Although efforts and progress have been made in crop classification using optical remote sensing images, it is still necessary to make full use of the high spatial, temporal, and spectral resolutions of remote sensing images. However, with the increasing volume of remote sensing data, a key emerging issue in the field of crop classification is how to find useful information from massive data to balance classification accuracy and processing time. To address this challenge, we developed a novel crop classification method, combining optimal feature selection (OFSM) with hybrid convolutional neural network-random forest (CNN-RF) networks for multi-temporal optical remote sensing images. This research used 234 features including spectral, segmentation, color, and texture features from three scenes of Sentinel-2 images to identify crop types in the Jilin province of northeast China. To effectively extract the effective features of remote sensing data with lower time requirements, the use of OFSM was proposed with the results compared with two traditional feature selection methods (TFSM): random forest feature importance selection (RF-FI) and random forest recursive feature elimination (RF-RFE). Although the time required for OFSM was 26.05 s, which was between RF-FI with 1.97 s and RF-RFE with 132.54 s, OFSM outperformed RF-FI and RF-RFE in terms of the overall accuracy (OA) of crop classification by 4% and 0.3%, respectively. On the basis of obtaining effective feature information, to further improve the accuracy of crop classification we designed two hybrid CNN-RF networks to leverage the advantages of one-dimensional convolution (Conv1D) and Visual Geometry Group (VGG) with random forest (RF), respectively. Based on the selected optimal features using OFSM, four networks were tested for comparison: Conv1D-RF, VGG-RF, Conv1D, and VGG. Conv1D-RF achieved the highest OA at 94.27% as compared with VGG-RF (93.23%), Conv1D (92.59%), and VGG (91.89%), indicating that the Conv1D-RF method with optimal feature input provides an effective and efficient method of time series representation for multi-temporal crop-type classification.

Chlorophyll-a (Chl-a) is an important characterized parameter of lakes. Monitoring it accurately through remote sensing is thus of great significance for early warnings of water eutrophication. Sentinel Multispectral Imager (MSI) images from May to September between 2020 and 2021 were used along with in-situ measurements to estimate Chl-a in Lake Chagan, which is located in Jilin Province, Northeast China. In this study, the extreme gradient boosting (XGBoost) and Random Forest (RF) models, which had similar performances, were generated by six single bands and six band combinations. The RF model was then selected based on the assessments (R² = 0.79, RMSE = 2.51 μg L⁻¹, MAPE = 9.86%), since its learning of the input features in the model conformed to the bio-optical properties of Case 2 waters. The study considered Chl-a concentrations in Lake Chagan as a seasonal pattern according to the K-Nearest-Neighbors (KNN) classification. The RF model also showed relatively stable performance for three seasons (spring, summer and autumn) and it was applied to map Chl-a in the whole lake. The research presents a more reliable machine learning (ML) model with higher precision than previous empirical models, as shown by the effects of the input features linked with the biological mechanisms of Chl-a. Its robustness was revealed by the temporal and spatial distributions of Chl-a concentrations, which were consistent with in-situ measurements in the map. This research was capable of revealing the current ecological situation in Lake Chagan and can serve as a reference in remote sensing of inland lakes.

Spring soil moisture (SM) is of great importance for monitoring agricultural drought and waterlogging in farmland areas. While winter snow cover has an important impact on spring SM, relatively little research has examined the correlation between winter snow cover and spring SM in great detail. To understand the effects of snow cover on SM over farmland, the relationship between winter snow cover parameters (maximum snow depth (MSD) and average snow depth (ASD)) and spring SM in Northeast China was examined based on 30 year passive microwave snow depth (SD) and SM remote-sensing products. Linear regression models based on winter snow cover were established to predict spring SM. Moreover, 4 year SD and SM data were applied to validate the performance of the linear regression models. Additionally, the effects of meteorological factors on spring SM also were analyzed using multiparameter linear regression models. Finally, as a specific application, the best-performing model was used to predict the probability of spring drought and waterlogging in farmland in Northeast China. Our results illustrated the positive effects of winter snow cover on spring SM. The average correlation coefficient (R) of winter snow cover and spring SM was above 0.5 (significant at a 95% confidence level) over farmland. The performance of the relationship between snow cover and SM in April was better than that in May. Compared to the multiparameter linear regression models in terms of fitting coefficient, MSD can be used as an important snow parameter to predict spring drought and waterlogging probability in April. Specifically, if the relative SM threshold is 50% when spring drought occurs in April, the prediction probability of the linear regression model concerning snow cover and spring SM can reach 74%. This study improved our understanding of the effects of winter snow cover on spring SM and will be beneficial for further studies on the prediction of spring drought.

Soil moisture (SM) retrieval from SMOS (the Soil Moisture and Ocean Salinity mission) and SMAP (the Soil Moisture Active/Passive mission) passive microwave data over forested areas with required accuracy is of great significance and poses some challenges. In this paper, we used Ground Wireless Sensor Network (GWSN) SM measurements from 9 September to 5 November 2015 to validate SMOS and SMAP Level 3 (L3) SM products over forested areas in northeastern China. Our results found that neither SMOS nor SMAP L3 SM products were ideal, with respective RMSE (root mean square error) values of 0.31 cm³/cm³ and 0.17 cm³/cm³. Nevertheless, some improvements in SM retrieval might be achievable through refinements of the soil dielectric model with respect to high percentage of soil organic matter (SOM) in the forested area. To that end, the potential of the semi-empirical soil dielectric model proposed by Jun Liu (Liu’s model) in improving SM retrieval results over forested areas was investigated. Introducing Liu’s model into the retrieval algorithms of both SMOS and SMAP missions produced promising results. For SMAP, the RMSE of L3 SM products improved from 0.16 cm³/cm³ to 0.07 cm³/cm³ for AM (local solar time around 06:00 am) data, and from 0.17 cm³/cm³ to 0.05 cm³/cm³ for PM (local solar time around 06:00 pm) data. For SMOS ascending orbit products, the accuracy was improved by 56%, while descending orbit products improved by 45%.

Snow cover plays a highly critical role in the global water cycle and energy exchange. Accurate snow depth (SD) data are important for research on hydrologic processes, climate change, and natural disaster prediction. However, existing passive microwave (PMW) SD products have high uncertainty in Northeast China owing to their coarse spatial resolution. Surface environment parameters should also be considered to reduce errors in existing SD products. Otherwise, it is difficult to accurately capture snow spatiotemporal variations, especially in a complex environment (e.g., mountain or forests areas). To improve the inversion accuracy and spatial resolution of existing SD products in Northeast China, a multifactor SD downscaling model was developed by combining PMW SD data from the AMSR2 sensor, optical snow cover extent data, and surface environmental parameters to produce fine scale (500 m × 500 m) and high precision SD data. Validations at 98 ground meteorological stations show that the developed model greatly improved the spatial resolution and inversion accuracy of the raw AMSR2 SD product; its root-mean-square error (RMSE) reduced from 26.15 cm of the raw product to 7.58 cm, and the correlation coefficient (R) increased from 0.39 to 0.53. For other SD products (WESTDC and FY), the multifactor SD downscaling model still has good applicability, it could further improve the performance of the WESTDC and FY SD products in time and space and achieve better inversion accuracy than raw SD products. Furthermore, the proposed model exhibited good agreement with the observed SD data in a field quadrat (3 km × 2 km) within the fine scale, with an error ranging between −2 and 2 cm. Compared with the existing downscaling methods, the proposed model presented the best performance.

Wide mode SAR images have an apparent incidence angle effect. The existing incident angle normalization methods assume that the relationship between the incident angle (θ) and the backscattering coefficient (σ_PQ) does not change with the growth stage of crops, which is in conflict with the real-life situation. Therefore, the normalization results of σ_PQ based on these existing methods will affect the accuracy of object classification, target recognition, and land surface parameter inversion. Here, the change in θ-σ_PQ relationship was investigated based on time-series (April to October) σ_PQ of maize canopies in northeast China, and a dynamic method based on normalized difference vegetation index (NDVI) was developed to normalize the effect of θ on σ_PQ. Through the accuracy evaluation, the following conclusions are obtained: (1) the dependence (referring to N) of Sentinel 1 C-band σ_PQ on θ varies with maize NDVI. In addition, the value of N changed from 9.35 to 0.66 at VV polarization from bare soil to biomass peak, and from 6.26 to 0.99 at VH polarization; (2) a dynamic method was proposed to quantify the change of N based on its strong correlation with NDVI, indicated by R² of 0.82 and 0.80 for VV and VH polarization, respectively; and (3) the overall root mean square error of normalized σ_PQ based on the newly-developed dynamic method is 0.51 dB, and this accuracy outperforms the original first-order cosine method (1.37 dB) and cosine square law method (1.08 dB) by about 63% and 53% on the whole. This study provides a dynamic framework for normalizing radar backscatter coefficient, improving the retrieval accuracy of land surface parameters from radar remote sensing.