Modeling relationships between a satellite-derived vegetation index and precipitation in the northern Great Plains

The Normalized Difference Vegetation Index (NDVI), derived from the Advanced Very High Resolution Radiometer data, plays a key role in detecting terrestrial vegetation greenness of large geographic areas. The quantitative relationship between the NDVI and precipitation is an important component in the understanding of climate and vegetation interactions. Although a large number of investigations have been conducted on this relationship, it has not been sufficiently studied due to the complexity of vegetation dynamics, including lag and seasonality effects on vegetation response to moisture availability.; To achieve a better understanding of the NDVI and precipitation relationship, a statistical time series analysis was performed on biweekly NDVI and precipitation data at the weather station level in the northern and central U.S. Great Plains. Regression models that incorporate lag and seasonality effects were used to quantify NDVI and precipitation interaction in grassland and cropland. It was found that the NDVI lag response to precipitation varied with plant growth stage, with the longest response duration occurring in the middle to late growing season. The regression analysis points to a close relationship between NDVI and lagged precipitation, with R2 values of 0.65 and 0.73 for grassland and cropland, respectively.; This result was applied to the study of vegetation response to drought. The 3-month Standardized Precipitation Index was found to have the best correlation with NDVI, but the highest correlations occurred during the middle of the growing season. This timing was associated with the “moisture-sensitive period” of plant growth. This study suggests seasonal timing should be taken into consideration when monitoring drought with NDVI.; The result was also applied to the creation of a Vegetation Greenness Forecast Model. The model implements an autoregressive distributed-lag function with a seasonal adjustment and uses historical climate and NDVI data as inputs. This model successfully predicts vegetation condition up to three months with high accuracy. A potential application of the model would be to produce maps of real-time forecasted greenness.