@Article{Yan_ERC_20190430,
 author		= {Binyan Yan and Jiafu Mao and Xiaoying Shi and Forrest M. Hoffman and Michael Notaro and Tianjun Zhou and Nate McDowell and Robert E. Dickinson and Min Xu and Lianhong Gu and Daniel M. Ricciuto},
 title		= {Predictability of Tropical Vegetation Greenness Using Sea Surface Temperatures},
 journal	= ERC,
 volume		= 1,
 number		= 3,
 pages		= {031003},
 doi		= {10.1088/2515-7620/ab178a},
 day		= 30,
 month		= apr,
 year		= {2019},
 abstract	= {Much research has examined the sensitivity of tropical terrestrial ecosystems to various environmental drivers. The predictability of tropical vegetation greenness based on sea surface temperatures (SSTs), however, has not been well explored. This study employed fine spatial resolution remotely-sensed Enhanced Vegetation Index (EVI) and SST indices from tropical ocean basins to investigate the predictability of tropical vegetation greenness in response to SSTs and established empirical models with optimal parameters for hindcast predictions. Three evaluation metrics were used to assess the model performance, i.e., correlations between historical observed and predicted values, percentage of correctly predicted signs of EVI anomalies, and percentage of correct signs for extreme EVI anomalies. Our findings reveal that the pan-tropical EVI was tightly connected to the SSTs over tropical ocean basins. The strongest impacts of SSTs on EVI were identified mainly over the arid or semi-arid tropical regions. The spatially-averaged correlation between historical observed and predicted EVI time series was 0.30 with its maximum value reaching up to 0.84. Vegetated areas across South America (25.76\%), Africa (33.13\%), and Southeast Asia (39.94\%) were diagnosed to be associated with significant SST-EVI correlations ($p < 0.01$). In general, statistical models correctly predicted the sign of EVI anomalies, with their predictability increasing from $\sim$60\% to nearly 100\% when EVI was abnormal (anomalies exceeding one standard deviation). These results provide a basis for the prediction of changes in greenness of tropical terrestrial ecosystems at seasonal to intra-seasonal scales. Moreover, the statistics-based observational relationships have the potential to facilitate the benchmarking of Earth System Models regarding their ability to capture the responses of tropical vegetation growth to long-term signals of oceanic forcings.}
}