Multi-temporal Imagery Research Articles

Deep-Learning for Change Detection Using Multi-Modal Fusion of Remote Sensing Images: A Review

Remote sensing images provide a valuable way to observe the Earth’s surface and identify objects from a satellite or airborne perspective. Researchers can gain a more comprehensive understanding of the Earth’s surface by using a variety of heterogeneous data sources, including multispectral, hyperspectral, radar, and multitemporal imagery. This abundance of different information over a specified area offers an opportunity to significantly improve change detection tasks by merging or fusing these sources. This review explores the application of deep learning for change detection in remote sensing imagery, encompassing both homogeneous and heterogeneous scenes. It delves into publicly available datasets specifically designed for this task, analyzes selected deep learning models employed for change detection, and explores current challenges and trends in the field, concluding with a look towards potential future developments.

Spatial Analysis of Modeling Potential Flood Areas in Padang City using Google Earth Engine

Flooding in Indonesia tends to increase from year to year, and one of the causes is high rainfall. Padang City is one of the areas that has high rainfall intensity which can trigger flooding. This research aims to present spatial information on flood-prone areas by processing data in Google Earth Engine (GEE) and determine the level of flood vulnerability in Padang City. GEE is a platform capable of processing geopatial data on a large scale using multi-temporal imagery and utilizing cloud computing technology that allows solving large problems in a short time. This research uses CHIRPS Dayli data: Climate Hazard Group Infrared Precipitation, NASA SRTM Digital Elevation 30 m, Sentinel 2-A and research administrative boundaries. This study uses rainfall, elevation, TPI, NDWI and NDVI parameters. Data analysis and processing in this study were fully carried out in GEE. The results showed that the area very prone to flooding has an area of 5569.46 ha or 8.03%. The prone area is 28064.57 ha or 40.46% and the non-prone area is 35726.20 ha or 51.51%. The results of this research are expected to help the government in flood mitigation in Padang City.

Utilizing multitemporal indices and spectral bands of Sentinel-2 to enhance land use and land cover classification with random forest and support vector machine

Multitemporal imagery offers a critical advantage by capturing seasonal variations, which are essential for differentiating between land use and land cover (LULC) types. While these types may appear similar when examined at one specific time, they exhibit distinct phenological patterns across different seasons. This temporal depth is crucial for enhancing model accuracy, particularly in heterogeneous landscapes where LULC transitions are frequent and complex. This paper made use of spectral bands and indices of Sentinel-2 from April to September 2020 for LULC classification using two advanced machine learning models: Random forest (RF) and support vector machine (SVM). The spectral indices include the normalized difference vegetation index (NDVI), normalized difference built-up index (NDBI), and modified normalized water index (MNDWI). The dataset was divided into four temporal feature sets: April-May, June-July, August-September, and the entire period from April-September. For each two-month period, the median values of the spectral bands and indices were used. Both models were evaluated based on overall accuracy, F1-score, Kappa coefficient, precision, and recall. Results indicate that incorporating multitemporal features enhanced the performance of the chosen models, with overall accuracy increasing from 82.4% to 94.03% for RF and from 75.4% to 93.54% for SVM. Additionally, the RF algorithm demonstrated higher accuracy than the SVM model across various time periods, with notable improvements in other performance metrics. These improvements also underscore the ability of the models to leverage the rich multitemporal data provided by Sentinel-2 for accurate LULC classification. This study highlights the importance of incorporating the dynamics of features in remote sensing applications to enhance the precision and reliability of LULC classification.

A Decade Analysis (2013-2023) of Paddy's Yield Productivity by Using Landsat 8 Imagery in Sukoharjo District, Indonesia

Sukoharjo District has the highest rice productivity in Central Java. Sukoharjo has a strategic location. It makes this area prone to land use changes. It makes a severe impact on Paddy's productivity. This condition needs to be monitored continually. Remote sensing can provide an efficient and accurate method to solve this condition. Using the NDVI from Landsat 8 imagery and ubinan data, the model can be built to calculate and analyze paddy’s productivity. The steps were 1) interpretation of paddy fields area; 2) calculation of NDVI’s mean values a month before harvesting; 3) interpretation accuracy test; 4) correlation value between NDVI and ubinan data in 2022; 5) calculation of paddy productivity; and 6) analysis. Within a decade (2013-2023), there was a reduction in paddy's yield area for 981.90 Ha. During that period, there was an increase in paddy's productivity, around 38.2 x 103 tons. Almost all sub-districts in Sukoharjo’s yield had been reduced except for Tawangsari and Weru. Kartasura and Grogol have experienced an intensive change in paddy yields to non-paddy yields. Intensive land use changes affected paddy’s productivity. Multi-temporal imagery combined with ubinan data can be used to monitor paddy’s productivity. Forty-one points were calculated for the y (productivity) and x (NDVI value) equation. The equation resulting from this method (y=6.2212x+6.7444) can be used as a reference for calculating productivity in Sukoharjo District in different periods. From different calculations, the accuracy obtained from this method was 86%.Keywords: Landsat 8; NDVI; Paddy’s Productivity; Sukoharjo; Ubinan data

Ensemble Learning for the Land Cover Classification of Loess Hills in the Eastern Qinghai–Tibet Plateau Using GF-7 Multitemporal Imagery

The unique geographic environment, diverse ecosystems, and complex landforms of the Qinghai–Tibet Plateau make accurate land cover classification a significant challenge in plateau earth sciences. Given advancements in machine learning and satellite remote sensing technology, this study investigates whether emerging ensemble learning classifiers and submeter-level stereoscopic images can significantly improve land cover classification accuracy in the complex terrain of the Qinghai–Tibet Plateau. This study utilizes multitemporal submeter-level GF-7 stereoscopic images to evaluate the accuracy of 11 typical ensemble learning classifiers (representing bagging, boosting, stacking, and voting strategies) and 3 classification datasets (single-temporal, multitemporal, and feature-optimized datasets) for land cover classification in the loess hilly area of the Eastern Qinghai–Tibet Plateau. The results indicate that compared to traditional single strong classifiers (such as CART, SVM, and MLPC), ensemble learning classifiers can improve land cover classification accuracy by 5% to 9%. The classification accuracy differences among the 11 ensemble learning classifiers are generally within 1% to 3%, with HistGBoost, LightGBM, and AdaBoost-DT achieving a classification accuracy comparable to CNNs, with the highest overall classification accuracy (OA) exceeding 93.3%. All ensemble learning classifiers achieved better classification accuracy using multitemporal datasets, with the classification accuracy differences among the three classification datasets generally within 1% to 3%. Feature selection and feature importance evaluation show that spectral bands (e.g., the summer near-infrared (NIR-S) band), topographic factors (e.g., the digital elevation model (DEM)), and spectral indices (e.g., the summer resident ratio index (RRI-S)) significantly contribute to the accuracy of each ensemble learning classifier. Using feature-optimized datasets, ensemble classifiers can improve classification efficiency. This study preliminarily confirms that GF-7 images are suitable for land cover classification in complex terrains and that using ensemble learning classifiers and multitemporal datasets can improve classification accuracy.

Characterization and Geomorphic Change Detection of Landslides Using UAV Multi-Temporal Imagery in the Himalayas, Pakistan

Characterization and Geomorphic Change Detection of Landslides Using UAV Multi-Temporal Imagery in the Himalayas, Pakistan

From agriculture to new town: land conversion towards new-build gentrification in the southwest of Jakarta, Indonesia

ABSTRACT Located in the southwest of Jakarta, Pagedangan has been known as one of the hotspots for large-scale residential complexes or ‘new town’ in the Jakarta Metropolitan Area, Indonesia. By using a diachronic approach, we aim to examine spatial transformations of rural Pagedangan at micro-scale from 1901 to 2022 which shows the trajectory of new-build gentrification in the study area. This study thoroughly examines how land conversion transforms the rural Pagedangan into gentrifying urban landscape using historical maps and multitemporal imageries. During the last decade, Pagedangan has experienced massive agricultural land conversion which increased land price in the area. It has become gentrified seen by the rise of new town’s urban economy. Meanwhile, pre-existing irregular settlements, or kampungs, have remained as an assemblage of enclaves in a new town, a realm that completely differs from their previous rural environment. Despite displacement has not yet been the main issue in Pagedangan, kampungs are still possibly at risk of disappearance. This study delivers a comprehensive empirical study on physical gentrification process to provide insights into urban studies discourse, especially in urban Asia. This study proposes kampung as an important spatial entity which government should consider in Indonesian land use planning and urban policy making.

Channel morphological change monitoring using high-resolution LiDAR-derived DEM and multi-temporal imageries

Natural processes and human activities both cause morphological changes in channels. Remote sensing products are often used to assess planform changes, but they tend to overlook vertical changes. However, considering both planform and vertical changes is crucial for a comprehensive evaluation of morphological changes. Using spatiotemporal aerial imagery and topographic data, remote sensing plays a vital role in evaluating channel morphological changes and flood-carrying capacity. This study aimed to investigate the morphological changes of a creek in an urban catchment using very high-resolution remote sensing products. In this study, we developed a new framework for investigating overall channel morphology change by employing very high-resolution aerial imagery and a LiDAR-derived digital elevation model (DEM). By digitizing channel boundaries using ArcGIS Pro 3.0, and analyzing various morphological parameters, erosion, and deposition patterns, we examined the impact of urban expansion and infrastructure development on channel adjustments. Channel adjustments have been performed in the case study catchment (Dry Creek, South Australia, Australia) due to urban expansion and development of infrastructure in the downstream reaches. Our findings revealed a significant southwest shift in the planform of the channel, with a maximum shift of 478 m and an average shift of 217 m since 1998. This alteration resulted in an increase in the sinuosity index reaching 1.2. Over the period from 2018 to 2022, the channel experienced a net deposition depth of 3.4 cm to 3.6 cm in downstream reaches. The annual deposition volume in the downstream reaches was 1963 m3, necessitating regular desilting to prevent channel capacity loss and flooding in the surrounding environment. This study also highlights the incremental growth of riparian vegetation within the channel, which affects surface roughness, channel slope, and carrying capacity. These findings provide a valuable baseline for future investigations into stream channel morphology changes and emphasize the importance of implementing appropriate measures such as desilting and vegetation management to mitigate deposition levels, reduce flood risks, and enhance the overall health and functionality of Dry Creek. The framework used in this study can be applied to other case studies employing reliable and high-resolution remote sensing data products.

StressNet: a spatial-spectral-temporal deformable attention-based framework for water stress classification in maize

In recent years, monitoring the health of crops has been greatly aided by deploying highthroughput crop monitoring techniques that integrate remotely captured imagery and deep learning techniques. Most methods rely mainly on the visible spectrum for analyzing the abiotic stress, such as water deficiency in crops. In this study, we carry out experiments on maize crop in a controlled environment of different water treatments. We make use of a multispectral camera mounted on an Unmanned Aerial Vehicle for collecting the data from the tillering stage to the heading stage of the crop. A pre-processing pipeline, followed by the extraction of the Region of Interest from orthomosaic is explained. We propose a model based on a Convolution Neural Network, added with a deformable convolutional layer in order to learn and extract rich spatial and spectral features. These features are further fed to a weighted Attention-based Bi-Directional Long Short-Term Memory network to process the sequential dependency between temporal features. Finally, the water stress category is predicted using the aggregated Spatial-Spectral-Temporal Characteristics. The addition of multispectral, multi-temporal imagery significantly improved accuracy when compared with mono-temporal classification. By incorporating a deformable convolutional layer and Bi-Directional Long Short-Term Memory network with weighted attention, our proposed model achieved best accuracy of 91.30% with a precision of 0.8888 and a recall of 0.8857. The results indicate that multispectral, multi-temporal imagery is a valuable tool for extracting and aggregating discriminative spatial-spectral-temporal characteristics for water stress classification.

Assessing spatial variations in soil organic carbon and C:N ratio in Northeast China's black soil region: Insights from Landsat-9 satellite and crop growth information

Monitoring soil properties, especially soil organic carbon (SOC) and Carbon-to-Nitrogen (C:N) ratios, is vital for understanding degradation and developing black soil conservation strategies. Most soil mapping research has leveraged satellite imagery from limited bare-soil periods. The predictive accuracy using time-series images capturing crop growth is underexplored. Moreover, while the new Landsat-9 satellite surpasses Landsat-8, its use in soil mapping is mostly uncharted. In this study, we compared the performance of single-date bare soil imagery (Landsat-8, Landsat-9, and Sentinel-2) and multi-temporal images (Landsat-8 and Landsat-9 combined, and Sentinel-2) using three machine learning techniques (Boosted Regression Tree, BRT; Random Forest, RF; Extreme Gradient Boosting, XGBoost) for mapping SOC and C:N ratio in a typical Northeast China black soil cropland region. The results revealed that single-date Landsat-9 exhibited great potential for soil mapping with the optimal XGBoost model, improving the prediction accuracy (in terms of R2) of SOC and C:N ratio by 15.01% and 30.07%, respectively, compared to Landsat-8, while also delivering performance slightly inferior to single-date Sentinel-2. Moreover, predictors derived from multi-temporal images significantly outperformed those derived from single-date images. The XGBoost models that utilized multi-temporal Sentinel-2 predictors achieved the highest prediction accuracy for both SOC (R2 = 0.676, RMSE = 1.928 g/kg, MAE = 1.580 g/kg, RPD = 1.535) and C:N ratio (R2 = 0.713, RMSE = 0.585, MAE = 0.484, RPD = 1.718). Interestingly, the combination of Landsat-8 and Landsat-9 demonstrated similar prediction accuracy but lower uncertainties in SOC and C:N ratio mapping compared to multi-temporal Sentinel-2 images. In addition, a comparison of the contemporary prediction soil maps with historical soil data revealed a continual decrease in SOC content and C:N ratio, suggesting a concerning trajectory that is detrimental to organic matter accumulation. Overall, this study highlights the efficacy of Landsat-9 and multi-temporal imagery incorporating crop growth information in accurately predicting soil properties and assessing their spatial variability.

Agricultural Land Cover Mapping through Two Deep Learning Models in the Framework of EU’s CAP Activities Using Sentinel-2 Multitemporal Imagery

The images of the Sentinel-2 constellation can help the verification process of farmers’ declarations, providing, among other things, accurate spatial explicit maps of the agricultural land cover. The aim of the study is to design, develop, and evaluate two deep learning (DL) architectures tailored for agricultural land cover and crop type mapping. The focus is on a detailed class scheme encompassing fifteen distinct classes, utilizing Sentinel-2 imagery acquired on a monthly basis throughout the year. The study’s geographical scope covers a diverse rural area in North Greece, situated within southeast Europe. These architectures are a Temporal Convolutional Neural Network (CNN) and a combination of a Recurrent and a 2D Convolutional Neural Network (R-CNN), and their accuracy is compared to the well-established Random Forest (RF) machine learning algorithm. The comparative approach is not restricted to simply presenting the results given by classification metrics, but it also assesses the uncertainty of the classification results using an entropy measure and the spatial distribution of the classification errors. Furthermore, the issue of sampling strategy for the extraction of the training set is highlighted, targeting the efficient handling of both the imbalance of the dataset and the spectral variability of instances among classes. The two developed deep learning architectures performed equally well, presenting an overall accuracy of 90.13% (Temporal CNN) and 90.18% (R-CNN), higher than the 86.31% overall accuracy of the RF approach. Finally, the Temporal CNN method presented a lower entropy value (6.63%), compared both to R-CNN (7.76%) and RF (28.94%) methods, indicating that both DL approaches should be considered for developing operational EO processing workflows.

Land cover multiclass classification of wonosobo, Indonesia with time series-based one-dimensional deep learning model

Mapping accurate land cover is critical to support authorities in producing a better land management policy, particularly in Wonosobo, which suffers from increased erosion and sedimentation due to extensive land conversion. Unfortunately, the conventional land cover map is mainly done by manual digitation on the screen, which is highly ineffective and time-consuming. This study presents a lightweight Deep Learning (DL) model to automate land cover mapping. We used one-dimensional CNN (Convolution Neuron Network) or CNN1D alongside Bi-GRU (Gated Recurrent Unit) and FCN (Fully Connected Network) layers to process pixel-based multivariate time-series dataset extracted from satellite imagery of Landsat 8 OLI (Operational Land Imager). Sampling points were gathered from the ground survey, and visual inspection via Google Earth was used to extract pixel values from multi-temporal imageries. During the model fitting which repeated 10 times, our model delivered a consistent performance with a mean peak validation score of 0.925 and a mean training score of 0.929. In the experiments stage, we added two more band indices into the main features; however, our model performance did not experience a noticeable improvement. Compared to other well-known DL models as well as traditional ML models, our model performed better and showed strong stability in accuracy, precision, and f1-score metrics for each land cover class. These results show that our proposed model can potentially improve the efficiency of the automation of land cover mapping.

A structural catalogue of the settlement morphology in refugee and IDP camps

In the past decade, the number of refugees and internally displaced people (IDP) has doubled. This prompted the construction of more refugee camps and the proliferation of existing camps with diverse structural morphologies. Satellite imagery and machine learning (ML) are increasingly utilized to map these camps. However, there exists no standardized inventory that systemizes the built-up structures of these camps. In this study, we conceptualize the settlement morphology of refugee and IDP camps from satellite images and create a structure catalogue. Using visual image interpretation (VII) of very-high-resolution and multitemporal imagery, we compile a global database of settlement structures from 285 camps across 1,053 observations. This catalogue is subsequently used to synthesize patterns in camp structures and temporal dynamics. The results show stark variations in settlement structures across camps. Despite some similar regional patterns, stark differences in morphologies are a testament to the global heterogeneous landscape of refugee and IDP camp structures. These findings highlight the importance of considering morphological differences in image analyses across camps in future designs of ML-based automated detection and monitoring efforts. Therein, the Structure Catalogue serves as an important foundation for future earth observation for humanitarian applications.

Vegetation and Built-Up Area Monitoring in Bandung City Using Multitemporal Imagery

Bandung is West Java's largest metropolitan city and Indonesia's third largest. The city of Bandung is very strategic in various aspects, such as accessibility, communication, public facilities, and the economy. The Increased population in Bandung indicates more complex ongoing human activities, which can then affect changes in land use. The land covers in urban areas tends to change more drastically over a short period e than in rural areas because of rapid urbanization. Therefore, urban phenomenon changes are ideally monitored and detected from satellite images with a multitemporal resolution. Vegetation greenness and built-up areas can identify through multitemporal remote sensing imagery. Changes in vegetation and built-up area can monitor using remote sensing with multitemporal imagery. The analysis of changes in vegetation and built-up area studied in Bandung City represents an area with rapid population growth. This study aims to: 1) Identify changes in vegetation greenness in Bandung City between 2014 and 2021, 2) Identify built-up area changes in Bandung City between 2014 and 2021, 3) Analyze the relevance between vegetation greenness and the built-up area in Bandung City—the correlation between NDBI and NDVI through selected samples is representative of all data in Landsat 8 imagery. The proportion between the values of NDBI and NDVI samples is 0.9034. So, it is concluded that the two variables are positively correlated. Therefore, the study’s results recommend preserving vegetated land cover to conserve natural resources and prevent increased land surface temperature.Keywords: Remote Sensing Imagery, Built-Up Area, Vegetation Greenness, Bandung City

Evaluation of Multi-temporal Imagery and Vegetation Indices for Mangrove Mapping (Case Study of The East Coast of Lampung)

Mangrove ecosystem monitoring aims to provide information on a regular basis to be used as a reference in sustainable management activities in an effort to maintain the functions and services of the mangrove ecosystem. Mangroves have a very high level of threat to exploitation, namely logging and land conversion which can cause increasingly damaged conditions. This study aims to map changes in mangrove density by comparing several methods using multi-temporal data. The satellite image data used are Landsat images in 2010, 2015 and 2020. The research method used is the transformation of the vegetation indices ARVI, DVI, EVI, GNDVI, MSAVI, NDVI, RVI and TSAVI. Based on the results of the analysis carried out, the identification of mangrove density shows that the index value is in the range -1 – 18.31 with the division of density classes into 5 classes, namely very low, low, medium, high and very high. The mangrove class is in the range of values from 0.4 to 0.8 with a medium to very high density class. The results of the map accuracy test which were validated using 30 samples in the field obtained an accuracy of 84.38%. This accuracy value is quite good with the multi-temporal image used is of medium resolution.

Identifying relict periglacial features in watershed landscape and deposits of Borisoglebsk Upland, Central European Russia

Watershed landscapes of the Central Russian Plain, which were subjected to the Late Pleistocene permafrost, bear witness to widespread cryogenic structuring. It is reflected both in semi-regular spots, blocks and polygons in soil and vegetation cover on arable fields and as vertically oriented wedge casts in surficial deposits. This study aimed to identify relict periglacial features in both the landscape and sedimentary sequence of the Borisoglebsk Upland to determine if there is a correspondence of these cast networks to the distribution of landforms and regular patterns of soil and vegetation cover. Combined interpretation of aerial and satellite multi-temporal imagery and detailed investigation of vertical cross-sections in gradually expanding geotechnical trenches including apparent magnetic susceptibility measurements allowed us to reveal and spatially correlate three generations of wedge casts, each buried by well-pronounced sedimentary units, and a specific bulbous type of cast. The type of cryogenesis, erosion and slope mass-movements changed from the Late Pleniglacial to Late Glacial, representing a good proxy in the watershed sedimentary cover for palaeoenvironmental reconstructions. However, regular polygonal patterns caused by these casts are almost erased in the modern topography whereas shallow dry gullies mostly inherit pre-Holocene wider linear depressions with no clear relationship to the buried cryogenic polygons.

Research on remote sensing classification of fruit trees based on Sentinel-2 multi-temporal imageries

Accurately obtaining the spatial distribution information of fruit tree planting is of great significance to the development of fruit tree growth monitoring, disease and pest control, and yield estimation. In this study, the Sentenel-2 multispectral remote sensing imageries of different months during the growth period of the fruit trees were used as the data source, and single month vegetation indices, accumulated monthly vegetation indices (∑VIs), and difference vegetation indices between adjacent months (∆VIs) were constructed as input variables. Four conventional vegetation indices of NDVI, PSRI, GNDVI, and RVI and four improved vegetation indices of NDVIre1, NDVIre2, NDVIre3, and NDVIre4 based on the red-edge band were selected to construct a decision tree classification model combined with machine learning technology. Through the analysis of vegetation indices under different treatments and different months, combined with the attribute of Feature_importances_, the vegetation indices of different periods with high contribution were selected as input features, and the Max_depth values of the decision tree model were determined by the hyperparameter learning curve. The results have shown that when the Max_depth value of the decision tree model of the vegetation indices under the three treatments was 6, 8, and 8, the model classification was the best. The accuracy of the three vegetation index processing models on the training set were 0.8936, 0.9153, and 0.8887, and the accuracy on the test set were 0.8355, 0.7611, and 0.7940, respectively. This method could be applied to remote sensing classification of fruit trees in a large area, and could provide effective technical means for monitoring fruit tree planting areas with medium and high resolution remote sensing imageries.

Estimation of Maize Yield and Flowering Time Using Multi-Temporal UAV-Based Hyperspectral Data

Maize (Zea mays L.) is one of the most consumed grains in the world. Within the context of continuous climate change and the reduced availability of arable land, it is urgent to breed new maize varieties and screen for the desired traits, e.g., high yield and strong stress tolerance. Traditional phenotyping methods relying on manual assessment are time-consuming and prone to human errors. Recently, the application of uncrewed aerial vehicles (UAVs) has gained increasing attention in plant phenotyping due to their efficiency in data collection. Moreover, hyperspectral sensors integrated with UAVs can offer data streams with high spectral and spatial resolutions, which are valuable for estimating plant traits. In this study, we collected UAV hyperspectral imagery over a maize breeding field biweekly across the growing season, resulting in 11 data collections in total. Multiple machine learning models were developed to estimate the grain yield and flowering time of the maize breeding lines using the hyperspectral imagery. The performance of the machine learning models and the efficacy of different hyperspectral features were evaluated. The results showed that the models with the multi-temporal imagery outperformed those with imagery from single data collections, and the ridge regression using the full band reflectance achieved the best estimation accuracies, with the correlation coefficients (r) between the estimates and ground truth of 0.54 for grain yield, 0.91 for days to silking, and 0.92 for days to anthesis. In addition, we assessed the estimation performance with data acquired at different growth stages to identify the good periods for the UAV survey. The best estimation results were achieved using the data collected around the tasseling stage (VT) for the grain yield estimation and around the reproductive stages (R1 or R4) for the flowering time estimation. Our results showed that the robust phenotyping framework proposed in this study has great potential to help breeders efficiently estimate key agronomic traits at early growth stages.

Assessment of vegetation dynamics and forest loss using google earth engine and multi-temporal sentinel-2 imagery

This study evaluated regional vegetation dynamics and changes between 2015 and 2020 using Google earth engine (GEE) platform and normalized difference vegetation index (NDVI) derived from the multi-petabyte catalogue of sentinel-2 imageries. Using the computational capability of GEE, yearly mean NDVI from 2015 to 2020 were computed using level C-1 product. Subsequently, each of the NDVI images was classified into four land cover classes; water bodies, non-vegetated, grassland /cropland /shrubs, and forest using NDVI threshold values of < 0.01, 0.01-0.20, 0.20-0.30 and > 0.30, respectively. The classified maps allowed for the assessment of yearly variation in vegetation and changes between 2015 and 2020. Result showed that non-vegetated area increased from 18.53% in 2015 to 42.56% in 2020 (~ 25.00% gain), the forest area reduced to 6.78% in 2020 compared to 23.76% measured in 2015 (~ 17.00% loss in forest); whereas water bodies and grassland/cropland/shrubs remained relatively constant (0.21 and ~ 50.00%, respectively) across the years studied. Presently, the forest land was estimated to be about 2, 371.131 km2 (~ 6.70%) of the total land mass, grassland/cropland/shrubs occupied 17, 770.79 km2 (~ 50.07%), non-vegetated area was slightly less than half with 15, 274.85 km2 (~ 43.04%) and water bodies occupied 75.68 km2 (~ 0.21%).

PREDICTING VEGETATION ATTRIBUTES WITH NEURAL NETWORKS AND SENTINEL-1 & 2

Abstract. Evidence suggests that plant traits, plant functional diversity, and species diversity are linked to ecosystem functions to different extents. However, these relationships are sometimes inconsistent because of the presence of environmental gradients (e.g. climate, topography, land use) and scale mismatches between sampling units and landscape processes. Relationships between satellite data and vegetation parameters seem to be also case-specific, which hinders the creation of generalizable models. We have built predictive models of structural parameters and species composition across a broad range of climatic and topoedaphic conditions and management practices across grasslands and forests in Germany. For that, we use Sentinel multitemporal imagery and neural networks. Our models manage to explain 50% of the data variability for structural parameters, show high stability, and can generalize well across environmental gradients and sites. We also found that prediction models of biodiversity parameters show lower predictive capabilities. Spatially continuous models of grassland and forest attributes provide vital information on ecosystem functions at landscape scale. Thus, they can contribute to studying the feedback mechanisms between biodiversity, ecosystem functions, and land management at the scales to which ecological processes occur.