Much of the local precipitation comes as snow in the high mountains and the late spring snowmelt is a key driver of regional hydrology. We have previously shown that regional patterns of winter-time climate variability, as captured in the NCEP/NCAR reanalysis, provide sufficient information about snow accumulation to skillfully forecast warm season river flows. This technique is particularly useful since accurate estimates of the snow pack are not available. As the bulk of the precipitation falls as snow during the cold season rather than as rain during the warm season, the snowmelt water is also a key driver of vegetation, both natural and agricultural. Thus, the same approach for forecasting warm-season river flows can also be used to forecast warm season vegetation.
The Normalized Difference Vegetation Index (NDVI) is used as an estimate of vegetation. Multivariate Empirical Orthogonal Function (EOF) analysis is applied to three variables: Nov-Mar 200hPa zonal wind, Nov-Mar model precipitation, and the subsequent Apr-Aug NDVI. As the time period for the climate variables (wind and precipitation) does not overlap the time period for the NDVI, this scheme can be directly implemented for forecasting. For parity with the climate variables the NDVI is regridded to 2.5x2.5 for the EOF calculations. Correlations to the original 8km NDVI data do exceed 0.8 in some areas, including key agricultural growing areas, demonstrating that there is predictability of vegetation even at 8km resolution. Aspects of recent droughts, including the 1999-2001 drought, are captured. Correlating the EOF results to sea surface temperatures (SSTs) shows a strong link to the tropical Pacific, consistent with our previous work on the 1999-2001 drought and suggestive of longer-lead predictability based on an SST predictor.