Satellite mapping of Baltic Sea Secchi depth with multiple regression models

摘要: Abstract Secchi depth is a measure of water transparency. In the Baltic Sea region, maps are used to assess eutrophication and as input for habitat models. Due their spatial temporal coverage, satellite data would be most suitable source such maps. But Sea’s optical properties so different from open ocean that globally calibrated standard models suffer large errors. Regional predictive take special into account thus needed. This paper tests how accurately generalized linear (GLMs) additive (GAMs) with MODIS/Aqua auxiliary inputs can predict at regional scale. It uses cross-validation test prediction accuracy hundreds GAMs GLMs up 5 variables. A GAM 3 variables (chlorophyll a, remote sensing reflectance 678 nm, long-term mean salinity) made accurate predictions. Tested against field observations not model selection calibration, best model’s absolute error (MAE) daily predictions was 1.07 m (22%), more than 50% lower other publicly available The MAE predicting monthly averages 0.86 m (15%). Thus, proposed process able find good accuracy. could useful environmental depth, using sensors, regions where non-standard needed mapping. Annual 2003–2012 come this Supplementary materials.