Search

Accurate crop type maps play an important role in food security due to their widespread applicability. Optical time series data (TSD) have proven to be significant for crop type mapping. However, filling in missing information due to clouds in optical imagery is always needed, which will increase the workload and the risk of error transmission, especially for imagery with high spatial resolution. The development of optical imagery with high temporal and spatial resolution and the emergence of deep learning algorithms provide solutions to this problem. Although the one-dimensional convolutional neural network (1D CNN), long short-term memory (LSTM), and gate recurrent unit (GRU) models have been used to classify crop types in previous studies, their ability to identify crop types using optical TSD with missing information needs to be further explored due to their different mechanisms for handling invalid values in TSD. In this research, we designed two groups of experiments to explore the performances and characteristics of the 1D CNN, LSTM, GRU, LSTM-CNN, and GRU-CNN models for crop type mapping using unfilled Sentinel-2 (Sentinel-2) TSD and to discover the differences between unfilled and filled Sentinel-2 TSD based on the same algorithm. A case study was conducted in Hengshui City, China, of which 70.3% is farmland. The results showed that the 1D CNN, LSTM-CNN, and GRU-CNN models achieved acceptable classification accuracies (above 85%) using unfilled TSD, even though the total missing rate of the sample values was 43.5%; these accuracies were higher and more stable than those obtained using filled TSD. Furthermore, the models recalled more samples on crop types with small parcels when using unfilled TSD. Although LSTM and GRU models did not attain accuracies as high as the other three models using unfilled TSD, their results were almost close to those with filled TSD. This research showed that crop types could be identified by deep learning features in Sentinel-2 dense time series images with missing information due to clouds or cloud shadows randomly, which avoided spending a lot of time on missing information reconstruction. [ABSTRACT FROM AUTHOR]

The relationship between land surface temperature (LST) and environmental factors is complex and nonlinear. To determine these relationships for China, this study analyzed the driving effects of air temperature, vegetation index, soil moisture, net surface radiation, precipitation, aerosols, evapotranspiration, and water vapor on LST based on remote-sensing and reanalysis data from 2003–2018, using a convergent cross-mapping method. During the study period, air temperature and net surface radiation were the dominant drivers of LST with a cross-mapping skill above 0.9. Vegetation index and evapotranspiration were the secondary drivers of LST with a cross-mapping skill that was higher than 0.5. Except for air temperature and net surface radiation, the direction and strength of the effects of the driving factors on LST were related to the climate type. The effects of air temperature and net radiation on LST diminished from north to south, indicating that LST was more sensitive to air temperature and net radiation in energy-limited regions. However, the effects of vegetation index and evapotranspiration on LST varied significantly across climate zones; that is, positive effects were mostly in non-monsoonal zones and negative effects were primarily in monsoonal zones. Our results quantified the driving role of environmental factors on LST and provided a comprehensive understanding of LST dynamics. [ABSTRACT FROM AUTHOR]

The HJ-1B satellite, which was launched on September 6, 2008, is one of the small ones placed in the constellation for disaster prediction and monitoring. HJ-1B imagery was simulated in this paper, which contains fires of various sizes and temperatures in a wide range of terrestrial biomes and climates, including RED, NIR, MIR and TIR channels. Based on the MODIS version 4 contextual algorithm and the characteristics of HJ-1B sensor, a contextual fire detection algorithm was proposed and tested using simulated HJ-1B data. It was evaluated by the probability of fire detection and false alarm as functions of fire temperature and fire area. Results indicate that when the simulated fire area is larger than 45 m² and the simulated fire temperature is larger than 800 K, the algorithm has a higher probability of detection. But if the simulated fire area is smaller than 10 m², only when the simulated fire temperature is larger than 900 K, may the fire be detected. For fire areas about 100 m², the proposed algorithm has a higher detection probability than that of the MODIS product. Finally, the omission and commission error were evaluated which are important factors to affect the performance of this algorithm. It has been demonstrated that HJ-1B satellite data are much sensitive to smaller and cooler fires than MODIS or AVHRR data and the improved capabilities of HJ-1B data will offer a fine opportunity for the fire detection. [ABSTRACT FROM AUTHOR]

• Both one-phase and two-phase trapezoids are examined in separating ET soil and ET veg. • Two typical methods determining four endmembers are integrated into two trapezoids. • Analytical deductions of differences of ET soil and ET veg fraction are made. • ET veg /ET are intercompared with those by five typical partitioning methods. Partition of land surface evapotranspiration (ET) into soil evaporation (ET soil) and vegetation transpiration (ET veg) is of great significance for scheduling agricultural irrigation, improving water-use efficiency of crop, and managing water resources. This study made a comprehensive evaluation of one-phase and two-phase surface temperature versus fractional vegetation cover trapezoids in the separation of soil evaporation from vegetation transpiration through analytical deductions and model applications, with surface temperatures at four end-members determined from the layered approach and the patch approach. The two trapezoids were tested on the MODIS data during June to September in 2012 at Daman superstation in Northwest China. The trapezoid-estimated ratios of transpiration to total ET (ET veg /ET) were intercompared with the ET veg /ET from five typical partitioning methods, namely, the stable isotope-method, the underlying water-use efficiency (uWUE) method, the transpiration estimation algorithm (TEA), the Pérez-Priego method, and the Wei method. Results showed that: 1) The two-phase trapezoid integrated with the layered approach and the one-phase trapezoid integrated with the layered approach performed the best and worst with root-mean-square errors of 52.6 W/m2 and 78.6 W/m2, respectively, when the estimated total ET was validated against the Bowen Ratio corrected eddy covariance measurements. 2) The five partitioning methods produced largely different ET veg /ET, with the highest values from the TEA method and the lowest values from the Pérez-Priego method. 3) The estimated vegetation transpiration and ET veg /ET by the two-phase trapezoid were generally higher than those by the one-phase trapezoid. 4) In the intercomparison of ET veg /ET, the layered approach agreed better with the five partitioning methods than the patch approach. 5) The two-phase trapezoid integrated with the layered approach overall produced the most consistent estimates of ET veg /ET with those from the five partitioning methods, with the lowest bias varying between -30.9% and 8.7% and root-mean-square differences varying between 16.8% and 36.8%. In summary, the two-phase trapezoid is theoretically more rational and appears to outperform the one-phase trapezoid. This study is beneficial for a better understanding of the differences, similarities, advantages and weaknesses of the one-phase and two-phase trapezoids in the partition of total ET to its soil and vegetation components. [ABSTRACT FROM AUTHOR]

Most algorithms for land surface temperature (LST) retrieval depend on acquiring prior knowledge. To overcome this drawback, we propose a novel LST retrieval method based on model-data-knowledge-driven and deep learning, called the MDK-DL method. Based on the expert knowledge and radiation transfer model, we deduce LST retrieval mechanism and determine the best combination of the thermal infrared (TIR) bands of the sensor. Then, we use the radiation transfer model simulation and reliable satellite-ground data to establish a training and test database, and finally use the deep learning neural network for optimal computation. Three typical high-, medium- and low-spatial-resolution TIR remote sensing datasets (from Gaofen, the Moderate Resolution Imaging Spectroradiometer (MODIS), and Fengyun) are used for theoretical simulation and application analysis. The simulation shows that the minimum mean absolute error (MAE) is less than 0.1 K (standard deviation: 0.04 K; correlation coefficient: 1.000) at a small viewing direction (<7.5°) and less than 0.8 K at a large viewing direction (<65°). The in situ validation shows that the minimum MAE obtained by the optimal band combination is approximately 1 K (root mean square error (RMSE) = 1.12 K; coefficient of determination (R2) = 0.902). The retrieval accuracy is improved by increasing the number of TIR bands in the atmospheric window, and adding accurate atmospheric water vapor information produces better results. In general, four TIR bands in the atmospheric window bands are sufficient to retrieve the LST with high accuracy. Likewise, three TIR bands plus atmospheric water vapor information are sufficient for the retrieval requirements. All analyses indicate that our method is feasible and reliably accurate and can also be used to help design the instrument band to retrieve the LST with high precision. [Display omitted] • A novel LST retrieval method is proposed based on model-data-knowledge-driven and DL. • Abduction is introduced from artificial intelligence to retrieve LST. • The DL-NN algorithm make good use of the relationship among geophysical parameters. • DL-NN can be used to solve and optimize complex LST retrieval calculations. [ABSTRACT FROM AUTHOR]

Timely and accurate estimation of the winter wheat planting area and its spatial distribution is essential for the implementation of crop growth monitoring and yield estimation, and hence for the development of national agricultural production and food security. In remotely sensed winter wheat mapping based on spectral similarity, the reference curve is obtained by averaging multiple standard curves, which limits mapping accuracy. We propose a spectral reconstruction method based on singular value decomposition (SR-SVD) for winter wheat mapping based on the unique growth characteristics of crops. Using Sentinel-2 A/B satellite data, we tested the SR-SVD method in Puyang County, and Shenzhou City, China. Performance was increased, with the optimal overall accuracy and the Kappa of Puyang County and Shenzhou City were 99.52% and 0.99, and 98.26% and 0.97, respectively. We selected the spectral angle mapper (SAM) and Euclidean Distance (ED) as the similarity measures. Compared to spectral similarity methods, the SR-SVD method significantly improves mapping accuracy, as it avoids excessive extraction, can identify more detailed information, and is advantageous in distinguishing non-winter wheat pixels. Three commonly used supervised classification methods, support vector machine (SVM), maximum likelihood (ML), and minimum distance (MD) were used for comparison. Results indicate that SR-SVD has the highest mapping accuracy and greatly reduces the number of misidentified pixels. Therefore, the SR-SVD method can achieve high-precision crop mapping and provide technical support for monitoring regional crop planting structure information. [ABSTRACT FROM AUTHOR]