A multivariate linear regression model for estimating chlorophyll-a concentration in Quan Son Reservoir (Hanoi, Vietnam) using Sentinel-2B Imagery

Abstract

Monitoring chlorophyll-a concentration (Chla) in inland waters is vital for environmental assessment. This study develops an empirical multivariate linear regression (MLR) model to directly estimate Chla in Quan Son Reservoir using Sentinel-2B (S2B) Level 2A images. Regression analysis of a 68-point in-situ Chla dataset measured in Quan Son Reservoir between 2021 and 2023, in conjunction with the corresponding S2B reflectance data, reveals a significant correlation between Chla and a combination of the blue (B2), green (B3), and red (B4) bands (coefficient of determination, R² = 0.95). The Chla estimation model is validated using a 30-point in-situ dataset collected on various dates (R² = 0.87; the root-mean-squared error RMSE < 5%). Subsequently, the model is applied to ten S2B images acquired from 2021 to 2023, revealing Chla's spatio-temporal distribution across the reservoir. Two key trends emerge: (1) Chla is lower during winter (November and December) than in summer and early autumn (July and September), and (2) The distribution of Chla undergoes noticeable spatial changes, particularly in July, with elevated levels observed in areas characterized by tourist hotspots. This approach shows promise for monitoring Chla in similar inland waters.