Spatial association of anomaly correlation for GCM seasonal forecasts of global precipitation

Global climate models (GCMs) are used by major climate centers worldwide for global climate forecasting, and predictive performance is one of the most important issues in GCM forecast applications. In addition to spatial plotting that illustrates anomaly correlation at individual grid cells, this study proposes a novel local indicator of spatial association (LISA) of anomaly correlation (herein, LISAAC) for GCM seasonal forecasts of global precipitation. LISAAC is built upon local Moran's I by relating anomaly correlation at neighboring grid cells to one another. While local Moran's I takes the grand mean of anomaly correlation as the benchmark, LISSAC considers the original value of anomaly correlation in the mathematical formulation. A case study is devised for the Climate Forecast System version 2 (CFSv2) seasonal forecasts, which are initialized in January, February, horizontal ellipsis , and June, of the global precipitation in June, July, and August. Three metrics-LISAAC, local Moran's I, and original anomaly correlation-are applied to investigate the predictive performance. In comparison with local Moran's I, LISAAC can identify clusters of positive, neutral, and negative anomaly correlations. In comparison with anomaly correlation, LISAAC can capture outliers of positive (negative) anomaly correlation surrounded by negative (positive) anomaly correlation. Overall, the results highlight that LISAAC can serve as a useful tool for evaluating the predictive performance of GCM seasonal forecasts of global precipitation.