Gauge-to-Grid Precipitation Error of CHIRPS v2 and CRU TS v4.10 at Mexican Stations, 1981–2020: Error Scale, Seasonality, Interannual Amplitude and Environmental Associations

Gauge-to-grid precipitation errors reflect environmental setting and the scale used to express disagreement. This study assessed CHIRPS v2 and CRU TS v4.10 against Mexican Servicio Meteorológico Nacional observations during 1981–2020. A 52-station network covered 25 combinations of latitude and elevation; 24 gauges added by coverage, before inspecting errors, formed a 76-station extension. Annual and quarterly metrics were examined with paired comparisons, adjusted models and sensitivity analyses. Adjusted models did not detect elevation or latitude associations in either network. CRU TS normalized mean absolute error (nMAE) showed an exploratory decrease with coastal distance, but error ratio models did not establish a product difference. July–September had lower relative error but higher absolute MAE than January–March at most stations, including matched years in the original network. Median annual standard deviation ratios of 0.61–0.65 indicated attenuated interannual amplitude. Median paired nMAE differences (CHIRPS minus CRU) were −3.12 percentage points in the original network (95% bootstrap interval from −4.72 to −1.36; Holm p = 0.006) and −3.06 in the extension (from −4.05 to −1.39; p < 0.001). Extraction from the nearest native cell preserved the original comparison without equalizing spatial support. Global residual Moran tests were not significant; local dependence remains unresolved. The results describe gauge-to-grid agreement in purposive networks, not independent predictive skill or nationwide causal controls.