Search

Timely and accurate crop yield information can ensure regional food security. In the field of predicting crop yields, deep learning techniques such as long short-term memory (LSTM) and convolutional neural networks (CNN) are frequently employed. Many studies have shown that the predictions of models combining the two are better than those of single models. Crop growth can be reflected by the vegetation index calculated using data from remote sensing. However, the use of pure remote sensing data alone ignores the spatial heterogeneity of different regions. In this paper, we tested a total of three models, CNN-LSTM, CNN and convolutional LSTM (ConvLSTM), for predicting the annual rice yield at the county level in Hubei Province, China. The model was trained by ERA5 temperature (AT) data, MODIS remote sensing data including the Enhanced Vegetation Index (EVI), Gross Primary Productivity (GPP) and Soil-Adapted Vegetation Index (SAVI), and a dummy variable representing spatial heterogeneity; rice yield data from 2000–2019 were employed as labels. Data download and processing were based on Google Earth Engine (GEE). The downloaded remote sensing images were processed into normalized histograms for the training and prediction of deep learning models. According to the experimental findings, the model that included a dummy variable to represent spatial heterogeneity had a stronger predictive ability than the model trained using just remote sensing data. The prediction performance of the CNN-LSTM model outperformed the CNN or ConvLSTM model. [ABSTRACT FROM AUTHOR]

Total phosphorus (TP) concentration is high in countless small inland waterbodies in Hubei province, middle China, which is threating the water environment. However, there are almost no ground-based water quality monitoring points in small inland waterbodies, because the cost of time, labor, and money is high and it does not meet the needs of spatiotemporal dynamic monitoring. Remote sensing provides an effective tool for TP concentration monitoring spatiotemporally. However, monitoring the TP concentration of small inland waterbodies is challenging for satellite remote sensing due to the inadequate spatial resolution. Recently, unmanned aerial vehicles (UAV) have been applied to quantitatively retrieve the spatiotemporal distribution of TP concentration without the challenges of cloud cover and atmospheric effects. Although state-of-the-art algorithms to retrieve TP concentration have been improved, specific models are only used for specific water quality parameters or regions, and there are no robust and reliable TP retrieval models for small inland waterbodies at this time. To address this issue, six machine learning methods optimized by intelligent optimization algorithms (IOA-ML models) have been developed to quantitatively retrieve TP concentration combined with the reflectance of original bands and selected band combinations of UAV multispectral images. We evaluated the performances of models in terms of coefficient of determination (R2), root mean squared error (RMSE), and residual prediction deviation (RPD). The results showed that the R2 of the six IOA-ML models for training, validation, and test sets were 0.8856–0.984, 0.8054–0.8929, and 0.7462–0.9045, respectively, indicating the methods had high precision and transferability. The extreme gradient boosting optimized by genetic algorithm (GA-XGB) performed best, with the highest precision for the validation and test sets. The spatial distribution of TP concentration of each flight derived from different models had similar distribution characteristics. This paper provides a reference for promoting the intelligent and automatic level of water environment monitoring in small inland waterbodies. [ABSTRACT FROM AUTHOR]

Forest logging detection is important for sustainable forest management. The traditional optical satellite images with visible and near-infrared bands showed the ability to identify intensive timber logging. However, less intensive logging is still difficult to detect with coarse spatial resolution such as Landsat or high spatial resolution in fewer spectral bands. Although more high-resolution remote sensing images containing richer spectral bands can be easily obtained nowadays, the questions of whether they facilitate the detection of logging patterns and which spectral bands are more effective in detecting logging patterns, especially in selective logging, remain unresolved. Therefore, this paper aims to evaluate the combinations of visible, near-infrared, red-edge, and short-wave infrared bands in detecting three different logging intensity patterns, including unlogged (control check, CK), selective logging (SL), and clear-cutting (CC), in north subtropical plantation forests with the random forest algorithm using Sentinel-2 multispectral imagery. This study aims to explore the recognition performance of different combinations of spectral bands (visual (VIS) and near-infrared bands (NIR), VIS, NIR combined with red-edge, VIS, NIR combined with short-wave infrared bands (SWIR), and full-spectrum bands combined with VIS, NIR, red edge and SWIR) and to determine the best spectral variables to be used for identifying logging patterns, especially in SL. The study was conducted in Taizishan in Hubei province, China. A total of 213 subcompartments of different logging patterns were collected and the random forest algorithm was used to classify logging patterns. The results showed that full-spectrum bands which contain the red-edge and short-wave infrared bands improve the ability of conventional optical satellites to monitor forest logging patterns and can achieve an overall accuracy of 85%, especially for SL which can achieve 79% and 64% for precision and recall accuracy, respectively. The red-edge band (698–713 nm, B5 in Sentinel-2), short-wave infrared band (2100–2280 nm, B12 in Sentinel-2), and associated vegetation indices (NBR, NDre2, and NDre1) enhance the sensitivity of the spectral information to logging patterns, especially for the SL pattern, and the precision and recall accuracy can improve by 10% and 6%, respectively. Meanwhile, both clear-cutting and unlogged patterns could be well-classified whether adding a red-edge or SWIR band or both in VIS and NIR bands; the best precision and recall accuracies for clear-cutting were enhanced to 97%, 95% and 81%, 91% for unlogged, respectively. Our results demonstrate that the optical images have the potential ability to detect logging patterns especially for the clear-cutting and unlogged patterns, and the selective logging detection accuracy can be improved by adding red-edge and short-wave infrared spectral bands. [ABSTRACT FROM AUTHOR]

Efficient and accurate landslide recognition is crucial for disaster prevention and post-disaster rescue efforts. However, compared to machine learning, deep learning approaches currently face challenges such as long model runtimes and inefficiency. To tackle these challenges, we proposed a novel knowledge distillation network based on Swin-Transformer (Distilled Swin-Transformer, DST) for landslide recognition. We created a new landslide sample database and combined nine landslide influencing factors (LIFs) with remote sensing images (RSIs) to evaluate the performance of DST. Our approach was tested in Zigui County, Hubei Province, China, and our quantitative evaluation showed that the combined RSIs with LIFs improved the performance of the landslide recognition model. Specifically, our model achieved an Overall Accuracy (OA), Precision, Recall, F1-Score (F1), and Kappa that were 0.8381%, 0.6988%, 0.9334%, 0.8301%, and 0.0125 higher, respectively, than when using only RSIs. Compared with the results of other neural networks, namely ResNet50, Swin-Transformer, and DeiT, our proposed deep learning model achieves the best OA (98.1717%), Precision (98.1672%), Recall (98.1667%), F1 (98.1615%), and Kappa (0.9766). DST has the lowest number of FLOPs, which is crucial for improving computational efficiency, especially in landslide recognition applications after geological disasters. Our model requires only 2.83 GFLOPs, which is the lowest among the four models and is 1.8242 GFLOPs, 1.741 GFLOPs, and 2.0284 GFLOPs less than ResNet, Swin, and DeiT, respectively. The proposed method has good applicability in rapid recognition scenarios after geological disasters. [ABSTRACT FROM AUTHOR]

Nighttime light (NTL) remote sensing data have been widely used to derive socioeconomic indicators at the national and regional scales to study regional economic development. However, most previous studies only chose a single measurement indicator (such as GDP) and adopted simple regression methods to investigate the economic development of a certain area based on DMSP-OLS or NPP-VIIRS stable NTL data. The status quo shows the problems of using a single evaluation index—it has a low evaluation precision. The LJ1-01 satellite is the first dedicated NTL remote sensing satellite in the world, launched in July 2018. The data provided by LJ1-01 have a higher spatial resolution and fewer blooming phenomena. In this paper, we compared the accuracy of the LJ1-01 data and NPP-VIIRS data in detecting county-level multidimensional economic development. In three provinces in China, namely, Hubei, Hunan and Jiangxi, 20 socioeconomic parameters were selected from the following five perspectives: economic conditions, people's livelihood, social development, public resources and natural vulnerability. Then, a County-level Economic Index (CEI) was constructed to evaluate the level of multidimensional economic development, with the spatial pattern of the multidimensional economic development also identified across the study area. The present study adopted the random forest (RF) and linear regression (LR) algorithms to establish the regression model individually, and the results were evaluated by cross-validation. The results show that the RF algorithm greatly improves the accuracy of the model compared with the LR algorithm, and thus is suitable for the study of NTL data. In addition, a better determinate coefficient (R2) based on the LJ1-01 data (0.8168) was obtained than that from the NPP-VIIRS data (0.7245) in the RF model, which reflects that the LJ1-01 data offer better potential in the evaluation of socioeconomic parameters and can be used to identify, both accurately and efficiently, multidimensional economic development at the county level. [ABSTRACT FROM AUTHOR]

Regional remote sensing image products are playing an important role in an increasing number of application fields. Aiming at multi-satellite imaging task planning for large-area image acquisition, this paper proposes a multi-objective modeling method. First, we analyzed the core requirements of regional mapping for multi-satellite imaging mission planning: Full coverage of the target area and low consumption of satellite resources. Second, an optimization model with two objective functions, namely the maximum target area coverage and minimum satellite resource utilization, was established. Using the selection of imaging strips and their swing angles as two types of decision variables, the regional decomposition and satellite resource allocation were integrated into the planning model. Third, two efficient algorithms, Vatti and non-dominated sorting genetic algorithm (NSGA-II), were used for objective function calculation and model solving, respectively. Finally, the experiments used Hubei, Finland, and Congo as the target areas and GF1, GF6, ZY1-02C, and ZY3 as imaging satellites to verify the modeling method proposed in this paper. The experiments showed that the proposed multi-objective modeling method could complete the coverage of regional targets with fewer satellite resources and improve the satellite application efficiency significantly. [ABSTRACT FROM AUTHOR]

As China is currently in a critical period of transition to green development, a quantitative analysis of the coupled synergy between economic growth, population migration, and environmental protection performance can provide a rational reference for the formulation of macro-policies in relevant regions. Based on these objectives, this study built a matching analysis framework between gross domestic production (GDP) density (GD), population density (PD), and gross ecosystem production (GEP) density (ED) to analyze the spatial heterogeneity of these three indicators among 362 municipal units in different regions of the Chinese Mainland from 2000 to 2020 based on satellite remote sensing images and statistical data. The spatial adaptability between them was explored by employing a center-of-gravity model. The findings of this study show that: (1) the GD, PD, and ED on the Chinese Mainland exhibited varying degrees of spatial heterogeneity on both sides of the Hu Line during the investigation period, with the general feature of being higher in the southeastern region and lower in the northwestern region; (2) the centers of gravity of GD, PD, and ED were all located in Hubei Province during the investigation period. The centers of gravity of PD and GD shifted 79.39 km and 109.72 km to the southwest, respectively, whereas the center of gravity of ED, remained relatively stable during the investigation period; and (3) the center of gravity distances between PD–GD, ED–PD, and ED–GD in 2020 were 99.31 km, 247.52 km, and 346.27 km, respectively, and the percentages of highly matched units ranked among the 362 samples were 72.93%, 23.48%, and 25.69% for GD–PD, GD–ED, and ED–PD, respectively. This study concluded that a synergistic spatial pattern of the population, economic layout, and land use on the Chinese Mainland has not yet been formed. Therefore, this study suggests that future policies should be committed to promoting the northwest Chinese Mainland movement of the center of gravity of GD and PD, as well as the southeast movement of the ED center. [ABSTRACT FROM AUTHOR]

The traditional mine remote sensing information pre-survey is mainly based on manual interpretation, and interpreters delineate the mine boundary shape. This work is difficult and susceptible to subjective judgment due to the large differences in the characteristics of mining complex within individuals and small differences between individuals. CondInst-VoV and BlendMask-VoV, based on VoVNet-v2, are two improved instance segmentation models proposed to improve the efficiency of mine remote sensing pre-survey and minimize labor expenses. In Hubei Province, China, Gaofen satellite fusion images, true-color satellite images, false-color satellite images, and Tianditu images are gathered to create a Key Open-pit Mine Acquisition Areas (KOMMA) dataset to assess the efficacy of mine detection models. In addition, regional detection was carried out in Daye Town. The result shows that the performance of improved models on the KOMMA dataset exceeds the baseline as well as the verification accuracy of manual interpretation in regional mine detection tasks. In addition, CondInst-VoV has the best performance on Tianditu image, reaching 88.816% in positioning recall and 98.038% in segmentation accuracy. [ABSTRACT FROM AUTHOR]

Pine wood nematode (PWN), Bursaphelenchus xyophilus, originating from North America, has caused great ecological and economic hazards to pine trees worldwide, especially affecting the coniferous forests and mixed forests of masson pine in subtropical regions of China. In order to prevent PWN disease expansion, the risk level and susceptivity of PWN outbreaks need to be predicted in advance. For this purpose, we established a prediction model to estimate the susceptibility and risk level of PWN with vegetation condition variables, anthropogenic activity variables, and topographic feature variables across a large-scale district. The study was conducted in Dangyang City, Hubei Province in China, which was located in a subtropical zone. Based on the location of PWN points derived from airborne imagery and ground survey in 2018, the predictor variables were conducted with remote sensing and geographical information system (GIS) data, which contained vegetation indices including normalized difference vegetation index (NDVI), normalized difference moisture index (NDMI), normalized burn ratio (NBR), and normalized red edge index (NDRE) from Sentinel-2 imagery in the previous year (2107), the distance to different level roads which indicated anthropogenic activity, topographic variables in including elevation, slope, and aspect. We compared the fitting effects of different machine learning algorithms such as random forest (RF), K-neighborhood (KNN), support vector machines (SVM), and artificial neural networks (ANN) and predicted the probability of the presence of PWN disease in the region. In addition, we classified PWN points to different risk levels based on the density distribution of PWN sites and built a PWN risk level model to predict the risk levels of PWN outbreaks in the region. The results showed that: (1) the best model for the predictive probability of PWN presence is the RF classification algorithm. For the presence prediction of the dead trees caused by PWN, the detection rate (DR) was 96.42%, the false alarm rate (FAR) was 27.65%, the false detection rate (FDR) was 4.16%, and the area under the receiver operating characteristic curve (AUC) was equal to 0.96; (2) anthropogenic activity variables had the greatest effect on PWN occurrence, while the effects of slope and aspect were relatively weak, and the maximum, minimum, and median values of remote sensing indices were more correlated with PWN occurrence; (3) modeling analysis of different risk levels of PWN outbreak indicated that high-risk level areas were the easiest to monitor and identify, while lower incidence areas were identified with relatively low accuracy. The overall accuracy of the risk level of the PWN outbreak was identified with an AUC value of 0.94. From the research findings, remote sensing data combined with GIS data can accurately predict the probability distribution of the occurrence of PWN disease. The accuracy of identification of high-risk areas is higher than other risk levels, and the results of the study may improve control of PWN disease spread. [ABSTRACT FROM AUTHOR]

Crop residue burning is the major biomass burning activity in China, strongly influencing the regional air quality and climate. As the cultivation pattern in China is rather scattered and intricate, it is a challenge to derive an accurate emission inventory for crop residue burning. In this study, we proposed a remote sensing-based method to estimate nitrogen oxide (NOx) emissions related to crop residue burning at the field level over Hubei, China. The new method considers differences in emission factors and the spatial distribution for different crop types. Fire radiative power (FRP) derived from moderate-resolution imaging spectroradiometer (MODIS) was used to quantify NOx emissions related to agricultural biomass combustion. The spatial distribution of different crops classified by multisource remote sensing data was used as an a priori constraint. We derived a new NOx emission database for Hubei from 2014 to 2016 with spatial resolution of 1 × 1 km. Significant seasonal patterns were observed from the NOx emission database. Peak NOx emission occurring in October was related to the residue burning in late autumn harvesting. Another peak was observed between January and April, which was due to the frequent burning of stubble before spring sowing. Our results were validated by comparing our emission inventory with geostationary satellite observations, previous studies, global fire emission database (GFED), NO2 vertical column densities (VCDs) from ozone monitoring instrument (OMI) satellite observations, and measurements from environmental monitoring stations. The comparisons showed NOx emission from GFED database was 47% lower than ours, while the evaluations from most of the statistical studies were significantly higher than our results. The discrepancies were likely related to the differences of methodology and data sources. The spatiotemporal variations of NOx emission in this study showed strong correlations with NO2 VCDs, which agreed well with geostationary satellite observations. A reasonable correlation between in situ NO2 observations and our results in agricultural regions demonstrated that our method is reliable. We believe that the new NOx emission database for crop residue burning derived in this study can potentially improve the understanding of pollution sources and can provide additional information for the design of pollution control measures. [ABSTRACT FROM AUTHOR]

With the continuous application of arsenic-containing chemicals, arsenic pollution in soil has become a serious problem worldwide. The detection of arsenic pollution in soil is of great significance to the protection and restoration of soil. Hyperspectral remote sensing is able to effectively monitor heavy metal pollution in soil. However, due to the possible complex nonlinear relationship between soil arsenic (As) content and the spectrum and data redundancy, an estimation model with high efficiency and accuracy is urgently needed. In response to this situation, 62 samples and 27 samples were collected in Daye and Honghu, Hubei Province, respectively. Spectral measurement and physical and chemical analysis were performed in the laboratory to obtain the As content and spectral reflectance. After the continuum removal (CR) was performed, the stable competitive adaptive reweighting sampling algorithm coupled the successive projections algorithm (sCARS-SPA) was used for characteristic band selection, which effectively solves the problem of data redundancy and collinearity. Partial least squares regression (PLSR), radial basis function neural network (RBFNN), and shuffled frog leaping algorithm optimization of the RBFNN (SFLA-RBFNN) were established in the characteristic wavelengths to predict soil As content. These results show that the sCARS-SPA-SFLA-RBFNN model has the best universality and high prediction accuracy in different land-use types, which is a scientific and effective method for estimating the soil As content. [ABSTRACT FROM AUTHOR]