An XGBoost-based inversion study on chlorophyll-a concentration in oceans observed by satellite-mounted GNSS-R

Chlorophyll-a (Chl-a) is a crucial pigment involved in photosynthesis in plants and serves as a key indicator of phytoplankton biomass. As a critical indicator for water quality evaluation, continuous monitoring of Chl-a concentration plays a significant role in assessing aquatic ecosystems' health. GNSS reflectometry has the advantage of a short re-entry period and is not affected by weather. This research utilizes the eXtreme Gradient Boosting (XGBoost) model to capture a nonlinear correlation, incorporating inputs from the Cyclone Global Navigation Satellite System (CyGNSS) alongside additional auxiliary data to forecast Chl-a concentration. The correlation coefficient between the model output and the Chl-a concentration of NOAA was 0.9328 with an RMSE of 0.1372 mg/m³ and an MAE of 0.0199 mg/m³ using XGBoost to model the Chl-a concentration. Additionally, it produces a RMSE of 0.1441 mg/m³ and a MAE of 0.0201 mg/m³.The model's output Chl-a concentration was further evaluated against Moderate Resolution Imaging Spectroradiometer-Aqua Chl-a concentration, yielding a correlation coefficient of 0.6754, a RMSE of 0.3741 mg/m³, and a MAE of 0.0421 mg/m³.