Remote Sensing, Vol. 11, Pages 2123: Climate Prediction of Satellite-Based Spring Eurasian Vegetation Index (NDVI) using Coupled Singular Value Decomposition (SVD) Patterns (Remote Sensing)

Satellite-based normalized difference vegetation index (NDVI) data are widely used for estimating vegetation greenness. Seasonal climate predictions of spring (April–May–June) NDVI over Eurasia are explored by applying the year-to-year increment approach. The prediction models were developed based on the coupled modes of singular value decomposition (SVD) analyses between Eurasian NDVI and climate factors. One synchronous predictor, the spring surface air temperature from the NCEP’s Climate Forecast System (SAT-CFS), and three previous-season predictors (winter (December–January–February) sea-ice cover over the Barents Sea (SICBS), winter sea surface temperature over the equatorial Pacific (SSTP), and winter North Atlantic Oscillation (NAO) were chosen to develop four single-predictor schemes: the SAT-CFS scheme, SICBS scheme, SSTP scheme, and NAO scheme. Meanwhile, a statistical scheme that involves the three previous-season predictors (i.e., SICBS, SSTP, and NAO) and a hybrid scheme that includes all four predictors are also proposed. To evaluate the prediction skills of the schemes, one-year-out cross-validation and independent hindcast results are analyzed, revealing the hybrid scheme as having the best prediction skill. The results indicate that the temporal correlation coefficients at 92% of grid points over Eurasia are significant at the 5% significance level in the hybrid scheme, which is the best among all the schemes. Furthermore, spatial correlation coefficients (SCCs) of the six schemes are significant at the 1% significance level in most years during 1983–2015, with the averaged SCC of the hybrid scheme being the highest (0.60). The grid-averaged root-mean-square-error of the hybrid scheme is 0.04. By comparing the satellite-based NDVI value with the independent hindcast results during 2010–2015, it can be concluded that the hybrid scheme shows high prediction skill in terms of both the spatial pattern and the temporal variability of spring Eurasian NDVI.