Comparison of the driving forces of spring phenology among savanna landscapes by including combined spatial and temporal heterogeneity.

Understanding spatial and temporal dynamics of land surface phenology (LSP) and its driving forces are critical for providing information relevant to short- and long-term decision making, particularly as it relates to climate response planning. With the third generation Global Inventory Monitoring and Modeling System (GIMMS3g) Normalized Difference Vegetation Index (NDVI) data and environmental data from multiple sources, we investigated the spatio-temporal changes in the start of the growing season (SOS) in southern African savannas from 1982 through 2010 and determined its linkage to environmental factors using spatial panel data models. Overall, the SOS occurs earlier in the north compared to the south. This relates in part to the differences in ecosystems, with northern areas representing high rainfall and dense tree cover (mainly tree savannas), whereas the south has lower rainfall and sparse tree cover (mainly bush and grass savannas). From 1982 to 2010, an advanced trend was observed predominantly in the tree savanna areas of the north, whereas a delayed trend was chiefly found in the floodplain of the north and bush/grass savannas of the south. Different environmental drivers were detected within tree- and grass-dominated savannas, with a critical division being represented by the 800 mm isohyet. Our results supported the importance of water as a driver in this water-limited system, specifically preseason soil moisture, in determining the SOS in these water-limited, grass-dominated savannas. In addition, the research pointed to other, often overlooked, effects of preseason maximum and minimum temperatures on the SOS across the entire region. Higher preseason maximum temperatures led to an advance of the SOS, whereas the opposite effects of preseason minimum temperature were observed. With the rapid increase in global change research, this work will prove helpful for managing savanna landscapes and key to predicting how projected climate changes will affect regional vegetation phenology and productivity.