Remote-sensing retrieval of total phosphorus in Tiande Lake by optimizing an XGBoost machine learning model DOI

Aimin LI,

Xuan Kang,

Xiangyu Yan

et al.

Published: Nov. 15, 2023

Traditional models for total phosphorus (TP) retrieval from remote-sensing images generally show low accuracy. Additionally, parameter adjustment and selection of combinations machine learning (ML) exert significant influences on the regressive prediction effect model performance. To solve these problems, this research proposed an extreme gradient boosting (XGBoost) optimized by Bayesian optimization (BO), that is, BO-XGB. The optimal parameters are sought automatically a small sample size through BO, which shortens training time. Taking Tiande Lake in Zhengzhou City (Henan Province, China) as region interest, BO-XGB TP is established based GF1-WFV satellite data water-quality data. Moreover, accuracy compared with those other four ML methods, namely, XGBoost model, k-nearest neighbors (KNN), multilayer perceptron (MLP) random forest (RF). Compared models, demonstrates highest accuracy, coefficients determination R2 , root mean square error (RMSE), relative (MRE) separately 0.923, 2.15 × 10-3 mg/L, 1.81%. Finally, adopted to retrieve spatial distribution concentration Lake. results optimizing using BO can significantly improve algorithm more suitable retrieving mass findings have implications inverse modelling non-optical water quality such nitrogen(TN).

Language: Английский

High spatial resolution inversion of chromophoric dissolved organic matter (CDOM) concentrations in Ebinur Lake of arid Xinjiang, China: Implications for surface water quality monitoring DOI Creative Commons
Zhihui Li, Cheng Chen,

Naixin Cao

et al.

International Journal of Applied Earth Observation and Geoinformation, Journal Year: 2024, Volume and Issue: 132, P. 104022 - 104022

Published: July 10, 2024

Utilizing satellite remote sensing for the assessment and temporal-spatial analysis of Chromophoric Dissolved Organic Matter (CDOM) is vital overseeing lake water health devising management plans. This study focused on saline, turbid arid Ebinur Lake, located in China's northwestern region. The Random Forest (RF) eXtreme Gradient Boosting (XGBoost) model algorithms were compared to select one with highest accuracy. It combined Sentinel-2 data situ measurement quantitative inversion CDOM. Monthly CDOM distribution maps generated a 10 m resolution non-frozen months May October from 2018 2022, followed by comprehensive temporal trends. primary conclusions are: (1) XGBoost yielded highly accurate estimates, training set coefficient determination (R2) 0.94, Root Mean Square Error (RMSE) 0.06 mg/L, Absolute Percentage (MAPE) 6.05 %, Relative Percent Difference (RPD) 4.07; test demonstrated an R2 0.41 RMSE 0.22 MAPE 22.74 RPD 1.35; (2) Throughout period, main portion displayed variable spatial patterns indicated higher concentrations central part than nearshore areas decreasing tandem seasonable water-surface shrinkage. findings offer hints evaluation color parameters Lake practical references monitoring arid-region quality via sensing.

Language: Английский

Citations

0
Water Quality in the Ma’an Archipelago Marine Special Protected Area: Remote Sensing Inversion Based on Machine Learning DOI Creative Commons
Zhixin Wang,

Zhenqi Zhang,

Hailong Li

et al.

Journal of Marine Science and Engineering, Journal Year: 2024, Volume and Issue: 12(10), P. 1742 - 1742

Published: Oct. 3, 2024

Due to the increasing impact of climate change and human activities on marine ecosystems, there is an urgent need study water quality. The use remote sensing for quality inversion offers a precise, timely, comprehensive way evaluate present state future trajectories In this paper, model utilizing machine learning was developed variations in Ma’an Archipelago Marine Special Protected Area (MMSPA) over long-time series Landsat images. concentrations chlorophyll-a (Chl-a), phosphate, dissolved inorganic nitrogen (DIN) sea area from 2002 2022 were inverted analyzed. spatial temporal characteristics these investigated. results indicated that random forest could reliably predict Chl-a, DIN MMSPA. Specifically, Chl-a showed coefficient determination (R2) 0.741, root mean square error (RMSE) 3.376 μg/L, absolute percentage (MAPE) 16.219%. Regarding distribution, parameters notably elevated nearshore zones, especially northwest, contrasted with lower offshore southeast areas. Predominantly, regions higher proximity aquaculture zones. Additionally, nutrients originating land sources, transported via rivers such as Yangtze River, well influenced by activities, have shaped nutrient distribution. Over long term, MMSPA has shown considerable interannual fluctuations during past two decades. As sanctuary, preserving superior healthy ecosystem very important. Efforts protection, restoration, management will demand labor. Remote demonstrated its worth proficient technology real-time monitoring, capable supporting sustainable exploitation resources safeguarding ecological environment.

Language: Английский

Citations

0
Satellite Remote Sensing of Turbidity in Lake Xingkai Using Seven Years of Olci Observations DOI
Li Jian, Yang Li,

Kaishan Song

et al.

Published: Jan. 1, 2024

Language: Английский

Citations

0
Remote-sensing retrieval of total phosphorus in Tiande Lake by optimizing an XGBoost machine learning model DOI

Aimin LI,

Xuan Kang,

Xiangyu Yan

et al.

Published: Nov. 15, 2023

Traditional models for total phosphorus (TP) retrieval from remote-sensing images generally show low accuracy. Additionally, parameter adjustment and selection of combinations machine learning (ML) exert significant influences on the regressive prediction effect model performance. To solve these problems, this research proposed an extreme gradient boosting (XGBoost) optimized by Bayesian optimization (BO), that is, BO-XGB. The optimal parameters are sought automatically a small sample size through BO, which shortens training time. Taking Tiande Lake in Zhengzhou City (Henan Province, China) as region interest, BO-XGB TP is established based GF1-WFV satellite data water-quality data. Moreover, accuracy compared with those other four ML methods, namely, XGBoost model, k-nearest neighbors (KNN), multilayer perceptron (MLP) random forest (RF). Compared models, demonstrates highest accuracy, coefficients determination R2 , root mean square error (RMSE), relative (MRE) separately 0.923, 2.15 × 10-3 mg/L, 1.81%. Finally, adopted to retrieve spatial distribution concentration Lake. results optimizing using BO can significantly improve algorithm more suitable retrieving mass findings have implications inverse modelling non-optical water quality such nitrogen(TN).

Language: Английский

Citations

0