Climate change — watershed modeling
What does a warmer climate do to a river's water?
We run a basin's SWAT+ model on downscaled projections of the past and the future from the same climate model, and compare the two. On one Michigan basin, 21 climate models agree on the direction of most changes and disagree on their size, and this page shows both.
21 climate models, one basin
Rogue River, Michigan (USGS 04118500, 671.48 km² modeled). Each dot below is one downscaled CMIP6 climate model under SSP2-4.5, run through one calibrated SWATGenX SWAT+ model. It shows the change from 1985–2014 to 2071–2100.
In 20 of the 21 climate models, the center of the year’s streamflow arrives earlier — by a median of about 12 days (range −22.2 days to +0.8 days).
Source: 21-model results (CSV)Outlet streamflow rises in 19 of 21 models, by a median of +16.1 %. The models range from −10.7 % to +32.6 %, and 2 show a decline. For this quantity the spread is the result.
Source: 21-model results (CSV)The highlighted dot, ACCESS-CM2, is among the warmest of the 21 models, yet its streamflow change ranks 13th of 21, below the median. How warm a climate model runs does not decide what it does to the river, which is why one climate model is not enough.
Source: 21-model results (CSV)Precipitation in the projections changes by a median of +8.4 % (range −3.2 % to +13.5 %).
How it works
Downscaled CMIP6 daily precipitation and temperature (LOCA2), past and future, from the same climate model.
Written to every weather station of the basin model, one set per period.
Historical and future runs of one model, identical except for the climate.
Future minus historical, per climate model, then the median and range across models.
Each climate model supplies its own simulated past (1985–2014) and its projected future under SSP2-4.5 (2071–2100), downscaled by LOCA2. The same SWAT+ model runs on both, identical except for the climate, so a climate model's own bias largely cancels in the difference. The projection supplies precipitation and daily maximum and minimum air temperature; solar radiation, humidity and wind come from the SWAT+ weather generator in both runs.
The expected direction of every change was written down before any run existed: warmer air, more evapotranspiration, less snow, and streamflow arriving earlier in the year. Every one held in the 21-model median. Streamflow volume was given no expected sign.
The model behind it: how SWATGenX builds a model, the methodology and how we check models.
What this is, and what it is not
- One basin. It says nothing about how any other river will respond.
- One SWAT+ model, calibrated to daily streamflow at USGS gauge 04118500 with one parameter set; parameter uncertainty is not included, so it is not a basis for a management decision.
- One emissions scenario (SSP2-4.5) and one pair of periods (1985–2014 against 2071–2100).
- A comparison between two model runs, not a forecast for a particular year.
Questions about this comparison
Which climate models and scenario are used?
21 downscaled CMIP6 climate models from LOCA2 (v20240915) under SSP2-4.5, comparing 2071–2100 with 1985–2014. All 21 requested climate models completed.
Why compare each model's future with its own simulated past rather than with observations?
Because a climate model that runs too wet or too cold does so in both periods, so that bias largely cancels in the difference.
Why 21 climate models instead of one?
They agree on the direction of most changes but not on their size. Outlet streamflow rises in 19 of the 21, and the change ranges from −10.7 % to +32.6 % across them. One model would show one of those values and hide the rest.
What does 'the center of the year's streamflow' mean?
The flow-weighted mean day of the water year (from 1 October) at the outlet, averaged over the complete water years (Stewart, Cayan and Dettinger, 2005). It moves earlier when more of the year's flow comes in winter and early spring.
Is this a forecast for the Rogue River?
No. It compares two runs of one model calibrated to daily streamflow at USGS gauge 04118500 under one emissions scenario. It demonstrates the method; it is not a management forecast, and it says nothing about other rivers.
Can this run on my own SWAT+ model?
Not yet as a self-service option. We run it on your SWAT+ model as a quoted service; write to us.
Climate scenarios for your own model
Not yet as a self-service option. We run it on your SWAT+ model as a quoted service; write to us.
References
- Pierce et al. (2023). Future Increases in North American Extreme Precipitation in CMIP6 Downscaled with LOCA. Journal of Hydrometeorology 24, 951-975. doi.org/10.1175/jhm-d-22-0194.1
- Eyring et al. (2016). Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization. Geoscientific Model Development 9, 1937-1958. doi.org/10.5194/gmd-9-1937-2016
- O'Neill et al. (2016). The Scenario Model Intercomparison Project (ScenarioMIP) for CMIP6. Geoscientific Model Development 9, 3461-3482. doi.org/10.5194/gmd-9-3461-2016
- Bieger et al. (2017). Introduction to SWAT+, A Completely Restructured Version of the Soil and Water Assessment Tool. JAWRA Journal of the American Water Resources Association 53, 115-130. doi.org/10.1111/1752-1688.12482
- Stewart, Cayan, Dettinger (2005). Changes toward Earlier Streamflow Timing across Western North America. Journal of Climate 18, 1136-1155. doi.org/10.1175/jcli3321.1