Feasibility of empirical remote sensing for estimating total nitrogen in the coral reef lagoon of Yoron Island, Japan

Land-derived nutrient enrichment is a major stressor affecting coral reef lagoon ecosystems, yet its spatial distribution remains poorly resolved. This study investigated the feasibility of using empirical remote sensing to estimate total nitrogen (T-N) concentrations in the eastern reef lagoon of Yoron Island, southwestern Japan, using Landsat imagery and limited in-situ observations. Empirical models were developed by testing multiple regression forms for individual spectral bands against five in-situ T-N measurements. Model performance was evaluated using the coefficient of determination (R2), F-test, t-test, and Root Mean Square Error (RMSE). Owing to the limited calibration dataset and the absence of an independent validation dataset, resulting model should be regarded as a preliminary, feasibility-level empirical relationship rather than a validated predictive model. Among the tested models, a quadratic polynomial based on Landsat-8 blue band reflectance showed the best relative performance and was applied to Landsat-8 imagery acquired in 2014, 2020, and 2023 to map the spatial distribution of T-N. Additional Landsat-5 TM and Landsat-7 ETM+ imagery (1984–2010) was used to provide qualitative historical context only because cross-sensor harmonization was not performed. Elevated T-N near groundwater discharge zones points to submarine groundwater discharge and runoff as key nutrient pathways into the lagoon. Though ecological impacts weren’t directly tested, the patterns flag areas of likely vulnerability, demonstrating that satellite-based modelling can map nutrient enrichment in shallow reef lagoons and setting up future integration with hydrological or ecological data.