Temporal Change Information Research Articles

Tropical cyclone tracking from geostationary infrared satellite images using deep learning techniques

ABSTRACT Accurate tracking of tropical cyclones (TCs) can provide regions of interest for intelligent forecasting of TC tracks and intensity. There have been few studies on algorithms for automatic TC tracking. This study proposes an effective TC tracking method based on deep learning combined with infrared satellite images. The study first constructed a TC tracking dataset based on the infrared images of the China Fengyun-2D geostationary satellite covering six different TC intensity levels between 2009 and 2012. This included 47 complete cases (video sequences) of TCs from generation to extinction. Based on deep learning, the visual tracking algorithm SiamRPN was used as the model framework. Combining Bi-GRU and TC cloud spatiotemporal evolution characteristics to improve the performance of the SiamRPN network, the SiamTCNet target-tracking model was designed to track TCs automatically. Considering that the shape and scale of TC changes with time, a TC is regarded as a typical non-rigid object with obvious timing characteristics, so the first frame of a TC video sequence is combined with the satellite images of the first three frames of the current frame as inputs to the proposed SiamTCNet model, which then extracts the evolution of the TC’s spatial structure and its bidirectional temporal change information. The experimental results show that the TC tracking of the proposed model is a significant improvement over the original SiamRPN model.

Pixel-based classification method for earthquake-induced landslide mapping using remotely sensed imagery, geospatial data and temporal change information

Pixel-based classification method for earthquake-induced landslide mapping using remotely sensed imagery, geospatial data and temporal change information

Hybrid Global Annual 1-km IGBP Land Cover Maps for the Period 2000–2020

Land cover products provide critical information for monitoring and analyzing land surface changes. However, notable disagreement and incompatible classification systems among existing land cover products bring challenges in using them. Here, we developed a hierarchical International Geosphere-Biosphere Programme (IGBP) classification system and integrated four widely used land cover products (i.e., MODIS-IGBP, ESA-CCI, GlobeLand30, and GLC_FCS30) based on their accuracy against a collection of global reference samples. We generated a hybrid global annual land cover product (HYBMAP) with ~1 km (1/120°, 30″) spatial resolution from 2000 to 2020. The HYBMAP integrates information from the four products of high- and medium-resolution and reduces the disagreement between them by up to 20.1%. The overall accuracy of the HYBMAP is 75.5%, which is higher than the best of the four products (MODIS-IGBP, 70.9%). HYBMAP also integrates the temporal change information from the four products and identifies a faster growth of built-up lands. The HYBMAP provides more consistent and reliable global land cover time series data for global change research. It is free to access at https://doi.org/10.5281/zenodo.10488191 .

Multivariate Time Series Prediction Based on Temporal Change Information Learning Method.

In the multivariate time series prediction tasks, the impact information of all nonpredictive time series on the predictive target series is difficult to be extracted at different time stages. Through the emphasis on optimal-related sequences in the target series, the deep learning model with the attention mechanism achieves a good predictive performance. However, temporal change information in the objective function and optimization algorithm is completely ignored in these models. To this end, a temporal change information learning (CIL) method is proposed in this article. First, mean absolute error (MAE) and mean squared error (MSE) losses are contained in the objective function to evaluate different amplitude errors. Meanwhile, the second-order difference technology is used in the correlation terms of the objective function to adaptively capture the impact of the abrupt and slow change information in each series on the target series. Second, the long short-term memory (LSTM) network with the transformation mechanism is used in the method so that temporal dependence information can be fully extracted (i.e., avoiding the supersaturation region). Third, to effectively obtain the optimal model parameters, the current and historical moment estimation information is adaptively memorized without the introduction of additional hyperparameters, and therefore, the acquisition ability of temporal change information in the error gradient flow is greatly enhanced by the proposed optimization algorithm. Finally, three datasets with different scales are used to verify the advantages of the CIL method in computational overhead and prediction effect.

Spectral–Temporal Transformer for Hyperspectral Image Change Detection

Deep-Learning-based (DL-based) approaches have achieved remarkable performance in hyperspectral image (HSI) change detection (CD). Convolutional Neural Networks (CNNs) are often employed to capture fine spatial features, but they do not effectively exploit the spectral sequence information. Furthermore, existing Siamese-based networks ignore the interaction of change information during feature extraction. To address this issue, we propose a novel architecture, the Spectral–Temporal Transformer (STT), which processes the HSI CD task from a completely sequential perspective. The STT concatenates feature embeddings in spectral order, establishing a global spectrum–time-receptive field that can learn different representative features between two bands regardless of spectral or temporal distance, thereby strengthening the learning of temporal change information. Via the multi-head self-attention mechanism, the STT is capable of capturing spectral–temporal features that are weighted and enriched with discriminative sequence information, such as inter-spectral correlations, variations, and time dependency. We conducted experiments on three HSI datasets, demonstrating the competitive performance of our proposed method. Specifically, the overall accuracy of the STT outperforms the second-best method by 0.08%, 0.68%, and 0.99% on the Farmland, Hermiston, and River datasets, respectively.

I-GCN: Incremental Graph Convolution Network for Conversation Emotion Detection

Sentiment analysis and emotion detection in conversation are becoming hot topics in regard to several applications. With the development of the social robot, social network, and intelligent voice assistant, emotion detection is attracting more attention as a key component in these research fields. Many approaches have been proposed to handle this problem in recent years. However, these previous approaches focus on either the temporal change information of the conversation or the semantic correlation information of the dialogue but ignore the combination of temporal information and semantic correlation information. In this paper, we propose an incremental graph convolution network (I-GCN) to handle emotion detection in conversation. We first utilize the graph structure to represent conversation at different times, which can represent the semantic correlation information of utterances. Then, we apply the incremental graph structure to imitate the process of dynamic conversation, which can preserve the temporal change information of conversation. Especially, for the first step of the process, we creatively propose utterance-level GCN (U-GCN) and speaker-level GCN (S-GCN) to learn the features of utterances for emotion detection. U-GCN focuses on the correlations among utterances and applies the multi-head attention model to find latent correlation information among utterances, which aims to further enhance the guidance of semantic relevance for feature learning. S-GCN focuses on the correlation between speaker and utterances, which can provide a different angle to guide the feature learning of utterances. In the learning of model parameters, we constantly utilize the new utterances to fine-tune the parameters of GNN for enhancement of the contribution of temporal change information. Detailed evaluations of the proposed method on three published conversation corpuses demonstrate the great effectiveness of our approach over several conventional competitive baselines.

Joint Variation Learning of Fusion and Difference Features for Change Detection in Remote Sensing Images

Remote sensing (RS) image change detection (CD) is an earth observation technique for detecting surface changes in the same area during a period. With the rapid development of deep learning, various deep neural networks especially Siamese ones have been widely used in the field of CD. However, they have the deficiency of insufficient contextual information aggregation, resulting in false and missed detections, and it is difficult to refine the detection of change edges. To alleviate these problems and obtain more accurate results, we propose an efficient self-weighted spatial-temporal attention network (SSANet). In contrast to the Siamese structure, our network is a novel joint learning framework composed of fusion sub-network, difference sub-network, and decoder. Fusion sub-network is used to extract multiscale object features where we propose a multi-core channel-aligning attention (MCA) module to capture the long-range semantic information for multi-scale context aggregation. Difference sub-network is used to extract the difference variation features, where we propose a feature differential reconfiguration (FDR) module to learn the temporal change information. FDR can effectively filter change information and reconstruct features to improve the perception of changed regions. To better balance the MCA and FDR modules, an asymmetric weighting (AW) module is proposed in the decoder to self-weight the multi-scale features and generate the change map. Experiments demonstrate the efficiency of proposed sub-networks and modules, and the state-of-the-art performance of SSANet.

Mapping Dongting Lake Wetland Utilizing Time Series Similarity, Statistical Texture, and Superpixels With Sentinel-1 SAR Data

Wetland is of significant ecological value, which is very important and challenging for large-scale mapping. Sentinel-1 can continuously record wetland changes with its all-day, all-weather working capability. How to fully utilize multidimensional information, such as time and statistical texture of Synthetic Aperture Radar (SAR) image, to classify wetlands more accurately has become a research focus. Thus, this paper constructed Time Series Similarity parameters to describe the temporal change information of targets in wetland and introduced G <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0</sup> statistical texture parameters to describe texture characteristics of SAR images. Combining multi-features and superpixels, we proposed a classification method based on Random Forest (RF) to map Dongting Lake wetland. The overall classification accuracy was 97.57%, and the Kappa coefficient was 0.97. The classification accuracy of reed beach, grass marsh etc. was above 95%. The results showed the proposed SAR Time Series Similarity features could effectively utilize dynamical information among classes and was helpful to identify mudflat, grass marsh and reed beach with high dynamics in wetland. The introduced statistical texture features expressed the heterogeneity of targets, and enhanced the recognition and extraction ability of forest beach, mudflat and reed beach. Compared with Support Vector Machine (SVM), Decision Tree (DT) and RF classification methods, RF with superpixels optimized not only got high precision but also could effectively reduce the pepper-salt error in classification, because of the consideration of superpixels context information.

Temporal Information Extraction for Afforestation in the Middle Section of the Yarlung Zangbo River Using Time-Series Landsat Images Based on Google Earth Engine

Afforestation is one of the most efficient ways to control land desertification in the middle section of the Yarlung Zangbo River (YZR) valley. However, the lack of a quantitative way to record the planting time of artificial forest (AF) constrains further management for these forests. The long-term archived Landsat images (including the Thematic Mapper (TM), Enhanced Thematic Mapper Plus (ETM+), and Operational Land Imager (OLI)) provide a good opportunity to capture the temporal change information about AF plantations. Under the condition that there would be an abrupt increasing trend in the normalized difference vegetation index (NDVI) time-series curve after afforestation, and this characteristic can be thought of as the indicator of the AF planting time. To extract the indicator, an algorithm based on the Google Earth Engine (GEE) for detecting this trend change point (TCP) on the maximum NDVI time series within the growing season (May to September) was proposed. In this algorithm, the time-series NDVI was initially smoothed and segmented into two subspaces. Then, a trend change indicator Sdiff was calculated with the difference between the fitting slopes of the subspaces before and after each target point. A self-adaptive method was applied to the NDVI series to find the right year with the maximum TCP, which is recorded as the AF planting time. Based on the proposed method, the AF planting time of the middle section of the YZR valley from 1988 to 2020 was derived. The detected afforestation temporal information was validated by 222 samples collected from the field survey, with a Pearson correlation coefficient of 0.93 and a root mean squared error (RMSE) of 2.95 years. Meanwhile, the area distribution of the AF planted each year has good temporal consistency with the implementation of the eco-reconstruction project. Overall, the study provides a good way to map AF planting times that is not only helpful for sustainable management of AF areas but also provides a basis for further research on the impact of afforestation on desertification control.

Mangrove Forest Landcover Changes in Coastal Vietnam: A Case Study from 1973 to 2020 in Thanh Hoa and Nghe An Provinces

Mangrove forests can ameliorate the impacts of typhoons and storms, but their extent is threatened by coastal development. The northern coast of Vietnam is especially vulnerable as typhoons frequently hit it during the monsoon season. However, temporal change information in mangrove cover distribution in this region is incomplete. Therefore, this study was undertaken to detect change in the spatial distribution of mangroves in Thanh Hoa and Nghe An provinces and identify reasons for the cover change. Landsat satellite images from 1973 to 2020 were analyzed using the NDVI method combined with visual interpretation to detect mangrove area change. Six LULC classes were categorized: mangrove forest, other forests, aquaculture, other land use, mudflat, and water. The mangrove cover in Nghe An province was estimated to be 66.5 ha in 1973 and increased to 323.0 ha in 2020. Mangrove cover in Thanh Hoa province was 366.1 ha in 1973, decreased to 61.7 ha in 1995, and rose to 791.1 ha in 2020. Aquaculture was the main reason for the loss of mangroves in both provinces. Overall, the percentage of mangrove loss from aquaculture was 42.5% for Nghe An province and 60.1% for Thanh Hoa province. Mangrove restoration efforts have contributed significantly to mangrove cover, with more than 1300 ha being planted by 2020. This study reveals that improving mangrove restoration success remains a challenge for these provinces, and further refinement of engineering techniques is needed to improve restoration outcomes.

Reconstruction of High-Temporal- and High-Spatial-Resolution Reflectance Datasets Using Difference Construction and Bayesian Unmixing

High-temporal- and high-spatial-resolution reflectance datasets play a vital role in monitoring dynamic changes at the Earth’s land surface. So far, many sensors have been designed with a trade-off between swath width and pixel size; thus, it is difficult to obtain reflectance data with both high spatial resolution and frequent coverage from a single sensor. In this study, we propose a new Reflectance Bayesian Spatiotemporal Fusion Model (Ref-BSFM) using Landsat and MODIS (Moderate Resolution Imaging Spectroradiometer) surface reflectance, which is then used to construct reflectance datasets with high spatiotemporal resolution and a long time series. By comparing this model with other popular reconstruction methods (the Flexible Spatiotemporal Data Fusion Model, the Spatial and Temporal Adaptive Reflectance Fusion Model, and the Enhanced Spatial and Temporal Adaptive Reflectance Fusion Model), we demonstrate that our approach has the following advantages: (1) higher prediction accuracy, (2) effective treatment of cloud coverage, (3) insensitivity to the time span of data acquisition, (4) capture of temporal change information, and (5) higher retention of spatial details and inconspicuous MODIS patches. Reflectance time-series datasets generated by Ref-BSFM can be used to calculate a variety of remote-sensing-based vegetation indices, providing an important data source for land surface dynamic monitoring.

A Cross-Resolution, Spatiotemporal Geostatistical Fusion Model for Combining Satellite Image Time-Series of Different Spatial and Temporal Resolutions

Dense time-series with coarse spatial resolution (DTCS) and sparse time-series with fine spatial resolution (STFS) data often provide complementary information. To make full use of this complementarity, this paper presents a novel spatiotemporal fusion model, the spatial time-series geostatistical deconvolution/fusion model (STGDFM), to generate synthesized dense time-series with fine spatial resolution (DTFS) data. Attributes from the DTCS and STFS data are decomposed into trend and residual components, and the spatiotemporal distributions of these components are predicted through novel schemes. The novelty of STGDFM lies in its ability to (1) consider temporal trend information using land-cover-specific temporal profiles from an entire DTCS dataset, (2) reflect local details of the STFS data using resolution matrix representation, and (3) use residual correction to account for temporary variations or abrupt changes that cannot be modeled from the trend components. The potential of STGDFM is evaluated by conducting extensive experiments that focus on different environments; spatially degraded datasets and real Moderate Resolution Imaging Spectroradiometer (MODIS) and Landsat images are employed. The prediction performance of STGDFM is compared with those of a spatial and temporal adaptive reflectance fusion model (STARFM) and an enhanced STARFM (ESTARFM). Experimental results indicate that STGDFM delivers the best prediction performance with respect to prediction errors and preservation of spatial structures as it captures temporal change information on the prediction date. The superiority of STGDFM is significant when the difference between pair dates and prediction dates increases. These results indicate that STGDFM can be effectively applied to predict DTFS data that are essential for various environmental monitoring tasks.

Change Analysis in Urban Areas Based on Statistical Features and Temporal Clustering Using TerraSAR-X Time-Series Images

The existing unsupervised multitemporal change detection approaches for synthetic aperture radar (SAR) images based on the pixel level usually suffer from the serious influence of speckle noise, and the classification accuracy of temporal change patterns is liable to be affected by the generation method of similarity matrices and the pre-specified cluster number. To address these issues, a novel time-series change detection method with high efficiency is proposed in this paper. Firstly, spatial feature extraction using local statistical information on patches is conducted to reduce the noise and for subsequent temporal grouping. Secondly, a density-based clustering method is adopted to categorize the pixel series in the temporal dimension, in view of its efficiency and robustness. Change detection and classification results are then obtained by a fast differential strategy in the final step. The experimental results and analysis of synthetic and realistic time-series SAR images acquired by TerraSAR-X in urban areas demonstrate the effectiveness of the proposed method, which outperforms other approaches in terms of both qualitative results and quantitative indices of macro F1-scores and micro F1-scores. Furthermore, we make the case that more temporal change information for buildings can be obtained, which includes when the first and last detected change occurred and the frequency of changes.

Identification and mapping of winter wheat by integrating temporal change information and Kullback–Leibler divergence

Crop acreage and its spatial distribution are a base for agriculture related works. Current research combining medium and low spatial resolution images focuses on data fusion and unmixing methods. The purpose of the former is to generate synthetic fine spatial resolution data instead of directly solving the problem. In the latter, high-resolution data is only used to provide endmembers and the result is usually an abundance map rather than the true spatial distribution data. To solve this problem, this paper designs a conceptual model which divides the study area into different types of pixels at a MODIS 250 m scale. Only three types of pixels contain winter wheat, i.e., pure winter wheat pixels (PA), the mixed pixels comprising winter wheat and other vegetation (MA) and the mixed pixels comprising winter wheat and other crops (MB). Different strategies are used in processing them. (1) Within the pure cultivated land pixels, the Kullback–Leibler (KL) divergence is employed to analyze the similarity between unknown pixels and the pure winter wheat samples on the temporal change characteristics of NDVI. Further PA is identified. (2) For MA, a proposed reverse unmixing method is firstly used to extract the temporal change information of cultivated land components, after which winter wheat is identified from the cultivated land components as previously described. (3) For MB which only appears on the border of PA, a mask is created by expanding the PA and temporal difference is utilized to identify winter wheat under the mask. Finally, these three results are integrated at a TM scale with the aid of 25 m resolution land use data. We applied the proposed solution and obtained a good result in the main agricultural area of the Yiluo River Basin. The identified winter wheat planting acreage is 161,050.00 hm2. The result is validated based on the five-hundred random validation points. Overall accuracy is 94.80% and Kappa coefficient is 0.85. This demonstrates that the temporal information reflecting crop growth is also an important indicator, and the KL divergence makes it more convenient in identifying winter wheat. This research provided a new perspective for the combination of low and medium spatial resolution remote sensing images. The proposed solution can also be effectively applied in other places and countries for the crop which has a clear temporal change characteristic that is different from others.

Social vulnerability change assessment: monitoring longitudinal demographic indicators of disaster risk in Germany from 2005 to 2015

Social vulnerability assessments of disaster risks related to natural or man-made hazards have become state of the art, while not undisputed. One critique—that most assessments are static—is addressed here by analysing indicators over a 10-year period at a research area in Germany using county-level administrative units. The indicators (and an index) are displayed as maps using publicly available statistical data from a single government source to establish consistency. In addition, maps of changes in indicator values per 5 or 10 years are visualised. As a result, spatial regions within Germany that have received increases in, for instance, higher numbers of elderly citizens can be visualised. Maps visualise the increasing susceptibility to hazards such as floods, heatwaves, storms or technological accidents. At the same time, increases or decreases in capacities, such as hospital density or care homes, highlight regions that have adjusted or still have to adjust to demographic change demands. The main purpose of this paper is to test stability and heterogeneity of indicators over time and offer advice on a selection of indicators that are not only based on static and singular assessments but also take dynamics and related richness of additional temporal change information into account for longitudinal monitoring of disaster risk with a focus on vulnerability.

Environmental monitoring and assessment of landscape dynamics in southern coast of the Caspian Sea through intensity analysis and imprecise land-use data.

A hierarchical intensity analysis of land-use change is applied to evaluate the dynamics of a coupled urban coastal system in Rasht County, Iran. Temporal land-use layers of 1987, 1999, and 2011 are employed, while spatial accuracy metrics are only available for 2011 data (overall accuracy of 94%). The errors in 1987 and 1999 layers are unknown, which can influence the accuracy of temporal change information. Such data were employed to examine the size and the type of errors that could justify deviations from uniform change intensities. Accordingly, errors comprising 3.31 and 7.47% of 1999 and 2011 maps, respectively, could explain all differences from uniform gains and errors including 5.21 and 1.81% of 1987 and 1999 maps, respectively, could explain all deviations from uniform losses. Additional historical information is also applied for uncertainty assessment and to separate probable map errors from actual land-use changes. In this regard, historical processes in Rasht County can explain different types of transition that are either consistent or inconsistent to known processes. The intensity analysis assisted in identification of systematic transitions and detection of competitive categories, which cannot be investigated through conventional change detection methods. Based on results, built-up area is the most active gaining category in the area and wetland category with less areal extent is more sensitive to intense land-use change processes. Uncertainty assessment results also indicated that there are no considerable classification errors in temporal land-use data and these imprecise layers can reliably provide implications for informed decision making.

Computer aided mass detection in mammography with temporal change analysis

This paper presents a method to extract change information from temporal mammogram pairs and to incorporate the temporal change information in the malignant mass classification. In this method, a temporal mammogram registration framework which is based on spatial relations between regions of interest and graph matching was used to create correspondences between regions of current mammogram and regions of previous mammogram. 18 image features were then used to capture the differences (temporal changes) between the matched regions. To assess the contribution of temporal change information to the mass detection, 5 methods were designed to combine mass classification on image features measured on single regions and mass classification on temporal features to improve overall mass classification. The method was tested on 95 pairs of temporal mammograms using k-fold cross validation procedure. The experimental results showed that, when combining two classification results using linear combination or by taking minimum value, the Az score of overall classification performance increased from 0.8843 to 0.8989 and 0.8863 respectively. The results demonstrated that registering temporal mammograms, measuring temporal changes from matched regions and incorporating the change information in the mass classification improves the overall mass detection.

Use of MODIS and Landsat time series data to generate high-resolution temporal synthetic Landsat data using a spatial and temporal reflectance fusion model

Due to cloud coverage and obstruction, it is difficult to obtain useful images during the critical periods of monitoring vegetation using medium-resolution spatial satellites such as Landsat and Satellite Pour l'Observation de la Terre (SPOT), especially in pluvial regions. Although high temporal resolution sensors, such as the Advanced Very High Resolution Radiometer (AVHRR) and Moderate Resolution Imaging Spectroradiometer (MODIS), can provide high-frequency data, the coarse ground resolutions of these sensors make them unsuitable to quantify the vegetation growth processes at fine scales. This paper introduces a new data fusion model for blending observations of high temporal resolution sensors (e.g., MODIS) and moderate spatial resolution satellites (e.g., Landsat) to produce synthetic imagery with both high-spatial and temporal resolutions. By detecting temporal change information from MODIS daily surface reflectance images, our algorithm produced high-resolution temporal synthetic Landsat data based on a Landsat-7 Enhanced Thematic Mapper Plus (ETM+) image at the beginning time (T1). The algorithm was then tested over a 185×185 km2 area located in East China. The results showed that the algorithm can produce high-resolution temporal synthetic Landsat data that were similar to the actual observations with a high correlation coefficient (r) of 0.98 between synthetic imageries and the actual observations.