The full-scale invasion of Ukraine on 24 February 2022 resulted in widespread disruption to its agricultural system. As winter crops were already planted in late 2021, this led to uncertainty regarding whether all the planted fields would be harvested. Monitoring the harvest status was therefore essential for reliable production estimates. As ground-based assessments were no longer feasible in conflict-affected areas, we relied on remote sensing techniques. We developed a method to monitor crop harvest status in-season with the capability to detect fields that were not-harvested.We monitored harvest from 13 June 2022 until 19 September 2022 and found that 94.1% and 87.5% of planted winter crops were harvested in government controlled and temporarily occupied regions, respectively. The highest intensity of not-harvested fields was observed along the occupation boundary. Validation using visually interpreted high-temporal-frequency Planet imagery yielded an overall accuracy of 85%, with an F1-score of 90% for the harvested class and 73% for the not-harvested class.

<p>Agricultural monitoring has been an important topic in remote sensing research since the inception of satellite Earth observations. With early national-scale crop yield estimation efforts dating back to the LACIE and AgriSTARS projects of the 1970s and 1980s, the value and importance of food production, combined with the high variability of crop genotypes and phenotypes, have continued to spur constant innovation in mapping and monitoring agricultural land from remote sensing data.</p><p>Cropland is a highly dynamic surface type that can be difficult to map with high precision and accuracy for myriad reasons, including: variability in crop type and crop rotations; intra-season growth cycles; multi-cropping practices; crop health; fallow land; farming practices; soil fertility; seed genetics; environmental factors; and more. Each of these variables also changes through time due to climate change, advancing technology, and changes in socio-economic or political drivers. Mapping and modeling agricultural variables, therefore, require constant recalibration and validation against in-situ observations that are representative of the cropping regime.</p><p>The Essential Agricultural Variables (EAVs), defined by the GEO Global Agricultural Monitoring (GEOGLAM) initiative, provide a basis for identifying the data variables and their functional requirements in terms of spatial and temporal resolution. EAVs are designed to help the GEOGLAM community prioritize the development of products that can be derived from Earth observations data to improve downstream insight into agricultural productivity. At the same time, the EAVs are instructive for developing methods and tools for collecting the in-situ data needed to evaluate the corresponding products.</p><p>Operating under the GEOGLAM Data Lifecycle, the EAVs, and the GEO data sharing and management principles, the NASA Harvest consortium on food security and agriculture collects and distributes thousands of in-situ observations for public use in the agricultureal R&D domain. These efforts are underpinned by freely accessible data collection platforms, searchable data discovery and distribution portals, and purpose-driven field measurement methodologies that balance project-specific requirements while ensuring the future reusability of the dataset.</p><p>In this presentation, we highlight the status of current in-situ datasets and tools available through NASA Harvest. Their relevance is contextualized within the GEOGLAM EAV framework, and we discuss practical issues of in-situ data collection for agricultural remote sensing applications including farmer data privacy, reducing enumerator errors, coordinating data collection campaigns, limitations of reusing data, and balancing measurement complexity with general utility. </p>

The Russian forces invaded Ukraine on 24th February 2022 leading to widespread disruption of Ukraine's agricultural system. Ukraine is a major exporter of crops , the invasion therefore poses a significant risk to global food security. Quantifying the extent of this impact is critical, and requires monitoring of Ukraine’s agricultural lands. Total production is one of the prime indicators in this regard. Production in turn is directly proportional to the total harvested area. Harvested areas at regional scales have previously been estimated from satellite data. The majority of these studies use a complete satellite derived phenological time series and make the assumption that senescence leads to harvest. Both these conditions are not applicable in this case, as harvest estimates are required in-season and all planted fields would not necessarily be harvested due to the conflict . A delayed harvest also results in a long browning phase prior to harvest, making it particularly difficult to differentiate from post-harvest signatures. Given these constraints and challenges, we developed a method to monitor crop harvest near-real time using high resolution Planet satellite imagery. Our method includes training a model to cluster change patterns on historic data and then identify harvest patterns in the current season. Samples used to train the model consist of information from two consecutive images. Such samples are collected across the season and spatially across four agro-climatic zones, ensuring we capture a complete representation of change patterns that exist. Clusters are assigned as ‘harvested’ or ‘non-harvested’ by visually inspecting imagery at a higher temporal resolution, using which, harvest can be seen as a clear change event. On clusters which are not fully separable, we apply a hierarchical approach to further separate them. Our method works in the absence of extensive training labels and does not use predefined thresholds or assumptions. We applied the method across the harvesting period for winter crops in Ukraine. Contrary to initial reports and expectations we found a higher percentage of harvested fields in Ukraine. In free Ukraine we found 94% of planted winter crops to be harvested and in occupied Ukraine it was 88% as of 19th September 2022. Strong visual patterns of non-harvested crops were observed along the occupation borders in eastern and southern Ukraine. Harvesting trends in the north and south were largely unaffected by the conflict. With no possibility to collect ground samples, we visually interpreted satellite imagery at a higher temporal frequency to generate statistically significant validation data for model accuracy calculation. We obtained an overall accuracy of 85% with an f1-score of 90% for the harvested class and 73% for the non-harvested class. Our assessments and analysis were directed to different organizations and agencies dealing with the Ukraine crisis and led to several key insights and derived interpretations.Following NASA EarthObservatory article was published based on this work: https://earthobservatory.nasa.gov/images/150590/larger-wheat-harvest-in-ukraine-than-expected

Spatial information on cropland distribution, often called cropland or crop maps, are critical inputs for a wide range of agriculture and food security analyses and decisions. However, high-resolution cropland maps are not readily available for most countries, especially in regions dominated by smallholder farming (e.g., sub-Saharan Africa). These maps are especially critical in times of crisis when decision makers need to rapidly design and enact agriculture-related policies and mitigation strategies, including providing humanitarian assistance, dispersing targeted aid, or boosting productivity for farmers. A major challenge for developing crop maps is that many regions do not have readily accessible ground truth data on croplands necessary for training and validating predictive models, and field campaigns are not feasible for collecting labels for rapid response. We present a method for rapid mapping of croplands in regions where little to no ground data is available. We present results for this method in Togo, where we delivered a high-resolution (10 m) cropland map in under 10 days to facilitate rapid response to the COVID-19 pandemic by the Togolese government. This demonstrated a successful transition of machine learning applications research to operational rapid response in a real humanitarian crisis. All maps, data, and code are publicly available to enable future research and operational systems in data-sparse regions.

This dataset provides a 10 m resolution map of cropland in Togo (togo_cropland_2019.zip). Each pixel represents a posterior probability (ranging 0 to 1) that the pixel contains crops, predicted using an LSTM classifier and multi-spectral time series of Sentinel-2 satellite observations. For more details on the method, please see Kerner and Tseng, et al. (full reference below). This dataset also provides the hand-labeled polygons used for training (crop_merged_v2.zi, noncrop_merged_v2.zip) and testing (togo_test_majority.zip) the model, which were created by experts based on photointerpretation of high-resolution imagery (primarily SkySat and PlanetScope) in QGIS and Google Earth Pro. If you use any part of this dataset, please cite the following paper: Hannah Kerner, Gabriel Tseng, Inbal Becker-Reshef, Catherine Nakalembe, Brian Barker, Blake Munshell, Madhava Paliyam, and Mehdi Hosseini. 2020. Rapid Response Crop Maps in Data Sparse Regions. In review for KDD ’20: ACMSIGKDD Conference on Knowledge Discovery and Data Mining Workshops, August 22–27, 2020, San Diego, CA.