Satellite-derived land cover data valuable for improving model temperature forecasts
|
Values of leaf area index calculated over a two-week period ending July 17, 1997, from the Advanced Very High Resolution Radiometer (AVHRR) data available from the NOAA polar orbiting satellites. The brown area across western Oklahoma is the region where winter wheat is grown, which typically is harvested in late May and early June. By July, the winter wheat region is characterized by very low values of leaf area index (due to harvest) as indicated in the image. |
Most numerical weather prediction models represent land surface effects rather crudely by assuming that vegetation characteristics do not change from year-to-year. In reality, vegetation (or land cover) characteristics change substantially from one year to the next, and these changes are difficult to predict. Scientists from NSSL's Models and Assimilation Team, collaborating with remote sensing specialists from the University of Nebraska-Lincoln, have developed and implemented a technique to determine various land cover characteristics from National Oceanic and Atmospheric Administration (NOAA) satellite data. The goal of this work is to insert the observed land cover characteristics into a numerical weather prediction model with a state-of-the-art land surface scheme.
Values of fractional vegetation cover, leaf area index, and surface albedo calculated from satellite data were used to initialize the model for selected days during July 1997 (see figure). The land surface scheme was implemented using the satellite-derived vegetation data, and high temperature forecasts were shown to be significantly more accurate than those produced with climatological values of vegetation parameters. Surprisingly though, high temperature forecasts were better when the sophisticated land surface scheme was not used at all. This emphasizes the important notion that adding complexity to a numerical model will not necessarily improve it.
However, the forecasts using high-resolution land cover data were significantly
better than those without the land-surface scheme at the warmest one-third
of the sites. This suggests that use of the high-resolution land cover
data will be most beneficial in improving forecasts of extreme temperatures,
where current operational numerical models perform poorly. The positive
results from this study illustrate the enormous value of using satellite-derived
land cover data. Future work will involve more accurate initializations
of soil moisture as well as an investigation of how the use of the high-resolution
land cover data improves forecasts of convective initiation.
