Abstract. Soil moisture (SM) is a critical variable governing land–atmosphere interactions and influencing ecohydrological and climatic processes. Despite substantial progress in estimating SM through remote sensing and land surface models, considerable uncertainties still remain, especially in near-natural and poorly monitored ecosystems interacting with deeper soil layers. In this study, the performance of four state-of-the-art gridded SM products (SMAP-L4, GLDAS-Noah, ERA5 and ERA5-Land) is evaluated against in situ observations at ten natural monitoring sites in central and southern Chile, covering different hydroclimatic conditions (five semi-arid and five humid sites). The evaluation is performed at a 3-hourly temporal resolution, using well-known statistical metrics of performance, including unbiased root mean square error, modified Kling–Gupta efficiency (KGE′), deseasonalised Spearman's rank correlation coefficient, and percent bias, each applied separately for surface soil moisture (SSM) and root zone soil moisture (RZSM). Finally, the dynamic SM responses to precipitation events is evaluated using rising time and amplitude SM signatures during the first and the most intense precipitation events of the year. Our results show that ERA5 and ERA5-Land consistently outperform SMAP-L4 and GLDAS-Noah on most metrics and in most regions, with ERA5-Land being particularly strong in humid areas. However, SMAP-L4 achieved the best SSM performance in selected northern arid locations, based on KGE′; while GLDAS-Noah performed the worst overall, with the exception of moderate correlation values in southern RZSM. During the first precipitation event of the year, all products systematically overestimated both rising times and amplitudes in the arid north, indicating challenges in capturing SM responses under dry antecedent conditions. In contrast, all the gridded products aligned more closely with in situ measurements during intense precipitation events, particularly in humid regions. Our findings suggest that both ERA5 and ERA5-Land are valuable datasets for monitoring SM variability in near-natural and data-scarce ecosystems, while highlighting the value of event-based SM signatures to complement traditional performance metrics. Finally, we recommend the use of the deseasonalised Spearman rank correlation to better detect inconsistencies in temporal dynamics, especially in regions with strong seasonal cycles, such as arid environments.
Nuñez-Ibarra et al. (Thu,) studied this question.