Improving estimates of soil moisture over Ghana

By Ewan Pinnington

This work aims to improve estimates of soil moisture over Ghana as part of the ERADACS project. In regions where the population relies on subsistence farming it is soil moisture, rather than precipitation per se, that is the critical factor in growing crops. The production of improved soil moisture forecasts should therefore enhance the drought resilience of these regions through improved capacity for early warning of agricultural drought. The seasonal cycles of precipitation and soil moisture over Ghana are shown in a video here, in this video we can see how the response of soil moisture to increased rainfall is lagged as it takes time for the rain to infiltrate into the soil.

Mathematical models of the land-surface are useful tools to inform soil moisture forecasts, but model errors are problematic. In order to improve soil moisture estimates from the Joint UK Land Environment Simulator (JULES) land surface model over Ghana we have combined satellite observations of precipitation and soil moisture with model predictions using the technique of data assimilation.

We have built a four-dimensional variational data assimilation system that ingests soil moisture observations from the European Space Agency (ESA) Climate Change Initiative to update the soil model parameters of JULES. In our experiments we drive the JULES model with precipitation observations from TAMSAT. Figure 1 shows a data assimilation experiment for one grid box. In this experiment we assimilated one year of soil moisture observations (2009) and then ran a five year hindcast (2010-2014) to judge the model performance against independent data. We can see that the JULES model being run with updated parameters after data assimilation (dark grey line) fits the ESA observations of soil moisture better than our prior model run. We can also see the reduction in bias for the hindcast period over the whole of Ghana in Figure 2, we see that before data assimilation the model is too wet over much of the country and that this bias is reduced after data assimilation.

Figure 1. Soil moisture data assimilation results for a north Ghana grid. Light grey line: prior JULES trajectory. Dark grey line: posterior JULES trajectory. Black dots: ESA CCI soil moisture observations. Faint grey vertical lines: error bars for observations. Vertical dashed line represents the end of the assimilation window.

Figure 2. JULES modelled soil moisture bias over Ghana for period 2010-2014. Left: before data assimilation. Right: after data assimilation.

Overall we find a 44% reduction in root-mean-squared error for our 5-year hindcast after assimilating a single year of soil moisture observations to update model parameters. The initial results of using this system are encouraging, but more work is needed to judge our results against “ground-truth” observations of soil moisture. From this work we also conclude that rainfall data has the greatest impact on model estimates during the seasonal wetting-up of soil, with the assimilation of remotely sensed soil moisture having greatest impact during drying down. For more information on this work please see our Hydrological and Earth System Sciences Discussions paper (Pinnington et al., 2018).

Reference
Pinnington, E., Quaife, T., and Black, E., 2017 (in review). Using satellite observations of precipitation and soil moisture to constrain the water budget of a land surface model. Hydrol. Earth Syst. Sci. Discuss., https://doi.org/10.5194/hess-2017-705

This entry was posted in Africa, Climate, Climate modelling, data assimilation, Hydrology, land use, Numerical modelling and tagged . Bookmark the permalink.