent basin by basin every day and marks the melt-out date as the curve crosses off, instead of leaving a desk to interpolate between a handful of survey dates.
Where the curve runs into trouble
Cloud cover is the obvious one. A multispectral optical pass can't see through it, so a string of overcast days leaves gaps that get filled by interpolation or by falling back to the last clear look. Shadowed north-facing terrain can also hang onto a thin, non-contributing snow patch long after it's stopped producing runoff, which drags the tail of the curve out longer than the water balance supports. Neither problem is fatal to the method. It just means the curve is read as a trend line, not a precision instrument, and cross-checked against SWE and streamflow when the basin is doing something unusual.
If your basin's curve is the thing standing between a routine forecast and a surprise rise on the river, it's worth seeing what a daily version of it looks like for your drainage.