Watershed delineation is one of the most satisfying analyses in QGIS, because it turns a simple elevation model into a map of drainage basins, streams, and outlets that looks like real hydrology. It is also a workflow where each step depends on the previous one, so a mistake early on, such as an unfilled sink or a wrong flow direction, propagates through the whole result. This guide walks through the standard DEM to watershed workflow in the correct order, with the checks that catch errors before they spread.

Begin with a digital elevation model clipped to an area slightly larger than your expected watershed, because watershed boundaries near the edge of a DEM are unreliable where flow data is cut off. Fill sinks in the DEM using the Fill Sinks tool, which removes small depressions caused by data noise that would otherwise trap flow and break the drainage network. Check the filled DEM against the original: the differences should be small and localised, and large filled areas may indicate real lakes or reservoirs that deserve separate treatment rather than blind filling.

Choose a DEM resolution suited to your purpose. Thirty metre SRTM or ALOS data suits catchment scale work, while higher resolution LiDAR data suits small urban or farm catchments but produces much larger files and longer processing times. Reproject the DEM to a projected CRS in metres, such as the local UTM zone, before analysis, because flow tools assume consistent horizontal and vertical units, and a geographic CRS in degrees will distort slope and flow calculations.

Next, run a flow direction tool, which assigns each cell the direction water would flow based on its neighbours, then run flow accumulation, which counts how many upstream cells drain into each cell. High accumulation cells form the stream network, and you extract them by setting a threshold: cells above the threshold become streams, cells below remain hillslope. The threshold controls stream density, so a low threshold produces a dense network of small channels, while a high threshold keeps only major rivers. There is no universal correct threshold; try several values and compare the result with mapped streams or satellite imagery.

Order the stream network if your report needs it, using Strahler ordering, where headwater streams are order one and orders increase where two streams of equal order join. This gives a standard way to describe stream size and compare catchments. Style the streams by order, with wider lines for higher orders, and check that the network looks natural: streams should follow valleys visible in the shaded relief, not cut across ridges or run in straight lines, which would signal a DEM or flow direction problem.

Place your outlet points on the stream network itself, snapped precisely to high accumulation cells, because an outlet placed even one cell off the stream will delineate a tiny incorrect catchment or fail entirely. Use the snap tool or check the accumulation value at your outlet location before running the watershed tool. Then run the watershed delineation tool, which traces all cells draining to each outlet and produces a polygon for each catchment, ready for area calculation, rainfall averaging, or land use analysis.

Validate the result before using it: compare catchment boundaries with topographic maps, check that areas are plausible, and look for catchments that are suspiciously small or oddly shaped. Common fixes include moving the outlet, refilling the DEM, or adjusting the stream threshold, and each fix should be tested one at a time so you know what changed. Done carefully, this workflow produces professional watershed maps from free data, and it teaches the core logic of hydrological modelling that underpins flood studies, water resource planning, and environmental assessment.