Crop modelling

Food security is one of the biggest challenges in the face of climate change and a fast-growing population. Within these challenges, the resilience of food crops to remain productive throughout changes to local weather regimes is a key focus. All essential aspects to a crop’s growth; water, radiation and soil, are all likely to be affected which could jeopardize the food security for regional and global communities.  Assimila is working with partners in the UK and China, using EO data and crop models to improve understanding of crop performance.

As the climate changes, past history of a crop can be an unreliable basis for future predictions.  Numerical crop models can be used to simulate crop growth based on meteorological data with crop specific parameters that describe how the target crop reacts to meteorological and hydrological pressures.

At Assimila we are using the WOFOST (WOrld FOod STudies) crop model which predicts certain crop characteristics on a day-by-day basis after a planting date. This is done by calculating the net energy available to the crop and partitioning it into different parts of the plant, based on the stage of the crop on that day and crop phenotype.

Winter wheat a few months after planting in Hebei Province, China
Leaf Area Index (LAI), retrieved from Sentinel 2 data compared to the possible outputs from the WOFOST model.

One of the critical issues facing crop modellers is to determine the right local calibration parameters to use.  We implement an innovative approach, generating an ensemble of multiple model runs and comparing them to measurements of biophysical parameters derived from satellites.  The satellite measurements inform us of site-specific field conditions as well as import phenological timing events, which helps to make the yield predictions for WOFOST allot more accurate on local scales and in individual fields.

For some applications, comparative or qualitative information of cope performance is required  to contextualize the current year’s crop development compared to previous years. This is made possible by interfacing with the Assimila DataCube which provides historical weather data to drive WOFOST. An example of this is shown below, where ensemble runs of WOFOST are compared to past growing season ensemble members. The progression of crop yield throughout the season is displayed on the right panel, whereas the last timestep for all growing seasons are compared in the boxplot on the left hand panel.

Comparisons of growing seasons using WOFOST Ensembles