10 The whole DEM and flood maps

Goal: see how water spreads over a whole area, a city or a valley with several rivers, rather than one watershed. Then make flood maps and an animation.

Settings used: MODEL_AREA, SAVE_FIELDS, FIELD_VARS, OUTPUT_INTERVAL_SECONDS.

With MODEL_AREA: whole_dem, MRRpy skips the outlet and the delineation and routes every cell of the DEM. Water leaves wherever it flows off the edge or into a no-data hole, so make the DEM cover the uphill land that drains into your area. SAVE_FIELDS records maps of depth, velocity and discharge at every output time.

Set it up

Notebook formConfig file + CLIPythonQGIS plugin

1. Terrain → Area to model: whole_dem. No outlet is asked.
2. Precipitation: design storm 30 mm/h for 2 hours.
3. Routing: simulation length 8; Save results every 600.
4. Outputs and run: tick Save maps over time, then run. The notebook's last cells draw the maps (see notebooks/configure_and_run.ipynb).

DEM_PATH: dem_90m.tif
MODEL_AREA: whole_dem
TARGET_CRS_EPSG: EPSG:32645
OUTPUT_DIR: results/whole/
RAIN_INTENSITY_MM_HR: 30.0
RAIN_DURATION_HOURS: 2.0
ROUTING_SCHEME: diffusive_implicit
TOTAL_SIMULATION_TIME_HOURS: 8.0
OUTPUT_INTERVAL_SECONDS: 600
SAVE_FIELDS: true

MRRpy run -c whole.yaml

Then draw the maps in Python (next tab).

from MRRpy import Config, run_pipeline, plot_field, animate_fields, export_peak_maps

cfg = Config(
    DEM_PATH="dem_90m.tif",
    MODEL_AREA="whole_dem",              # no OUTPUT_POINT needed
    TARGET_CRS_EPSG="EPSG:32645",
    OUTPUT_DIR="results/whole/",
    RAIN_INTENSITY_MM_HR=30.0,
    RAIN_DURATION_HOURS=2.0,
    ROUTING_SCHEME="diffusive_implicit",
    TOTAL_SIMULATION_TIME_HOURS=8.0,
    OUTPUT_INTERVAL_SECONDS=600,         # one saved map every 10 minutes
    SAVE_FIELDS=True,
)
out = run_pipeline(cfg)

plot_field(out, "depth")                       # the deepest water each cell reached
plot_field(out, "discharge", time=2.5)         # the flow 2.5 h into the storm
animate_fields(out, ["depth", "discharge"])    # results/whole/depth_discharge_animation.gif
export_peak_maps(out)                          # max_depth.tif, … for QGIS

1. Tab 1 · DEM & Watershed → Area to model: The whole DEM (no outlet), then Use whole DEM.
2. Tab 4 · Routing: tick Save maps over time.
3. Run. On tab 5 · Results, Load Peak Maps adds the peak depth and discharge maps to the project; Save Flow Animation… writes a GIF.

Results

A 30 mm/h, 2-hour storm over the whole 90 m Kathmandu-valley DEM (diffusive_implicit), made with animate_fields(out, ["depth", "discharge"]). Left: water depth; the valley floor ponds. Middle: discharge as river lines that widen with the flow, draining off several edges of the DEM, the main river to the south-west. Right: the flow at the main exit and all water leaving the DEM.

plot_field(out, "discharge"): the highest flow each river reached.

What changes in a whole-DEM run:

• hydrograph.csv: Q_m3s is the flow at the main exit, where the largest river leaves the DEM (the log prints where). An extra column, Q_total_outflow_m3s, is all water leaving the DEM. For flow at other places, add virtual gauges.
• watershed.tif / watershed.geojson keep their names but hold the whole modelled area, so Earth Engine downloads and the QGIS layers work as usual.
• The automatic channel size is estimated at the main exit instead of at an outlet point.

About the maps: animate_fields takes one quantity or a list. Discharge is drawn as river lines along the flow network, wider and brighter with more water (pass lines=False for plain cells); flows below 0.1 % of the run's peak are left out so the rivers stand out (min_value changes that). Every frame uses the same colour scale and shows its time. Pass out_path="flow.mp4" for a video (needs ffmpeg) or every=3 for a shorter GIF. export_peak_maps writes, for each saved quantity, max_<var>.tif and time_of_max_<var>_hours.tif (when the peak came).

How far does the water spread?

The routing passes water from cell to cell along one flow path, so these depth maps show the rivers as a one-cell ribbon. For flood depth and extent beside the rivers, turn on INUNDATION_MAP (Example 12).

Memory

The maps stay in memory until the run ends: about saved times × cells × quantities × 4 bytes. The log prints the estimate and warns above 2 GB. For a large DEM, raise OUTPUT_INTERVAL_SECONDS or FIELD_STRIDE, or save only FIELD_VARS: [depth].

Other options to try

Change	Setting	See
maps for a single watershed	MODEL_AREA: watershed with SAVE_FIELDS: true	6.2
fewer maps, less memory	FIELD_STRIDE: 3, FIELD_VARS: [depth]	6.2
flow at named places	ROUTING_GAUGES	6.1
a faster explicit run on the GPU	ROUTING_SCHEME: kinematic, BACKEND: gpu	6.4