Is there any skill in daily global precipitation forecasts over the Maritime Continent?

Abstract The presence of and spatial variation in daily precipitation forecast skill from a global deterministic Numerical Weather Prediction (NWP) model is explored over the Maritime Continent (MC). Location specificity and the variation of forecast performance with topographic height were of primary interest. The localised Fractions Skill Score (LFSS) was used alongside one well‐known derivative, the Skilful Spatial Scale (SSS). In this study a new temporal factorisation, the Lead Time Potential (LTP), is introduced and applied. It was found that whilst scores can be high, these do not necessarily imply useful skill. This is because of the impact of observed frequency of occurrence or base rate f in the 0.5 + f /2 useful skill criterion which underpins the SSS and LTP. For regions with high rainfall, such as the MC, it is not uncommon for f → 1, especially for modest thresholds, such that 0.5 + f /2 → 1. This means that even a near‐perfect score is not associated with any useful skill. Thus, if f /2 is ignored, the scales with useful skill are either considered as being smaller than they are, or worse, suggesting there is useful skill when there is none. The new LTP identifies many locations where useful skill is present up to seven days ahead. Typhoons are associated with some of the highest scores and skill, whereas persistence over land can be locally as skilful as the model forecast, though model forecasts show larger regions with better (lower) SSS and better (higher) LTP.