Crop-type identification is very important in agricultural regions. Most researchers in this area have focused on exploring the ability of synthetic-aperture radar (SAR) sensors to identify crops. This paper uses multi-source (Sentinel-1, Sentinel-2, and Landsat-8) and multi-temporal data to identify crop types. The change detection method was used to analyze spectral and indices information in time series. Significant differences in crop growth status during the growing season were found. Then, three obviously differentiated time features were extracted. Three advanced machine learning algorithms (Support Vector Machine, Artificial Neural Network, and Random Forest, RF) were used to identify the crop types. The results showed that the detection of (Vertical-vertical) VV, (Vertical-horizontal) VH, and Cross Ratio (CR) changes was effective for identifying land cover. Moreover, the red-edge changes were obviously different according to crop growth periods. Sentinel-2 and Landsat-8 showed different normalized difference vegetation index (NDVI) changes also. By using single remote sensing data to classify crops, Sentinel-2 produced the highest overall accuracy (0.91) and Kappa coefficient (0.89). The combination of Sentinel-1, Sentinel-2, and Landsat-8 data provided the best overall accuracy (0.93) and Kappa coefficient (0.91). The RF method had the best performance in terms of identity classification. In addition, the indices feature dominated the classification results. The combination of phenological period information with multi-source remote sensing data can be used to explore a crop area and its status in the growing season. The results of crop classification can be used to analyze the density and distribution of crops. This study can also allow to determine crop growth status, improve crop yield estimation accuracy, and provide a basis for crop management.

Soil total nitrogen (STN) is an important indicator of soil quality and plays a key role in global nitrogen cycling. Accurate prediction of STN content is essential for the sustainable use of soil resources. Synthetic aperture radar (SAR) provides a promising source of data for soil monitoring because of its all-weather, all-day monitoring, but it has rarely been used for STN mapping. In this study, we explored the potential of multi-temporal Sentinel-1 data to predict STN by evaluating and comparing the performance of boosted regression trees (BRTs), random forest (RF), and support vector machine (SVM) models in STN mapping in the middle reaches of the Heihe River Basin in northwestern China. Fifteen predictor variables were used to construct models, including land use/land cover, multi-source remote sensing-derived variables, and topographic and climatic variables. We evaluated the prediction accuracy of the models based on a cross-validation procedure. Results showed that tree-based models (RF and BRT) outperformed SVM. Compared to the model that only used optical data, the addition of multi-temporal Sentinel-1A data using the BRT method improved the root mean square error (RMSE) and the mean absolute error (MAE) by 17.2% and 17.4%, respectively. Furthermore, the combination of all predictor variables using the BRT model had the best predictive performance, explaining 57% of the variation in STN, with the highest R² (0.57) value and the lowest RMSE (0.24) and MAE (0.18) values. Remote sensing variables were the most important environmental variables for STN mapping, with 59% and 50% relative importance in the RF and BRT models, respectively. Our results show the potential of using multi-temporal Sentinel-1 data to predict STN, broadening the data source for future digital soil mapping. In addition, we propose that the SVM, RF, and BRT models should be calibrated and evaluated to obtain the best results for STN content mapping in similar landscapes.

The objective of this research was to investigate the impact of seasonality on urban land-cover mapping and to explore better classification accuracy by using multi-season Sentinel-1A and GF-1 wide field view (WFV) images, and the combinations of both types of images in subtropical monsoon-climate regions in Southeast China. We obtained multi-season Sentinel-1A and GF-1 WFV images, as well as the combinations of both data, by using a support vector machine (SVM) and a random forest (RF) classifier. The backscatter intensity, texture, and interference-coherence images were extracted from Sentinel-1A images, and different combinations of these Sentinel-1A-derived images were used to evaluate their ability to map urban land cover. The results showed that the performance of winter images was better than that of any other season, while the summer images performed the worst. Higher classification accuracy was achieved by using multi-season images, and satisfactory classification results were obtained when using Sentinel-1A images from only three seasons. The best classification result was achieved using a combination of all Sentinel-1A data from all four seasons and GF-1 WFV data from winter, with an overall accuracy of up to 96.02% and a kappa coefficient reaching 0.9502. The performance of textures was slightly better than that of the backscatter-intensity images. Although the coherence data performed the worst, it was still able to distinguish urban impervious surfaces well. In addition, the overall classification accuracy of RF was better than that of SVM.

In this study, a low-cost, self-balancing crop spectral reflectance sensor (CSRS) was designed for real-time, nondestructive monitoring of the spectral reflectance and vegetation index of crops such as tomato and rapeseed. The sensor had a field of view of 30°, and a narrow-band filter was used for light splitting. The filter’s full width at half-maximum was 10 nm, and the spectral bands were 710 nm and 870 nm. The sensor was powered by a battery and used WiFi for communication. Its software was based on the Contiki operating system. To make the sensor work in different light intensity conditions, the photoelectric conversion automatic gain circuit had a total of 255 combinations of amplification. The gimbal of the sensor was mainly composed of an inner ring and an outer ring. Under the gravity of the sensor, the central axis of the sensor remained vertical, such that the up-facing and down-facing photosensitive units stayed in the horizontal position. The mechanical components of the sensor were designed symmetrically to facilitate equal mass distribution and to meet the needs of automatic balancing. Based on the optical signal transmission process of the sensor and the dark-current characteristics of the photodetector, a calibration method was theoretically deduced, which improved the accuracy and stability of the sensor under different ambient light intensities. The calibration method is also applicable for the calibration of other crop growth information sensors. Next, the standard reflectance gray scale was taken as the measurement variable to test the accuracy of the sensor, and the results showed that the root mean square error of the reflectance measured by the sensor at 710 nm and 870 nm was 1.10% and 1.27%, respectively; the mean absolute error was 0.95% and 0.89%, respectively; the relative error was below 4% and 3%, respectively; and the coefficient of variation was between 1.0% and 2.5%. The reflectance data measured by the sensor under different ambient light intensities suggested that the absolute error of the sensor was within ±0.5%, and the coefficients of variation at the two spectral bands were 1.04% and 0.39%, respectively. With tomato and rapeseed as the monitoring targets, the proposed CSRS and a commercial spectroradiometer were used to measure at the same time. The results showed that the reflectance measured by the two devices was very close, and there was a linear relationship between the normalized difference vegetation index of the CSRS and that of the commercial spectroradiometer. The coefficient of determination (R²) for tomato and rapeseed were 0.9540 and 0.9110, respectively.

Accurate estimation of the degree of regeneration in tropical dry forest (TDF) is critical for conservation policymaking and evaluation. Hyperspectral remote sensing and light detection and ranging (LiDAR) have been used to characterize the deterministic successional stages in a TDF. These successional stages, classified as early, intermediate, and late, are considered a proxy for mapping the age since the abandonment of a given forest area. Expanding on the need for more accurate successional forest mapping, our study considers the age attributes of a TDF study area as a continuous expression of relative attribute scores/levels that vary along the process of ecological succession. Specifically, two remote-sensing data sets: HyMap (hyperspectral) and LVIS (waveform LiDAR), were acquired at the Santa Rosa National Park Environmental Monitoring Super Site (SRNP-EMSS) in Costa Rica, were used to generate age-attribute metrics. These metrics were then used as entry-level variables on a randomized nonlinear archetypal analysis (RNAA) model to select the most informative metrics from both data sets. Next, a relative attribute learning (RAL) algorithm was adapted for both independent and fused metrics to comparatively learn the relative attribute levels of the forest ages of the study area. In this study, four HyMap indices and five LVIS metrics were found to have the potential to map the forest ages of the study area, and compared with these results, a significant improvement was found through the fusion of the metrics on the accuracy of the generated forest age maps. By linking the age group mapping and the relative attribute mapping results, a dynamic gradient of the age-attribute transition patterns emerged.

The use of multi-mode remote sensing data for biomass prediction is of potential value to aid planting management and yield maximization. In this study, an advanced biomass estimation approach for sugarcane fields is proposed based on multi-source remote sensing data. Since feature interpretability in agricultural data mining is significant, a feature extraction method of deep archetypal analysis (DAA) that has good model interpretability is introduced and aided by principal component analysis (PCA) for feature mining from the multi-mode multispectral and light detection and ranging (LiDAR) remote sensing data pertaining to sugarcane. In addition, an integrated regression model integrating random forest regression, support vector regression, K-nearest neighbor regression and deep network regression is developed after feature extraction by DAA to precisely predict biomass of sugarcane. In this study, the biomass prediction performance achieved using the proposed integrated learning approach is found to be predominantly better than that achieved by using conventional linear methods in all the time periods of plant growth. Of more significance, according to model interpretability of DAA, only a small set of informative features maintaining their physical meanings (four informative spectral indices and four key LiDAR metrics) can be extracted which eliminates the redundancy of multi-mode data and plays a vital role in accurate biomass prediction. Therefore, the findings in this study provide hands-on experience to planters with indications of the key or informative spectral or LiDAR metrics relevant to the biomass to adjust the corresponding planting management design.