Models for estimation of grassland production and spatial inversion based on MODIS data in Songnen Plain

Using NASA/MODIS data to calculate NDVI, RVI, MSAVI and EVI four vegetation indexes, combining grass production(green weight)data, by means of statistical methods, estimation models of grass production in Songnen Plain were established, for the sake of estimating yield of grass reasonably and regulating grassland resources. With the correlation analysis, it was found that all of the four prophase and synchronous vegetation indexes were correlated with grass production(green weight)significantly, among them NDVI had the highest correlation coefficient and EVI the lowest. The best estimation model of grass production in Songnen Plain was S curve model based on NDVI, with estimation accuracy 78%. Calculating yield of grass in Songnen Plain through the best estimation model, total fresh yield was 18 850 thousand tons, equivalent to 5 890 thousand tons of hay, and fresh weight per unit area was 5 717 kg/hm2, equaling to 1 687 kg/hm2 of dry weight. Total fresh yield belong to Heilongjiang Province was 13 560 thousand tons, equivalent to 4 240 thousand tons of hay, and yield of fresh grass and hay belong to Jilin Province was 5 310 and 1 660 thousand tons separately. Using vegetation indexes to predicting future sixteen days’ grass yield was proved nice, and the prediction accuracy of exponential model based on NDVI was 74%. The research shows that it is feasible to study grass production of Songnen Plain with vegetation indexes.