Machine-learning-derived surface nitrate dataset for the Northwest European Shelf

This dataset provides daily gridded surface nitrate (NO₃) concentrations [mmol N m⁻³] for the Northwest European Shelf (NWES) covering the period 1998–2018. The dataset is generated using a machine learning (neural network) model trained on in situ nitrate observations and environmental driver variables. The methodology follows the approach described in: Improved understanding of nitrate trends, eutrophication indicators, and risk areas using machine learningwhere a neural network model was developed to reconstruct a gap-free, bi-decadal nitrate dataset from sparse observations across the NWES. The model uses a combination of satellite-constrained variables, atmospheric reanalysis, riverine inputs, and spatial–temporal features to predict surface nitrate concentrations. Due to observational sparsity, nitrate measurements in the NWES are limited in space and time. This dataset addresses that limitation by providing a spatially and temporally complete reconstruction, enabling analysis of nutrient dynamics, eutrophication patterns, and long-term variability. The effective resolution of the dataset is coarser than the native grid (~7 km, daily), but is well suited for monthly and regional-scale analyses. This nitrate product has been further applied in subsequent work to improve phytoplankton forecasting through data assimilation, demonstrating significant forecast skill improvements. (Assimilation of machine-learning-predicted nitrate to improve the quality of phytoplankton forecasting in the shelf-sea environment) Domain: Northwest European ShelfTemporal coverage: 1998–2018Resolution: ~7 km grid, daily (effective resolution coarser)Units: mmol N m⁻³ This dataset is intended for research applications including marine biogeochemistry, eutrophication studies, and model evaluation.