What question did this study set out to answer?

The research aims to compare the effectiveness of statistical and dynamical downscaling methods for predicting significant wave height in the Cuban archipelago.

February 14, 2026

Comparison between statistical and dynamical downscaling for estimating significant wave height (Hsig) in the Cuban Archipelago

Key Points

The research aims to compare the effectiveness of statistical and dynamical downscaling methods for predicting significant wave height in the Cuban archipelago.
Compared historical significant wave height data with future projections from 2081-2099.
Utilized ERA5 reanalysis and CMIP5 models under RCP4.5 and RCP8.5 scenarios.
Employed statistical methods, including quantile regression and bias correction techniques.
Statistical downscaling showed high accuracy in mean conditions, with a skill score of approximately 23%.
Dynamic downscaling outperformed statistical methods for extreme wave height events, showing lower error rates at high percentiles.

Abstract

This study evaluates significant wave height (Hsig) along the Cuban archipelago by comparing the future period 2081–2099 with the historical baseline 1986–2004. Using ERA5 reanalysis and a weighted multi-model mean (MMM) derived from six CMIP5 GCMs under RCP4.5 and RCP8.5 scenarios, downscaling was performed via dynamic (DD) and statistical (SD) approaches, the latter based on quantile regression between Hsig and wind speed. The SD coefficients showed temporal invariance—a prerequisite for projection, validated via Wald tests with robust variance estimation through Moving Block Bootstrap (MBB)—in over 93% of grid points at extreme percentiles (90th, 95th) and 79% at the 50th percentile. Bias correction with Quantile Delta Mapping (QDM) reduced the global quantile deviation (qqdev) from 0.213 m to 0.164 m (skill score ≈ 23%). SD exhibited highest accuracy for mean regimes (45th–55th percentiles), with errors below instrumental uncertainty, and outperformed DD in temporal correlation ( r ≈ 0.77 vs. 0.45) and distribution center precision after correction. Conversely, DD yielded 20–53% lower errors at the 90th and 95th percentiles, confirming its superior representation of extreme-event physics. These results demonstrate that DD better captures extremes, while SD—due to its computational efficiency and accuracy for mean conditions—offers a pragmatic alternative for coastal risk assessments in resource-limited Caribbean environments.

Demander à l'IA

Bookmark