Monitoring water quality of inland lakes and reservoirs is a great concern for the public and government in China. Water turbidity is a reliable and direct indicator that can reflect the water quality. Remote sensing has become an efficient technology for monitoring large-scale water turbidity. This study aims to search an optimal regression model to accurately predict water turbidity using remote sensing data. To achieve this goal, 187 water samples were collected from field campaigns across Northeast China, in 2018, of which the samples were gathered within ±6 days of Sentinel-2 overpasses. The spectral reflectance data was used as independent variables for modeling. The simple regression, partial least squares regression, support vector regression, extreme learning machine, back-propagation neural network, classification and regression tree, gradient boosting decision tree (GBDT), random forest (RF), and K-nearest neighbor were used to compare. From model validation, we identified GBDT as the best regression model ( R ² = 0.88, RMSE = 9.90 NTU, MAE = 6.71 NTU). We applied GBDT to retrieve the water turbidity and obtained a satisfactory result. Feature selection technique from tree-based ensemble method was also tested. We selected B2, B3, B4, and B5 as the important variables because of their high ability to explain the variation of turbidity. These results demonstrated the significance of using a promising method to retrieve water turbidity using Sentinel-2 imagery at the regional scale. It is beneficial to monitor the spatial-temporal distribution of water turbidity; support water quality management and inland water environment protection.

Lithology classification is important in mineral resource exploration, engineering geological exploration, and disaster monitoring. Traditional laboratory methods for the qualitative analysis of rocks are limited by sampling conditions and analytical techniques, resulting in high costs, low efficiency, and the inability to quickly obtain large-scale geological information. Hyperspectral remote sensing technology can classify and identify lithology using the spectral characteristics of rock, and is characterized by fast detection, large coverage area, and environmental friendliness, which provide the application potential for lithological mapping at a large regional scale. In this study, ZY1-02D hyperspectral images were used as data sources to construct a new two-layer extreme gradient boosting (XGBoost) lithology classification model based on the XGBoost decision tree and an improved greedy search algorithm. A total of 153 spectral bands of the preprocessed hyperspectral images were input into the first layer of the XGBoost model. Based on the tree traversal structural characteristics of the leaf nodes in the XGBoost model, three built-in XGBoost importance indexes were split and combined. The improved greedy search algorithm was used to extract the spectral band variables, which were imported into the second layer of the XGBoost model, and the bat algorithm was used to optimize the modeling parameters of XGBoost. The extraction model of rock classification information was constructed, and the classification map of regional surface rock types was drawn. Field verification was performed for the two-layer XGBoost rock classification model, and its accuracy and reliability were evaluated based on four indexes, namely, accuracy, precision, recall, and F1 score. The results showed that the two-layer XGBoost model had a good lithological classification effect, robustness, and adaptability to small sample datasets. Compared with the traditional machine learning model, the two-layer XGBoost model shows superior performance. The accuracy, precision, recall, and F1 score of the verification set were 0.8343, 0.8406, 0.8350, and 0.8157, respectively. The variable extraction ability of the constructed two-layer XGBoost model was significantly improved. Compared with traditional feature selection methods, the GREED-GFC method, when applied to the two-layer XGBoost model, contributes to more stable rock classification performance and higher lithology prediction accuracy, and the smallest number of extracted features. The lithological distribution information identified by the model was in good agreement with the lithology information verified in the field.

Snow depth (SD) is an indispensable parameter for many studies. Launched in 2018, the Ice, Cloud and land Elevation Satellite-2 (ICESat-2) is designed to obtain global glacial elevations, but it can also acquire canopy and terrain elevations. Whether the depth of seasonal snow can be estimated by directly comparing the difference in elevations in snow-cover and snow-free cases, many people may reasonably ask. In this letter, we conduct such an investigation in Altay, Northwest China, using ICESat-2 ATL08 elevation products. Our investigation suggests: 1) in mountainous areas, the answer maybe is no because the estimation is obviously affected by rugged topography; 2) but in flat regions, SDs have been effectively estimated. (The $R^{2}$ is up to 0.88 between estimates and ground measurements.); and 3) as expected, land-cover types also affect the accuracy of the results, and the best estimation happens over the type of bare land. Therefore, estimating the depth of seasonal snow from the ICESat-2 product may be feasible, but we must check the results carefully.

Optical remote sensing images with high temporal resolution can be used to monitor lake ice phenology, a periodic freezing and thawing cycle of ice resulting from seasonal and inter-annual climate variations. In the research reported here, we used MODIS satellite data to establish the time series of lake ice extent and extracted lake ice phenology dates and durations for eight large typical lakes in Northeast China for the hydrological years from 2003 to 2016. The MODIS-derived results were validated against ice records at hydrological stations. The mean absolute error for a freeze-up start (FUS), freeze-up end (FUE), break-up start (BUS) and break-up end (BUE) was 3.1, 4.8, 6.6 and 6.6 days, respectively. Our findings indicated that the investigated lakes were tending to freeze later and melt earlier and were frozen for a shortened period over time. FUS was experiencing a delay of 0.65 days per year and BUE was advancing by 0.19 days per year, implying a decrease of frozen duration (FD) of 0.84 days per year taking all eight lakes into consideration. The lake ice duration increased with latitude, and the lakes with a relatively smaller area had a higher yearly rate of change and were more variable compared with the larger ones. The relationship between lake ice phenology and other influencing factors was evaluated using correlation coefficients and partial least squares regression. The results showed that the freeze-up process was more dependent on the lake morphometry, while the break-up process was more dependent on climate changes, particularly on air temperature, which had the highest correlation coefficient (r = −0.69, p < 0.01).