Weather, both directly and indirectly, is the critical factor in the success of a harvest and farmers' livelihoods. Severe weather events, such as hail, high winds, tornados, and flash floods can destroy an entire harvest in a very short period. However, many agricultural decisions require more accurate forecasts of the weather and the resultant soil conditions. Precise soil temperature and soil moisture forecasts are critical to the timely application of pesticides, seed and fertilizer selection, and to efficient irrigation practices.
RAL has been collaborating with industry to develop agricultural decision support capabilities that optimizes the timing of pesticide application and irrigation. These projects typically utilize advanced weather and land surface models and an intelligent data fusion technology that continuously optimizes the weather and soil predictions. This research has led to improvements in the High-Resolution Land Data Assimilation System (HRLDAS), Dynamic, Integrated Forecast System (DICAST®), and Noah Land Surface Model. This research is instrumental in providing critical feedback to the weather and land surface modeling, and satellite communities and represents a cross disciplinary effort. Continued work in this area will lead to more precise prediction of weather and soil condition and more efficient and profitable agricultural operations.
The NASA Agriculture Decision Support Project combines advanced weather and soil forecast systems (DICast and HRLDAS) with the goal of producing an accurate soil temperature and moisture forecast at high temporal and spatial resolution. The output of the land–surface model will be communicated to Meteorlogix (RAL's commercial partner in this project). After an evaluation period, the output will be used to drive agriculture–specific models and incorporated into the Meteorlogix agricultural DSS which has roughly 80,000 subscribers. NASA's interest in funding the project is to determine whether incorporation of their MODIS products into the land–surface model can improve the soil forecasts.
The vegetation state and land use are key factors in the land–surface model. Prior to this project, HRLDAS has used static land use and climatological vegetation data sets developed in the 1970s by the USGS. This project evaluated the use of MODIS satellite data sets to improve the initial conditions provided to HRLDAS. The MODIS land use data sets are static, but are of higher spatial resolution and are much newer than the USGS data sets. Rather than monthly climatological averages, the MODIS Leaf Area Index (LAI) and other products are updated weekly and better represent the current vegetation state. As anticipated, the use of these products improved the soil temperature and moisture forecasts, but great care must be taken when utilizing these data types for this purpose.
Model runs covering 2005–2006 were compared to soil temperature and moisture observations. The soil temperature errors at 5 cm were reduced by roughly 10% and the 10 cm soil temperature errors were reduced by approximately 50%. These are critical depths for agriculture.
MODIS data, as well as enhancements made by the HRLDAS developers during this project, were incorporated into the prototype soil temperature prediction system. Continued land–surface model development will attempt to further improve the modeled heat transfer at the surface. The results from retrospective studies and the 2009 growing season using these model improvements were incorporated into the end–of–project report to NASA.