POSTER SESSION 2
10:30–12:30, Tuesday, September 15
Poster Session | 1 | 2 | 3 | 4 | Instructions | Floorplan | Schedule at a Glance
ABSTRACT 1223 | POSTER 234
ESTIMATING TOTAL IRON IN DUTCH INLAND WATERS FROM SENTINEL-2
Iron dynamics in Dutch lowland waters reflect hydrological inputs and redox transformations and influence phosphorus retention. Fe-bearing material can influence light absorption and scattering, contributing to variation in water-leaving reflectance. Water-quality monitoring by Rijkswaterstaat and regional water authorities provides an extensive record of total-Fe measurements, enabling linkage to satellite observations. Published satellite applications have focused mainly on iron-rich acid mine drainage. To our knowledge, this is the first national-scale evaluation of Sentinel-2 for predicting total Fe across heterogeneous lowland freshwaters outside mining settings.
We linked 775 field measurements from 151 locations to satellite observations and compared models using spectral predictors, environmental predictors or both. After accounting for suspended solids, organic carbon and chlorophyll-a, changes in total Fe within the same monitoring location were most clearly associated with reflectance in Sentinel-2’s red band (B4). The best-performing configuration was an XGBoost regression model combining five reflectance-derived predictors with water-body width, seven-day antecedent precipitation, calendar month and relative elevation.
Nested location-grouped evaluation, withholding complete monitoring locations from training, yielded R² = 0.68 and RMSE = 0.66 mg L⁻¹. Forward-year evaluation yielded R² = 0.68, and spatial-block evaluation, withholding complete 25 × 25 km areas, yielded R² = 0.57. Concurrent measurements of Fe fractions, redox speciation and optical properties could help explain residual prediction error and support more chemically specific retrievals. Sentinel-2 reflectance, combined with environmental context, supported empirical estimation of total Fe within the evaluated Dutch freshwater domain and added spatial and temporal context to field monitoring
Marit van Oostende, Vrije Universiteit Amsterdam, [email protected], https://orcid.org/0000-0003-3364-3785
Ype van der Velde, Vrije Universiteit, [email protected], https://orcid.org/0000-0002-2183-2573
Poster Session | 1 | 2 | 3 | 4 | Instructions | Floorplan | Schedule at a Glance
