The ongoing war in Ukraine, starting in February 2022, has had devastating consequences beyond human lives and property damage; it has severely disrupted agricultural production, a cornerstone of the Ukrainian economy and a significant contributor to global food supplies. This disruption has far-reaching implications, particularly for middle- and low-income countries heavily reliant on Ukrainian food exports. The war has directly impacted local crop production, leading to a global food insecurity crisis and exacerbating existing inequalities worldwide. The World Food Programme (WFP) has highlighted a substantial increase in the number of people facing acute food insecurity, partially attributed to the conflict. The crisis is further compounded by existing global challenges such as the COVID-19 pandemic, climate shocks, and trade restrictions imposed in response to the war, leading to unprecedented price volatility in international agricultural markets. The damage to Ukrainian energy infrastructure further worsened food insecurity, halting grain production, export, and humanitarian aid. The ensuing food insecurity disproportionately affects vulnerable populations already struggling with hunger before the conflict. Short-term solutions have been proposed to mitigate this crisis, including expanding cropland production, bolstering humanitarian aid, and lifting export restrictions on food, feed, and fertilizers. Other suggestions include flour blending and promoting plant-based diets. However, long-term strategies remain debated. The complexities are highlighted by Ukraine's significant role in global sunflower oil, wheat, and maize markets. While short-term fixes are necessary, the lack of consensus on long-term strategies to strengthen global food systems remains a significant challenge. Addressing the food crisis hinges on accurate, quantitative assessments of agricultural losses in Ukraine due to the war. The proposed solutions, while promising, have associated costs, such as increased carbon emissions from expanding cropland or the affordability challenges for plant-based diets in lower-income countries. A comprehensive evaluation of agricultural losses is essential to inform policymakers and allow a balanced assessment of potential solutions against their costs. Such evaluations are critical for directing emergency relief efforts, ensuring that food assistance reaches those most in need effectively and proactively.

Several studies have attempted to assess grain losses due to the Russia-Ukraine war, offering valuable insights. However, these studies have limitations. Many rely on descriptive approaches or hypothetical scenarios based on assumptions about the war's trajectory, often failing to capture spatial and temporal heterogeneities. While earth observation satellites are valuable tools for monitoring cropland and agricultural activities during wartime, existing research often focuses on single crops or lacks precise species distinctions. Previous research has shown a considerable decrease in Ukrainian crop production in 2022, but the precise extent of the war's impact remained unclear. This study aims to address these knowledge gaps by providing a detailed, spatially explicit evaluation of agricultural losses caused by the war.

This research combines satellite remote sensing imagery, machine learning, and quantitative regression analysis to provide a spatially explicit evaluation of agricultural losses due to the Russia-Ukraine war. The methodology addresses three key questions: mapping major crops cost-effectively, quantifying crop production reduction and economic losses, and determining the relationship between war intensity and agricultural production loss. The study focuses on major crops in Ukraine (wheat, maize, sunflower, and rapeseed, accounting for approximately 90% of agricultural production in 2021). Publicly available satellite imagery (Sentinel-1 and 2, Google Earth, and Google Street View) was used. A semi-supervised learning approach was developed to classify these crops at a 10-m spatial resolution. The accuracy of this classification was assessed using validation sample sets. The study then focused on five high-risk eastern provinces (Crimea, Donetsk, Kherson, Luhansk, and Zaporizhzhya). Time-series satellite imagery and a threshold-based classification method were employed to assess planting and growing status, allowing the estimation of direct and indirect economic losses due to production changes. Nighttime light (NTL) data from the Visible Infrared Imaging Radiometer Suite (VIIRS) satellite served as a proxy for war intensity. The relationship between NTL changes (before and after the war) and agricultural production loss was investigated using buffer zones around urban areas. Quantitative regression analyses assessed the impact of the war, climate variables (precipitation and growing degree days), and human activity (NTL) on crop production. Two regression models were developed: a fixed-effects model to evaluate the overall impact of the war and a multiple linear regression model to analyze the attribution of changes in crop production. The models accounted for grid fixed effects, time fixed effects, and random error terms. The analysis used publicly available satellite datasets through the Google Earth Engine (GEE) platform and the global urban boundary (GUB) data from Pengcheng Laboratory.

A semi-supervised learning approach was successfully implemented to create a high-resolution (10-m) crop classification map for Ukraine in 2020, achieving an overall accuracy of 89.83% and a Kappa coefficient of 0.87. This map revealed the spatial distribution of major crops, with maize dominant in the northwest and wheat, sunflower, and rapeseed prevalent in the southeast. The study then focused on the five eastern provinces, where wheat, sunflower, and rapeseed constitute 76.8% of the cropland area. Analysis of planting conditions in these provinces indicated that about 18.11 ± 2.47% (16,956 ± 2308 km²) of cropland for these three crops remained unplanted in 2022. Kherson experienced the highest proportion of unplanted areas (24.84 ± 6.48%). Among the planted areas, significant reductions in crop production were observed. Compared to the pre-war average (2019-2021), the mean NDVI loss for planted wheat, sunflower, and rapeseed was 37.19%, 36.68%, and 36.39%, respectively. Donetsk showed the largest NDVI loss for wheat (44.66%), while Kherson and Luhansk exhibited the largest reductions for sunflower and rapeseed, respectively. Economic analysis revealed that indirect losses due to reduced production substantially exceeded direct losses from unplanted land. The estimated indirect economic losses were $520.36 ± 22.52 million (wheat), $427.59 ± 24.62 million (sunflower), and $205.02 ± 11.53 million (rapeseed), significantly higher than their respective direct losses. The analysis demonstrated a strong correlation between war intensity (measured by NTL changes) and NDVI differences, indicating that areas closer to urban centers with greater NTL reduction experienced more significant production decreases. Regression analysis confirmed that while climatic factors significantly influenced crop production, the war caused a substantial overall reduction. A one-unit reduction in NTL accounted for a 1.4% decrease in crop NDVI.

This study's findings provide a spatially detailed and quantitative assessment of the impact of the Russia-Ukraine war on agricultural production in the five high-risk eastern provinces. The combined use of satellite imagery, machine learning, and regression analysis offers a robust methodology for evaluating war-related agricultural damage. The significant finding that indirect losses substantially outweighed direct losses underscores the far-reaching consequences of the conflict beyond simply the land left unplanted. The strong correlation between war intensity (NTL changes) and crop production reduction highlights the importance of considering the wider impacts of conflict on agricultural activities. These results provide critical information for policymakers and aid organizations involved in directing emergency relief and developing long-term strategies for food security in Ukraine and globally.

This research provides a robust methodology and spatially explicit data on the significant agricultural losses resulting from the Russia-Ukraine war. The findings highlight the substantial impact of the war on crop production, with indirect losses surpassing direct losses. The strong correlation between war intensity and production decline emphasizes the need to consider the widespread consequences of conflict. Future research could extend this methodology to other conflict zones, incorporate additional factors influencing crop yield (e.g., soil quality, fertilizer usage), and explore the long-term recovery of Ukrainian agriculture.

While this study provides valuable insights, several limitations should be acknowledged. The analysis focuses on four major crops, excluding other crops due to data limitations. Using pre-war averages as a baseline may introduce potential biases due to interannual variations in crop choices. The economic loss estimates only cover the direct and indirect effects on crop production, not considering downstream effects on food processing, distribution, and market dynamics. The use of NDVI data, while common, might introduce uncertainties in yield estimation compared to other indices or ground-based measurements. Finally, the analysis captures only the first year of the war's effects; the long-term cumulative impacts might be even greater.