No abstract

Accurate prediction of future observations based on past data is the key to near real-time disturbance detection using satellite image time series (SITS). To overcome the limitations of existing methods, we present an attention-based long-short-term memory (LSTM) encoder-decoder model in which the historical time series of a pixel is encoded with a bidirectional LSTM encoder while the future time series is produced by another LSTM decoder. An attention mechanism is integrated into the encoder-decoder model to align the input time series with the output time series and to dynamically choose the most relevant contextual information while forecasting. Based on the proposed model, we develop a framework for near real-time disturbance detection and verify its effectiveness in the case of burned area mapping. The prediction accuracy of the proposed model is evaluated using moderate resolution imaging spectroradiometer (MODIS) time series and compared with state-of-the-art models. Experimental results show that our model achieves the best results in terms of lower prediction error and higher model fitness. We also evaluate the disturbance detection ability of the proposed framework. The proposed approach improves the detection rate of disturbances while suppressing false alarms, and increases the temporal accuracy. We suggest that the proposed methods provide new tools for enhancing current early warning systems in real time. Index Terms-Attention mechanism, encoder-decoder, longshort-term memory (LSTM), near real-time disturbance detection, satellite image time series (SITS). I. INTRODUCTIONT HE dramatic advances in data storage technology and high-performance computing over the past decade provide the opportunity to conduct time-series analyses with unprecedented volumes of data. Recently, near real-time disturbance Manuscript

Global navigation satellite system-reflectometry (GNSS-R) can retrieve Earth's surface parameters, such as soil moisture (SM) using the reflected signals from GNSS constellations with advantages of noncontact, all-weather, real-time, and continuity, particularly the space-borne cyclone GNSS (CYGNSS) mission. However, the accuracy and efficiency of SM estimation from CYGNSS still need to improve. In this article, the global SM is estimated using machine learning (ML) regression aided by a preclassification strategy. The total observations are classified by land types and corresponding subsets are built for constructing ML regression submodels. Ten-fold cross-validation technique is adopted. The overall performance of SM estimation with/without preclassification is compared, and the results show that the SM estimations using different ML algorithms all have substantial improvement with the preclassification strategy. Then, the optimal XGBoost predicted model with root-mean-square error (RMSE) of 0.052 cm 3 /cm 3 is adopted. In addition, the satisfactory daily and seasonal SM prediction outcomes with an overall correlation coefficient value of 0.86 and an RMSE value of 0.056 cm 3 /cm 3 are achieved at a global scale, respectively. Furthermore, the extensive temporal and spatial variations of CYGNSS SM predictions are evaluated. It shows that the reflectivity plays a main role among the predictors in SM estimation, and the next is vegetation. In some extremely dry places, the roughness may become more important. The value of SM is positively correlated with RMSE and also another limit condition that will constrain the variation of predictors, thus affecting correlation coefficient R and RMSE. Also, we compare both SMAP and CYGNSS SM predictions against in situ SM measurements from 301 stations. Similar

Deep learning has shown great potential in smart agriculture, especially in the field of pest recognition. However, existing methods require large datasets and do not exploit the semantic associations between multimodal data. To address these problems, this paper proposes a multimodal fine-grained transformer (MMFGT) model, a novel pest recognition method that improves three aspects of transformer architecture to meet the needs of few-shot pest recognition. On the one hand, the MMFGT uses self-supervised learning to extend the transformer structure to extract target features using contrastive learning to reduce the reliance on data volume. On the other hand, fine-grained recognition is integrated into the MMFGT to focus attention on finely differentiated areas of pest images to improve recognition accuracy. In addition, the MMFGT further improves the performance in pest recognition by using the joint multimodal information from the pest’s image and natural language description. Extensive experimental results demonstrate that the MMFGT obtains more competitive results compared to other excellent models, such as ResNet, ViT, SwinT, DINO, and EsViT, in pest recognition tasks, with recognition accuracy up to 98.12% and achieving 5.92% higher accuracy compared to the state-of-the-art DINO method for the baseline.