Quantitative assessment of NMME seasonal forecast skill of temperature and precipitation over Southeast Asia

Seasonal climate forecasts are critical for Southeast Asia (SEA), a region highly exposed to climate variability and hydroclimate extremes. Here we quantify the seasonal forecast skill of the North American Multi-Model Ensemble (NMME) for monthly precipitation and near-surface temperature using hindcasts (1982–2010) and real-time forecasts (2011–2024). Using the Kling-Gupta Efficiency (KGE) metric and its components, we assess 17 dynamical models (and multi-model ensembles) across five SEA subregions and two representative target months (January and July) that reflect contrasting monsoon regimes. Temperature generally shows higher and more spatially coherent skill than precipitation, with the highest skill over mainland SEA and reduced skill over maritime regions; skill decreases with increasing lead time. Forecasts targeting January show lower precipitation skill and larger inter-model spread than those targeting July. Model performance depends strongly on variable, region, and lead-time; however, CFSv2, GEOSS2S, and CCSM4 frequently rank among the top-performing models and match or exceed the multi-model ensemble mean (MME). Across all metrics, CFSv2 achieves the highest mean KGE score of 0.71 (precipitation: 0.70; temperature: 0.72), while the MME mean KGE score is 0.53 (precipitation: 0.50; temperature: 0.55). Because temperature skill can be influenced by long-term trends, we also report a sensitivity using detrended anomalies. Overall, the findings provide a regional evidence base for advancing NMME-based climate services for agriculture, water management, and disaster preparedness across SEA.