Lake Urmia's shallow depths were more accurately mapped by a tree model
WHAT THE EVIDENCE SAYS
Using field depth observations and Landsat-8 predictors, authors compared four regressions. CatBoost attained held-out R² about 0.93 and RMSE near 9 cm under repeated evaluation, outperforming linear regression under the sampled lake conditions.
WHY IT MATTERS / OUR READ
Satellite tools can improve water-depth mapping in challenging shallow lakes when calibrated to local observations.
CAVEATS / UNCERTAINTY
Accuracy pertains to a specific lake scene and holdout design; transfer across salinity, season, sediment and depth requires independent validation.
SOURCE
Mojtahedi et al., Scientific Reports (9 Nov 2026)
View original evidence ↗
A satellite and field-comparison study shows promising bathymetry in one difficult hypersaline setting.
Read the Wild Lands assessment →