Can models describe the translation of rainfall and temperature to not only discharge but soil moisture and ET?
The hydrological cycle across the landscape can be simply defined as precipitation (P) onto the landscape is equal to (partitioned by) to runoff (R), evapotranspiration (ET) and changes in soil moisture (SM) over time (P = R + ET + SM). Runoff not only provides the direct water flow for multiple issues (downstream transport, filling reservoirs, causing floods), but it a parameter that can be directly measured. If it can be demonstrated that a model can reproduce R (at gauges), then there is more confidence in calculations of ET and SM. As such ET represents a loss to the landscape, exacerbated in dry regions where some plants have produce high ET. Soil moisture is important for both agriculture (relative to droughts) and flooding (a region with high SM is more prone to flooding). Therefore the first order of business for the modeling is to be able to demonstrate that it is capable of addressing the question,
Could the models calculate other results or products, for example soil moisture and evapotranspiration, as well as fluvial discharge?
Hence the first step here is to examine the ability of the models to describe R, ET and SM across Espírito Santo. That is, how well can the hydrological cycle of Espírito Santo be modeled (by VIC)? An important aspect of this process is to be able to “visualize” how the hydrological processes are distributed, and, eventually to be able to be queried for computed results at specific points.
The setup and calibration and validation of VIC (and DHSVM) is described under the tab on Data to Models. Given the sparse observed records and uncertainty in soil and vegetation data, VIC does a reasonable job at simulating streamflow. VIC performs better for the larger basins since it uses parameterization of spatial variability in soil properties, and topographic effects is based on a simplifying assumption that the large-scale effects can be represented adequately without assigning infiltration parameters to specific subgrid locations. Uncertainties in climate forcing are more averaged out.
In the interactive graph in the figure below, each station can be queried, with each year of record compared to others.
|VIC results: observed (obs) vs simulated model (t0/p100)|
With the ability of the model to translate rainfall to runoff established, the next problem is to examine the resulting patterns of ET and changes in soil moisture also computed by the model. In this figure of annual averages, P and T lead to ET and SM. Each point on the map can be queried With the ability to “see” such spatial maps, it would be straight-forward to track evolving conditions.
Annual average P (from Climate Re-analysis tab) driving VIC to produce ET and SM (as well as R). All scales increasing left to right. Showing seasonality, wet conditions of January with high ET and SM, drying out through September, then getting wet again.
Animating model results can produce a more dynamic view of the co-variance of P, R, ET, and SM, under different scenarios of climate forcing variables. This gives information on the possible range of responses of the landscape to climate, providing a sensitivity analysis of fluxes the interactions of P, T, R, and SM. By changing the model forcing elements of P and T, the effects on a watershed can be seen over time. This is shown in the four inserts, below. Each is a snapshot of the watershed on March 1, 1980 showing average precipitation for the month and modeled runoff for the day. Precipitation was modeled to be light in March and yet runoff is significant as January and February were modeled to be wet.
The four model runs show differences. The upper-right insert shows the additional model runoff on March 1, 1980 when the model runs precipitation levels at 110% of base. The lower-left insert shows the base model run. The upper-right shows 100% precipitation but a temperature increase of three degrees Celsius. The lower-right shows the decrease in runoff on March 1, 1980 when the model runs precipitation levels at only 70% of base. By comparing the four inserts, one gets sense of the sensitivity of runoff, ET, soil moisture to preciptiation and temperature.
Click on any one of the inserts to see an animated video of modeled average precipitation and resultant modeled runoff during the years 1970-2007 for the four scenarios depicted.
Results show that the modeling environment is capable of addressing not only the distributions of runoff, but of ET and soil moisture.