This study aims to apply Moderate Resolution Imaging Spectroradiometer (MODIS) data to monitor water quality parameters of total phosphorus (TP) and total nitrogen (TN) in Balik Lake since TP and TN are important factors for eutrophication as well as it is essential to monitor TP and TN concentrations accurately for the management of water environments. Water chemical variables, including TN and TP, are soluble and optically inactive. For such a purpose, remote sensing (RS) enables an alternative way to monitor the optical activity of water with some limitations. In this study, TN and TP concentrations of Balik Lake are examined by RS data. The artificial neural networking (ANN) model was used to reveal the relation between the reflectance bands of MODIS along with the water quality parameters of TP and TN concentrations in aquatic environments by utilizing ground truth water samples collected concurrently with the MODIS overpass. The dataset used in the study consists of MODIS Aqua bands from 1 to 7 and ground truth observations of TP and TN from the study area of Balik Lake during 2017–2019. The development and testing of a feedforward back-propagation multilayer perceptron artificial neural network were performed by using the Levenberg-Marquardt training algorithm to predict TP and TN. The accuracy of the ANN model was determined to be relatively higher in the testing stage of TP prediction with R = 0.59 for Model-1 (3-10-1) and R = 0.56 for Model-2 (7-10-1) while it is lower in the testing stage of TN prediction with R = 0.36 for Model-1 and Model-2. The findings of this study indicated that the reflectance bands of MODIS have a moderate potential to monitor the solubility variables of the water quality for the management of the lake. However, the ANN model introduced in this study is expected to ensure an improvement in the prediction accuracy of inactive water quality variables based on the RS data.

本研究旨在应用中分辨率成像光谱仪 (MODIS) 数据监测巴厘岛湖中总磷 (TP) 和总氮 (TN) 的水质参数，因为 TP 和 TN 是富营养化的重要因素，而且监测必不可少。用于水环境管理的 TP 和 TN 浓度准确。水的化学变量，包括 TN 和 TP，是可溶的且不具有光学活性。为此，遥感 (RS) 提供了一种替代方法来监测水的光学活动，但存在一些限制。在这项研究中，巴厘湖的 TN 和 TP 浓度通过 RS 数据进行了检查。利用人工神经网络 (ANN) 模型，利用与 MODIS 立交桥同时采集的地面实况水样，揭示了 MODIS 反射带与水生环境中 TP 和 TN 浓度的水质参数之间的关系。研究中使用的数据集包括 1 到 7 的 MODIS Aqua 波段以及 2017-2019 年期间来自 Balik 湖研究区的 TP 和 TN 的地面实况观测。前馈反向传播多层感知器人工神经网络的开发和测试是通过使用 Levenberg-Marquardt 训练算法来预测 TP 和 TN 的。在TP预测的测试阶段，ANN模型的准确性被确定为相对较高，模型1（3-10-1）的R = 0.59和R = 0。Model-2 (7-10-1) 为 56，而在 TN 预测的测试阶段较低，Model-1 和 Model-2 的 R = 0.36。这项研究的结果表明，MODIS 的反射带具有监测水质溶解度变量以管理湖泊的中等潜力。然而，本研究中引入的人工神经网络模型有望确保提高基于 RS 数据的非活性水质变量的预测精度。