Search

This paper introduces a methodology based on machine learning and remote sensing for detecting military-induced damages to agricultural lands in Ukraine using free Sentinel-2 satellite data. The most informative spectral bands (B2, B3) and vegetation indices (NDVI, GCI) were experimentally selected for recognizing damaged fields through the Random Forest classification algorithm. Additionally, an anomaly detection method based on the estimation of deviations of pixel values from the mean within each field was applied to determine local damage in the identified affected fields. The proposed methodology demonstrated high classification accuracy with an f1-score of 0.87%, producer’s accuracy of 0.89%, user’s accuracy of 0.85, and sensitivity for detecting local damage. The developed anomaly detection method allows to recognize damage visible on the 10-meter pixel of the Sentinel-2 satellite, but does not identify small craters. Cloudiness of satellite images can significantly impair the accuracy of damage detection, and the method of local damage detection can respond to non-military anomalies and requires careful selection of threshold coefficients for each field. The study conducted a comprehensive assessment of damages inflicted on Ukrainian agricultural fields during the period 2022-2023, revealing that a total of 1,544,952 hectares, equivalent to 5.72% of the total agricultural area, experienced damage. This included 509,107 ha of wheat, 114,302 ha of sunflower, 68,830 ha of maize, 4,029 ha of rapeseed, and 16,561 ha of other crops. The most affected regions were Donetsk, Zaporizhia, and Kherson oblasts. The comprehensive findings of this research provide valuable insights for monitoring the state of agriculture and formulating strategic plans for the recovery of agricultural resources amidst the ongoing military conflict.

One of the problems in dealing with optical images for large territories (more than 10,000 sq. km) is the presence of clouds and shadows that result in having missing values in data sets. In this paper, a new approach to classification of multi-temporal optical satellite imagery with missing data due to clouds and shadows is proposed. First, self-organizing Kohonen maps (SOMs) are used to restore missing pixel values in a time series of satellite imagery. SOMs are trained for each spectral band separately using nonmissing values. Missing values are restored through a special procedure that substitutes input sample's missing components with neuron's weight coefficients. After missing data restoration, a supervised classification is performed for multi-temporal satellite images. An ensemble of neural networks, in particular multilayer perceptrons (MLPs), is proposed. Ensembling of neural networks is done by the technique of average committee, i.e. to calculate the average class probability over classifiers and select the class with the highest average posterior probability for the given input sample. The proposed approach is applied for regional scale crop classification using multi temporal Landsat-8 images for the JECAM test site in Ukraine in 2013. It is shown that ensemble of MLPs provides better performance than a single neural network in terms of overall classification accuracy, kappa coefficient, and producer's and user's accuracies for separate classes. The overall accuracy more than 85% is achieved. The obtained classification map is also validated through estimated crop areas and comparison to official statistics.

Winter wheat crop yield forecasting at national, regional and local scales is an extremely important task. This paper aims at assessing the efficiency (in terms of prediction error minimization) of satellite and biophysical model based predictors assimilation into winter wheat crop yield forecasting models at different scales (region, county and field) for one of the regions in central part of Ukraine. Vegetation index NDVI, as well as different biophysical parameters (LAI and fAPAR) derived from satellite data and WOFOST crop growth model are considered as predictors of winter wheat crop yield forecasting model. Due to very short time series of reliable statistics (since 2000) we consider single factor linear regression. It is shown that biophysical parameters (fAPAR and LAI) are more preferable to be used as predictors in crop yield forecasting regression models at each scale. Correspondent models possess much better statistical properties and are more reliable than NDVI based model. The most accurate result in current study has been obtained for LAI values derived from SPOT-VGT (at 1 km resolution) on county level. At field level, a regression model based on satellite derived LAI significantly outperforms the one based on LAI simulated with WOFOST.

Abstract  This article presents the architecture of a system that is used for regional water resource quality monitoring of environmental parameters using heterogeneous data sources such as remote sensing data, model data, and data of in-situ observations. The system’s architecture and components are developed that are reusable and can be applied to solving various monitoring problems. The monitoring of the aquatic environment of the Dnieper estuary is selected as an example of such a problem. The distinctive features of this system consist of using a Grid approach to the distribution of complex computations and also the realization of computationally complicated computations on supercomputers of the SKIT family. Typical components of the monitoring system are presented, namely, those of data acquisition, data processing, modeling, and result representation. [ABSTRACT FROM AUTHOR]

Open burning is illegal in Ukraine, yet Ukraine has, on average, 300 times more fire activity per year (2001-2019) than most European countries. In 2016 and 2017, 47% of Ukraine was identified as cultivated area, with a total of 70% of land area dedicated to agricultural use. Over 57% of all active fires in Ukraine detected using space-borne Visible Infrared Imaging Radiometer Suite (VIIRS) during 2016 and 2017 were associated with pre-planting field clearing and post-harvest crop residue removal, meaning that the majority of these fires are preventable. Due to the small size and transient nature of cropland burns, satellite-based burned area (BA) estimates are often underestimated. Moreover, traditional spectral-based BA algorithms are not suitable for distinguishing burned from plowed fields, especially in the black soil regions of Ukraine. Therefore, we developed a method to estimate agricultural BA by calibrating VIIRS active fire data with exhaustively mapped cropland reference areas (42 958 fields). Our study found that cropland BA was significantly underestimated (by 30%-63%) in the widely used Moderate Resolution Imaging Spectroradiometer-based MCD64A1 BA product, and by 95%-99.9% in Ukraine's National Greenhouse Gas Inventory. Although crop residue burns are smaller and emit far less emissions than larger wildfires, reliable monitoring of crop residue burning has a number of important benefits, including (a) improving regional air quality models and the subsequent understanding of human health impacts due to the proximity of crop residue burns to urban locations, (b) ensuring an accurate representation of predominantly smaller fires in regional emission inventories, and (c) increasing awareness of often illegal managed open burning to provide improved decision-making support for policy and resource managers.

Knowledge of the spatial distribution of agricultural abandonment following the collapse of the Soviet Union is highly uncertain. To help improve this situation, we have developed a new map of arable and abandoned land for 2010 at a 10 arc-second resolution. We have fused together existing land cover and land use maps at different temporal and spatial scales for the former Soviet Union (fSU) using a training data set collected from visual interpretation of very high resolution (VHR) imagery. We have also collected an independent validation data set to assess the map accuracy. The overall accuracies of the map by region and country, i.e. Caucasus, Belarus, Kazakhstan, Republic of Moldova, Russian Federation and Ukraine, are 90±2%, 84±2%, 92±1%, 78±3%, 95±1%, 83±2%, respectively. This new product can be used for numerous applications including the modelling of biogeochemical cycles, land-use modelling, the assessment of trade-offs between ecosystem services and land-use potentials (e.g., agricultural production), among others.

In this article, the use of time series of satellite imagery to flood hazard mapping and flood risk assessment is presented. Flooded areas are extracted from satellite images for the flood-prone territory, and a maximum flood extent image for each flood event is produced. These maps are further fused to determine relative frequency of inundation (RFI). The study shows that RFI values and relative water depth exhibit the same probabilistic distribution, which is confirmed by Kolmogorov-Smirnov test. The produced RFI map can be used as a flood hazard map, especially in cases when flood modeling is complicated by lack of available data and high uncertainties. The derived RFI map is further used for flood risk assessment. Efficiency of the presented approach is demonstrated for the Katima Mulilo region (Namibia). A time series of Landsat-5/7 satellite images acquired from 1989 to 2012 is processed to derive RFI map using the presented approach. The following direct damage categories are considered in the study for flood risk assessment: dwelling units, roads, health facilities, and schools. The produced flood risk map shows that the risk is distributed uniformly all over the region. The cities and villages with the highest risk are identified. The proposed approach has minimum data requirements, and RFI maps can be generated rapidly to assist rescuers and decisionmakers in case of emergencies. On the other hand, limitations include: strong dependence on the available data sets, and limitations in simulations with extrapolated water depth values., (© 2013 Society for Risk Analysis.)