Evaluating kilometre-scale statistical downscaling of daily climate extremes in Ecuador’s complex terrain: performance, limitations and implications for impact studies

Journal article

Abstract Climate change is expected to increase the frequency and severity of extreme weather events, creating demand for high-resolution climate information to support impact modelling and adaptation planning. However, the performance of daily, kilometre-scale statistical downscaling frameworks remains poorly constrained, particularly in topographically complex, data-sparse regions. Here, we apply the globally standardised ISIMIP3 bias adjustment and statistical downscaling (ISIMIP3BASD) framework to daily temperature and precipitation from 15 CMIP6 models to generate 1-km resolution climate datasets over Ecuador, a region characterised by complex terrain and strong ENSO-driven variability. We evaluate performance for 2005-2014 against the CHELSA-W5E5 reference dataset and station observations, assessing reproduction of spatial climatology, temporal variability, and a suite of ETCCDI extreme climate indices. Because CHELSA-W5E5 contributes to the downscaling workflow, agreement with it is interpreted as reference-based consistency rather than independent validation. The downscaled data reproduce spatial gradients and seasonal cycles well, with the strongest skill at monthly to annual timescales. In contrast, station-based evaluation indicates weak anomaly skill, particularly for precipitation, showing limited ability to reproduce local variability and event sequencing. Performance is highest for extremes defined by absolute thresholds, whereas percentile-based extremes are less reliably captured. Mid-century projections under SSP2-4.5 indicate intensifying temperature extremes and spatially heterogeneous precipitation changes across Ecuador, with downscaled projections reflecting both large-scale GCM signals and local geographic controls. Overall, for Ecuador, ISIMIP3BASD can provide spatially detailed, daily climate realisations suitable for applications focused on climatological means, seasonal cycles, long-term climate change signals, and fixed-threshold exceedance frequencies. However, the data should not be treated as reconstructions of observed weather sequences and are less appropriate for applications requiring realistic persistence, antecedent conditions, or rare-event tail behaviour. By identifying these capabilities and limitations, this study supports more appropriate use of high-resolution downscaled climate data in observationally sparse regions.

Authors: Dobson, R; Howard, C; Willis, SG

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: IOP Publishing
• Journal: Environmental Research Letters
• Publication date: 2026-07-20
• DOI: 10.1088/1748-9326/ae8d3f
• EISSN: 1748-9326
• ISSN: 1748-9326
• Language: English
• Keywords: Spatial ecology; Terrain; Climate change; Anomaly (physics); GCM transcription factors; Climate model; Precipitation; Downscaling; Climate extremes