Mapping Smallholder Yield Heterogeneity at Multiple Scales in Eastern Africa

Predicting wheat yield from 2001 to 2020 in Hebei Province at county and pixel levels based on synthesized time series images of Landsat and MODIS

To obtain seasonable and precise crop yield information with fine resolution is very important for ensuring the food security. However, the quantity and quality of available images and the selection of prediction variables often limit the performance of yield prediction. In our study, the synthesized images of Landsat and MODIS were used to provide remote sensing (RS) variables, which can fill the missing values of Landsat images well and cover the study area completely. The deep learning (DL) was used to combine different vegetation index (VI) with climate data to build wheat yield prediction model in Hebei Province (HB). The results showed that kernel NDVI (kNDVI) and near-infrared reflectance (NIRv) slightly outperform normalized difference vegetation index (NDVI) in yield prediction. And the regression algorithm had a more prominent effect on yield prediction, while the yield prediction model using Long Short-Term Memory (LSTM) outperformed the yield prediction model using Light Gradient Boosting Machine (LGBM). The model combining LSTM algorithm and NIRv had the best prediction effect and relatively stable performance in single year. The optimal model was then used to generate 30 m resolution wheat yield maps in the past 20 years, with higher overall accuracy. In addition, we can define the optimum prediction time at April, which can consider simultaneously the performance and lead time. In general, we expect that this prediction model can provide important information to understand and ensure food security.

Maize yield prediction and condition monitoring at the sub-county scale in Kenya: synthesis of remote sensing information and crop modeling

Agricultural production assessments are crucial for formulating strategies for closing yield gaps and enhancing production efficiencies. While in situ crop yield measurements can provide valuable and accurate information, such approaches are costly and lack scalability for large-scale assessments. Therefore, crop modeling and remote sensing (RS) technologies are essential for assessing crop conditions and predicting yields at larger scales. In this study, we combined RS and a crop growth model to assess phenology, evapotranspiration (ET), and yield dynamics at grid and sub-county scales in Kenya. We synthesized RS information from the Food and Agriculture Organization (FAO) Water Productivity Open-access portal (WaPOR) to retrieve sowing date information for driving the model simulations. The findings showed that grid-scale management information and progressive crop growth could be accurately derived, reducing the model output uncertainties. Performance assessment of the modeled phenology yielded satisfactory accuracies at the sub-county scale during two representative seasons. The agreement between the simulated ET and yield was improved with the combined RS-crop model approach relative to the crop model only, demonstrating the value of additional large-scale RS information. The proposed approach supports crop yield estimation in data-scarce environments and provides valuable insights for agricultural resource management enabling countermeasures, especially when shortages are perceived in advance, thus enhancing agricultural production.

Exploring the use of Sentinel-2 datasets and environmental variables to model wheat crop yield in smallholder arid and semi-arid farming systems

Low levels of agricultural productivity are associated with the persistence of food insecurity, poverty, and other socio-economic stresses. Mapping and monitoring agricultural dynamics and production in real-time at high spatial resolution are essential for ensuring food security and shaping policy interventions. However, an accurate yield estimation might be challenging in some arid and semi-arid regions since input datasets are generally scarce, and access is restricted due to security challenges. This work examines how well Sentinel-2 satellite sensor-derived data, topographic and climatic variables, can be used as covariates to accurately model and predict wheat crop yield at the farm level using statistical models in low data settings of arid and semi-arid regions, using Sulaimani governorate in Iraq as an example. We developed a covariate selection procedure that assessed the correlations between the covariates and their relationships with wheat crop yield. Potential non-linear relationships were investigated in the latter case using regression splines. In the absence of substantial non-linear relationships between the covariates and crop yield, and residual spatial autocorrelation, we fitted a Bayesian multiple linear regression model to model and predict crop yield at 10 m resolution. Out of the covariates tested, our results showed significant relationships between crop yield and mean cumulative NDVI during the growing season, mean elevation, mean end of the season, mean maximum temperature and mean the start of the season at the farm level. For in-sample prediction, we estimated an R² value of 51 % for the model, whereas for out-of-sample prediction, this was 41 %, both of which indicate reasonable predictive performance. The calculated root-mean-square error for out-of-sample prediction was 69.80, which is less than the standard deviation of 89.23 for crop yield, further showing that the model performed well by reducing prediction variability. Besides crop yield estimates, the model produced uncertainty metrics at 10 m resolution. Overall, this study showed that Sentinel-2 data can be valuable for upscaling field measurement of crop yield in arid and semi-arid regions. In addition, the environmental covariates can strengthen the model predictive power. The method may be applicable in other areas with similar environments, particularly in conflict zones, to increase the availability of agricultural statistics.

Monitoring impacts of soil bund on spatial variation of teff and finger millet yield with Sentinel-2 and spectroradiometric data in Ethiopia

Data from remote sensing devices are essential for monitoring environmental protection practices and estimating crop yields. However, yield estimates in Ethiopia are based on time-consuming surveys. We used Sentinel-2, spectroradiometeric, and ground-truthing data to estimate the grain yield (GY) of two major crops, teff, and finger millet, in Ethiopia's Aba Gerima catchment in 2020 and 2021. At the flowering stage, we performed supervised classification on October Sentinel-2 images and spectral reflectance measurement. We used regression models to identify and predict crop yields, as evaluated by the coefficient of determination (adjusted R²) and root mean square error (RMSE). The enhanced vegetation index (EVI) and normalized-difference vegetation index (NDVI) provided the best fit to the data among the vegetation indices used to predict teff and finger millet GY. Soil bund construction increased the majority of vegetation indices and GY of both crops. We discovered a strong correlation between GY and the satellite EVI and NDVI. However, NDVI and EVI had the greatest influence on teff GY (adjusted R² = 0.83; RMSE = 0.14 ton/ha), while NDVI had the greatest influence on finger millet GY (adjusted R² = 0.85; RMSE = 0.24 ton/ha). Teff GY ranged from 0.64 to 2.16 ton/ha for bunded plots and 0.60 to 1.85 ton/ha for non-bunded plots using Sentinel-2 data. Besides, finger millet GY ranged from 1.92 to 2.57 ton/ha for bunded plots and 1.81 to 2.38 ton/ha for non-bunded plots using spectroradiometric data. Our findings show that Sentinel-2- and spectroradiometeric-based monitoring can help farmers manage teff and finger millet to achieve higher yields, more sustainable food production, and better environmental quality in the area. The study's findings revealed a link between VIs and soil management practices in soil ecological systems. Model extrapolation to other areas will necessitate local validation.

Maize Yield Estimation in Intercropped Smallholder Fields Using Satellite Data in Southern Malawi

Satellite data provide high potential for estimating crop yield, which is crucial to understanding determinants of yield gaps and therefore improving food production, particularly in sub-Saharan Africa (SSA) regions. However, accurate assessment of crop yield and its spatial variation is challenging in SSA because of small field sizes, widespread intercropping practices, and inadequate field observations. This study aimed to firstly evaluate the potential of satellite data in estimating maize yield in intercropped smallholder fields and secondly assess how factors such as satellite data spatial and temporal resolution, within-field variability, field size, harvest index and intercropping practices affect model performance. Having collected in situ data (field size, yield, intercrops occurrence, harvest index, and leaf area index), statistical models were developed to predict yield from multisource satellite data (i.e., Sentinel-2 and PlanetScope). Model accuracy and residuals were assessed against the above factors. Among 150 investigated fields, our study found that nearly half were intercropped with legumes, with an average plot size of 0.17 ha. Despite mixed pixels resulting from intercrops, the model based on the Sentinel-2 red-edge vegetation index (VI) could estimate maize yield with moderate accuracy (R2 = 0.51, nRMSE = 19.95%), while higher spatial resolution satellite data (e.g., PlanetScope 3 m) only showed a marginal improvement in performance (R2 = 0.52, nRMSE = 19.95%). Seasonal peak VI values provided better accuracy than seasonal mean/median VI, suggesting peak VI values may capture the signal of the dominant upper maize foliage layer and may be less impacted by understory intercrop effects. Still, intercropping practice reduces model accuracy, as the model residuals are lower in fields with pure maize (1 t/ha) compared to intercropped fields (1.3 t/ha). This study provides a reference for operational maize yield estimation in intercropped smallholder fields, using free satellite data in Southern Malawi. It also highlights the difficulties of estimating yield in intercropped fields using satellite imagery, and stresses the importance of sufficient satellite observations for monitoring intercropping practices in SSA.

Cropping Patterns of Annual Crops: A Remote Sensing Review

Cropping patterns are defined as the sequence and spatial arrangement of annual crops on a piece of land. Knowledge of cropping patterns is crucial for crop production and land-use intensity. While cropping patterns are related to crop production and land use intensity, they are rarely reported in agricultural statistics, especially those relating to small farms in developing countries. Remote sensing has enabled mapping cropping patterns by monitoring crops’ spatial and temporal dynamics. In this paper, we reviewed remote sensing studies of single, sequential and intercropping patterns of annual crops practiced at local and regional scales. A total of 90 studies were selected from 753 publications based on their cropping pattern types and relevance to the scope of this review. The review found that despite the increase in single cropping pattern studies due to the Sentinel missions, studies on intercropping patterns are rare, suggesting that mapping intercropping is still challenging. More so, microwave remote sensing for mapping intercropping has not been fully explored. Given the complexities in mapping intercropping, our review highlights how less frequently used vegetation indices (VIs) that benefit from red-edge and SWIR spectral bands may improve intercropping mapping.

Winter Wheat Yield Estimation Based on Optimal Weighted Vegetation Index and BHT-ARIMA Model

This study aims to use remote sensing (RS) time-series data to explore the intrinsic relationship between crop growth and yield formation at different fertility stages and construct a high-precision winter wheat yield estimation model applicable to short time-series RS data. Sentinel-2 images were acquired in this study at six key phenological stages (rejuvenation stage, rising stage, jointing stage, heading stage, filling stage, filling-maturity stage) of winter wheat growth, and various vegetation indexes (VIs) at different fertility stages were calculated. Based on the characteristics of yield data continuity, the RReliefF algorithm was introduced to filter the optimal vegetation index combinations suitable for the yield estimation of winter wheat for all fertility stages. The Absolutely Objective Improved Analytic Hierarchy Process (AOIAHP) was innovatively proposed to determine the proportional contribution of crop growth to yield formation in six different phenological stages. The selected VIs consisting of MTCI(RE2), EVI, REP, MTCI(RE1), RECI(RE1), NDVI(RE1), NDVI(RE3), NDVI(RE2), NDVI, and MSAVI were then fused with the weights of different fertility periods to obtain time-series weighted data. For the characteristics of short time length and a small number of sequences of RS time-series data in yield estimation, this study applied the multiplexed delayed embedding transformation (MDT) technique to realize the data augmentation of the original short time series. Tucker decomposition was performed on the block Hankel tensor (BHT) obtained after MDT enhancement, and the core tensor was extracted while preserving the intrinsic connection of the time-series data. Finally, the resulting multidimensional core tensor was trained with the Autoregressive Integrated Moving Average (ARIMA) model to obtain the BHT-ARIMA model for wheat yield estimation. Compared to the performance of the BHT-ARIMA model with unweighted time-series data as input, the weighted time-series input significantly improves yield estimation accuracy. The coefficients of determination (R2) were improved from 0.325 to 0.583. The root mean square error (RMSE) decreased from 492.990 to 323.637 kg/ha, the mean absolute error (MAE) dropped from 350.625 to 255.954, and the mean absolute percentage error (MAPE) decreased from 4.332% to 3.186%. Besides, BHT-ARMA and BHT-CNN models were also used to compare with BHT-ARIMA. The results indicated that the BHT-ARIMA model still had the best yield prediction accuracy. The proposed method of this study will provide fast and accurate guidance for crop yield estimation and will be of great value for the processing and application of time-series RS data.

Methodology to Differentiate Legume Species in Intercropping Agroecosystems Based on UAV with RGB Camera

Mixed crops are one of the fundamental pillars of agroecological practices. Row intercropping is one of the mixed cropping options based on the combination of two or more species to reduce their impacts. Nonetheless, from a monitoring perspective, the coexistence of different species with different characteristics complicates some processes, requiring a series of adaptations. This article presents the initial development of a procedure that differentiates between chickpea, lentil, and ervil in an intercropping agroecosystem. The images have been taken with a drone at the height of 12 and 16 m and include the three crops in the same photograph. The Vegetation Index and Soil Index are used and combined. After generating the index, aggregation techniques are used to minimize false positives and false negatives. Our results indicate that it is possible to differentiate between the three crops, with the difference between the chickpea and the other two legume species clearer than that between the lentil and the ervil in images gathered at 16 m. The accuracy of the proposed methodology is 95% for chickpea recognition, 86% for lentils, and 60% for ervil. This methodology can be adapted to be applied in other crop combinations to improve the detection of abnormal plant vigour in intercropping agroecosystems.

Understanding the Requirements for Surveys to Support Satellite-Based Crop Type Mapping: Evidence from Sub-Saharan Africa

This paper provides recommendations on how large-scale household surveys should be conducted to generate the data needed to train models for satellite-based crop type mapping in smallholder farming systems. The analysis focuses on maize cultivation in Malawi and Ethiopia, and leverages rich, georeferenced plot-level data from national household surveys that were conducted in 2018–20 and integrated with Sentinel-2 satellite imagery and complementary geospatial data. To identify the approach to survey data collection that yields optimal data for training remote sensing models, 26,250 in silico experiments are simulated within a machine learning framework. The best model is then applied to map seasonal maize cultivation from 2016 to 2019 at 10-m resolution in both countries. The analysis reveals that smallholder plots with maize cultivation can be identified with up to 75% accuracy. Collecting full plot boundaries or complete plot corner points provides the best quality of information for model training. Classification performance peaks with slightly less than 60% of the training data. Seemingly little erosion in accuracy under less preferable approaches to georeferencing plots results in the total area under maize cultivation being overestimated by 0.16–0.47 million hectares (8–24%) in Malawi.

Monitoring and Analyzing Yield Gap in Africa through Soil Attribute Best Management Using Remote Sensing Approaches: A Review

Africa has the largest population growth rate in the world and an agricultural system characterized by the predominance of smallholder farmers. Improving food security in Africa will require a good understanding of farming systems yields as well as reducing yield gaps (i.e., the difference between potential yield and actual farmer yield). To this end, crop yield gap practices in African countries need to be understood to fill this gap while decreasing the environmental impacts of agricultural systems. For instance, the variability of yields has been demonstrated to be strongly controlled by soil fertilizer use, irrigation management, soil attribute, and the climate. Consequently, the quantitative assessment and mapping information of soil attributes such as nitrogen (N), phosphorus (P), potassium (K), soil organic carbon (SOC), moisture content (MC), and soil texture (i.e., clay, sand and silt contents) on the ground are essential to potentially reducing the yield gap. However, to assess, measure, and monitor these soil yield-related parameters in the field, there is a need for rapid, accurate, and inexpensive methods. Recent advances in remote sensing technologies and high computational performances offer a unique opportunity to implement cost-effective spatiotemporal methods for estimating crop yield with important levels of scalability. However, researchers and scientists in Africa are not taking advantage of the opportunity of increasingly available geospatial remote sensing technologies and data for yield studies. The objectives of this report are to (i) conduct a review of scientific literature on the current status of African yield gap analysis research and their variation in regard to soil properties management by using remote sensing techniques; (ii) review and describe optimal yield practices in Africa; and (iii) identify gaps and limitations to higher yields in African smallholder farms and propose possible improvements. Our literature reviewed 80 publications and covered a period of 22 years (1998-2020) over many selected African countries with a potential yield improvement. Our results found that (i) the number of agriculture yield-focused remote sensing studies has gradually increased, with the largest proportion of studies published during the last 15 years; (ii) most studies were conducted exclusively using multispectral Landsat and Sentinel sensors; and (iii) over the past decade, hyperspectral imagery has contributed to a better understanding of yield gap analysis compared to multispectral imagery; (iv) soil nutrients (i.e., NPK) are not the main factor influencing the studied crop productivity in Africa, whereas clay, SOC, and soil pH were the most examined soil properties in prior papers.

Mapping Crop Types and Cropping Systems in Nigeria with Sentinel-2 Imagery

Reliable crop type maps from satellite data are an essential prerequisite for quantifying crop growth, health, and yields. However, such maps do not exist for most parts of Africa, where smallholder farming is the dominant system. Prevalent cloud cover, small farm sizes, and mixed cropping systems pose substantial challenges when creating crop type maps for sub-Saharan Africa. In this study, we provide a mapping scheme based on freely available Sentinel-2A/B (S2) time series and very high-resolution SkySat data to map the main crops—maize and potato—and intercropping systems including these two crops on the Jos Plateau, Nigeria. We analyzed the spectral-temporal behavior of mixed crop classes to improve our understanding of inter-class spectral mixing. Building on the Framework for Operational Radiometric Correction for Environmental monitoring (FORCE), we preprocessed S2 time series and derived spectral-temporal metrics from S2 spectral bands for the main temporal cropping windows. These STMs were used as input features in a hierarchical random forest classification. Our results provide the first wall-to-wall crop type map for this key agricultural region of Nigeria. Our cropland identification had an overall accuracy of 84%, while the crop type map achieved an average accuracy of 72% for the five relevant crop classes. Our crop type map shows distinctive regional variations in the distribution of crop types. Maize is the dominant crop, followed by mixed cropping systems, including maize–cereals and potato–maize cropping; potato was found to be the least prevalent class. Plot analyses based on a sample of 1166 fields revealed largely homogeneous mapping patterns, demonstrating the effectiveness of our classification system also for intercropped classes, which are temporally and spatially highly heterogeneous. Moreover, we found that small field sizes were dominant in all crop types, regardless of whether or not intercropping was used. Maize–legume and maize exhibited the largest plots, with an area of up to 3 ha and slightly more than 10 ha, respectively; potato was mainly cultivated on fields smaller than 0.5 ha and only a few plots were larger than 1 ha. Besides providing the first spatially explicit map of cropping practices in the core production area of the Jos Plateau, Nigeria, the study also offers guidance for the creation of crop type maps for smallholder-dominated systems with intercropping. Critical temporal windows for crop type differentiation will enable the creation of mapping approaches in support of future smart agricultural practices for aspects such as food security, early warning systems, policies, and extension services.

Using big data for evaluating development outcomes: A systematic map

BACKGROUND

Policy makers need access to reliable data to monitor and evaluate the progress of development outcomes and targets such as sustainable development outcomes (SDGs). However, significant data and evidence gaps remain. Lack of resources, limited capacity within governments and logistical difficulties in collecting data are some of the reasons for the data gaps. Big data-that is digitally generated, passively produced and automatically collected-offers a great potential for answering some of the data needs. Satellite and sensors, mobile phone call detail records, online transactions and search data, and social media are some of the examples of big data. Integrating big data with the traditional household surveys and administrative data can complement data availability, quality, granularity, accuracy and frequency, and help measure development outcomes temporally and spatially in a number of new ways.The study maps different sources of big data onto development outcomes (based on SDGs) to identify current evidence base, use and the gaps. The map provides a visual overview of existing and ongoing studies. This study also discusses the risks, biases and ethical challenges in using big data for measuring and evaluating development outcomes. The study is a valuable resource for evaluators, researchers, funders, policymakers and practitioners in their effort to contributing to evidence informed policy making and in achieving the SDGs.

OBJECTIVES

Identify and appraise rigorous impact evaluations (IEs), systematic reviews and the studies that have innovatively used big data to measure any development outcomes with special reference to difficult contexts.

SEARCH METHODS

A number of general and specialised data bases and reporsitories of organisations were searched using keywords related to big data by an information specialist.

SELECTION CRITERIA

The studies were selected on basis of whether they used big data sources to measure or evaluate development outcomes.

DATA COLLECTION AND ANALYSIS

Data collection was conducted using a data extraction tool and all extracted data was entered into excel and then analysed using Stata. The data analysis involved looking at trends and descriptive statistics only.

MAIN RESULTS

The search yielded over 17,000 records, which we then screened down to 437 studies which became the foundation of our systematic map. We found that overall, there is a sizable and rapidly growing number of measurement studies using big data but a much smaller number of IEs. We also see that the bulk of the big data sources are machine-generated (mostly satellites) represented in the light blue. We find that satellite data was used in over 70% of the measurement studies and in over 80% of the IEs.

AUTHORS' CONCLUSIONS

This map gives us a sense that there is a lot of work being done to develop appropriate measures using big data which could subsequently be used in IEs. Information on costs, ethics, transparency is lacking in the studies and more work is needed in this area to understand the efficacies related to the use of big data. There are a number of outcomes which are not being studied using big data, either due to the lack to applicability such as education or due to lack of awareness about the new methods and data sources. The map points to a number of gaps as well as opportunities where future researchers can conduct research.

An integrated approach of field, weather, and satellite data for monitoring maize phenology

Efficient, more accurate reporting of maize (Zea mays L.) phenology, crop condition, and progress is crucial for agronomists and policy makers. Integration of satellite imagery with machine learning models has shown great potential to improve crop classification and facilitate in-season phenological reports. However, crop phenology classification precision must be substantially improved to transform data into actionable management decisions for farmers and agronomists. An integrated approach utilizing ground truth field data for maize crop phenology (2013–2018 seasons), satellite imagery (Landsat 8), and weather data was explored with the following objectives: (i) model training and validation—identify the best combination of spectral bands, vegetation indices (VIs), weather parameters, geolocation, and ground truth data, resulting in a model with the highest accuracy across years at each season segment (step one) and (ii) model testing—post-selection model performance evaluation for each phenology class with unseen data (hold-out cross-validation) (step two). The best model performance for classifying maize phenology was documented when VIs (NDVI, EVI, GCVI, NDWI, GVMI) and vapor pressure deficit (VPD) were used as input variables. This study supports the integration of field ground truth, satellite imagery, and weather data to classify maize crop phenology, thereby facilitating foundational decision making and agricultural interventions for the different members of the agricultural chain.

Studying the Feasibility of Assimilating Sentinel-2 and PlanetScope Imagery into the SAFY Crop Model to Predict Within-Field Wheat Yield

Spatial information embedded in a crop model can improve yield prediction. Leaf area index (LAI) is a well-known crop variable often estimated from remote-sensing data and used as an input into crop models. In this study, we evaluated the assimilation of LAI derived from high-resolution (both spatial and temporal) satellite imagery into a mechanistic crop model, a simple algorithm for yield estimate (SAFY), to assess the within-field crop yield. We tested this approach on spring wheat grown in Israel. Empirical LAI models were derived from the biophysical processor for Sentinel-2 LAI and spectral vegetation indices from Sentinel-2 and PlanetScope images. The predicted grain yield obtained from the SAFY model was compared against the harvester’s yield map. LAI derived from PlanetScope and Sentinel-2 fused images achieved higher yield prediction (RMSE = 69 g/m2) accuracy than that of Sentinel-2 LAI (RMSE = 88 g/m2). Even though the spatial yield estimation was only moderately correlated to the ground truth (R2 = 0.45), this is consistent with current studies in this field, and the potential to capture within-field yield variations using high-resolution imagery has been demonstrated. Accordingly, this is the first application of PlanetScope and Sentinel-2 images conjointly used to obtain a high-density time series of LAI information to model within-field yield variability.

Implications of intra-plot heterogeneity for yield estimation accuracy: Evidence from smallholder maize systems in Ethiopia

Intra-plot heterogeneity in yield is often observed in smallholder farming systems, although its implications for yield measurement remain under-investigated. Using a unique dataset on smallholder maize production in Ethiopia, we quantify the magnitude of inter- and intra-plot heterogeneity, describe the relationship between intra-plot heterogeneity and maize productivity, and document the implications of intra-field heterogeneity on the accuracy of alternative yield estimation protocols. Our data include five common yield estimation protocols, as well as full plot harvests of 230 smallholder maize fields. We surveyed agronomic decisions, biophysical variables, and accessibility characteristics of the surveyed fields. We quantify intra-plot heterogeneity using the coefficient of variation (CV) of stand density, cob weight, and maize grain yield. A generalized linear mixed model is used to explore the relationship between these variables and the method- and heterogeneity-dependence of yield estimation accuracy. We find inter-plot CV values ranging from 32 to 56 %, 22 to 73 % and 39 to 49 % in population density, cob weight and grain yield, respectively. Intra-plot heterogeneity constituted most of this variation, with across-method mean CV values of 41 %, 82 % and 63 %, respectively, of the total variability in population density, cob weight and grain yield. A rise in intra-plot heterogeneity of 0.5 % to 0.8 % is associated with a significant increase in yield estimation error under alternative yield estimation protocols. Regression analysis shows that interactions in agronomic decisions, input intensity and plot accessibility factors dictate intra-plot heterogeneity and method accuracy in smallholder systems. Intra-plot heterogeneity is larger than inter-plot heterogeneity in the current study area. Our analysis shows that the effect of intra-plot heterogeneity on yield estimation accuracies is method-dependent and yield estimation methods that fail to capture true intra-plot heterogeneity are more error-prone. Results of such estimations should be considered with caution when used as the basis of decision-making.

Using satellite imagery to understand and promote sustainable development

Accurate and comprehensive measurements of a range of sustainable development outcomes are fundamental inputs into both research and policy. We synthesize the growing literature that uses satellite imagery to understand these outcomes, with a focus on approaches that combine imagery with machine learning. We quantify the paucity of ground data on key human-related outcomes and the growing abundance and improving resolution (spatial, temporal, and spectral) of satellite imagery. We then review recent machine learning approaches to model-building in the context of scarce and noisy training data, highlighting how this noise often leads to incorrect assessment of model performance. We quantify recent model performance across multiple sustainable development domains, discuss research and policy applications, explore constraints to future progress, and highlight research directions for the field.

Remote Crop Mapping at Scale: Using Satellite Imagery and UAV-Acquired Data as Ground Truth

Timely and accurate agricultural information is needed to inform resource allocation and sustainable practices to improve food security in the developing world. Obtaining this information through traditional surveys is time consuming and labor intensive, making it difficult to collect data at the frequency and resolution needed to accurately estimate the planted areas of key crops and their distribution during the growing season. Remote sensing technologies can be leveraged to provide consistent, cost-effective, and spatially disaggregated data at high temporal frequency. In this study, we used imagery acquired from unmanned aerial vehicles to create a high-fidelity ground-truth dataset that included examples of large mono-cropped fields, small intercropped fields, and natural vegetation. The imagery was acquired in three rounds of flights at six sites in different agro-ecological zones to capture growing conditions. This dataset was used to train and test a random forest model that was implemented in Google Earth Engine for classifying cropped land using freely available Sentinel-1 and -2 data. This model achieved an overall accuracy of 83%, and a 91% accuracy for maize specifically. The model results were compared with Rwanda’s Seasonal Agricultural Survey, which highlighted biases in the dataset including a lack of examples of mixed land cover.

Estimating Yields of Household Fields in Rural Subsistence Farming Systems to Study Food Security in Burkina Faso

Climate change has an increasing impact on food security and child nutrition, particularly among rural smallholder farmers in sub-Saharan Africa. Their limited resources and rainfall dependent farming practices make them sensitive to climate change-related effects. Data and research linking yield, human health, and nutrition are scarce but can provide a basis for adaptation and risk management strategies. In support of studies on child undernutrition in Burkina Faso, this study analyzed the potential of remote sensing-based yield estimates at household level. Multi-temporal Sentinel-2 data from the growing season 2018 were used to model yield of household fields (median 1.4 hectares (ha), min 0.01 ha, max 12.6 ha) for the five most prominent crops in the Nouna Health and Demographic Surveillance (HDSS) area in Burkina Faso. Based on monthly metrics of vegetation indices (VIs) and in-situ harvest measurements from an extensive field survey, yield prediction models for different crops of high dietary importance (millet, sorghum, maize, and beans) were successfully generated producing R² between 0.4 and 0.54 (adj. R² between 0.32 and 0.5). The models were spatially applied and resulted in a yield estimation map at household level, enabling predictions of up to 2 months prior to harvest. The map links yield on a 10-m spatial resolution to households and consequently can display potential food insecurity. The results highlight the potential for satellite imagery to provide yield predictions of smallholder fields and are discussed in the context of health-related studies such as child undernutrition and food security in rural Africa under climate change.

Deep Neural Networks and Transfer Learning for Food Crop Identification in UAV Images

Accurate projections of seasonal agricultural output are essential for improving food security. However, the collection of agricultural information through seasonal agricultural surveys is often not timely enough to inform public and private stakeholders about crop status during the growing season. Acquiring timely and accurate crop estimates can be particularly challenging in countries with predominately smallholder farms because of the large number of small plots, intense intercropping, and high diversity of crop types. In this study, we used RGB images collected from unmanned aerial vehicles (UAVs) flown in Rwanda to develop a deep learning algorithm for identifying crop types, specifically bananas, maize, and legumes, which are key strategic food crops in Rwandan agriculture. The model leverages advances in deep convolutional neural networks and transfer learning, employing the VGG16 architecture and the publicly accessible ImageNet dataset for pretraining. The developed model performs with an overall test set F1 of 0.86, with individual classes ranging from 0.49 (legumes) to 0.96 (bananas). Our findings suggest that although certain staple crops such as bananas and maize can be classified at this scale with high accuracy, crops involved in intercropping (legumes) can be difficult to identify consistently. We discuss the potential use cases for the developed model and recommend directions for future research in this area.

Prediction of Winter Wheat Yield Based on Multi-Source Data and Machine Learning in China

Wheat is one of the main crops in China, and crop yield prediction is important for regional trade and national food security. There are increasing concerns with respect to how to integrate multi-source data and employ machine learning techniques to establish a simple, timely, and accurate crop yield prediction model at an administrative unit. Many previous studies were mainly focused on the whole crop growth period through expensive manual surveys, remote sensing, or climate data. However, the effect of selecting different time window on yield prediction was still unknown. Thus, we separated the whole growth period into four time windows and assessed their corresponding predictive ability by taking the major winter wheat production regions of China as an example in the study. Firstly we developed a modeling framework to integrate climate data, remote sensing data and soil data to predict winter wheat yield based on the Google Earth Engine (GEE) platform. The results show that the models can accurately predict yield 1~2 months before the harvesting dates at the county level in China with an R2 > 0.75 and yield error less than 10%. Support vector machine (SVM), Gaussian process regression (GPR), and random forest (RF) represent the top three best methods for predicting yields among the eight typical machine learning models tested in this study. In addition, we also found that different agricultural zones and temporal training settings affect prediction accuracy. The three models perform better as more winter wheat growing season information becomes available. Our findings highlight a potentially powerful tool to predict yield using multiple-source data and machine learning in other regions and for crops.

Combining Optical, Fluorescence, Thermal Satellite, and Environmental Data to Predict County-Level Maize Yield in China Using Machine Learning Approaches

Maize is an extremely important grain crop, and the demand has increased sharply throughout the world. China contributes nearly one-fifth of the total production alone with its decreasing arable land. Timely and accurate prediction of maize yield in China is critical for ensuring global food security. Previous studies primarily used either visible or near-infrared (NIR) based vegetation indices (VIs), or climate data, or both to predict crop yield. However, other satellite data from different spectral bands have been underutilized, which contain unique information on crop growth and yield. In addition, although a joint application of multi-source data significantly improves crop yield prediction, the combinations of input variables that could achieve the best results have not been well investigated. Here we integrated optical, fluorescence, thermal satellite, and environmental data to predict county-level maize yield across four agro-ecological zones (AEZs) in China using a regression-based method (LASSO), two machine learning (ML) methods (RF and XGBoost), and deep learning (DL) network (LSTM). The results showed that combining multi-source data explained more than 75% of yield variation. Satellite data at the silking stage contributed more information than other variables, and solar-induced chlorophyll fluorescence (SIF) had an almost equivalent performance with the enhanced vegetation index (EVI) largely due to the low signal to noise ratio and coarse spatial resolution. The extremely high temperature and vapor pressure deficit during the reproductive period were the most important climate variables affecting maize production in China. Soil properties and management factors contained extra information on crop growth conditions that cannot be fully captured by satellite and climate data. We found that ML and DL approaches definitely outperformed regression-based methods, and ML had more computational efficiency and easier generalizations relative to DL. Our study is an important effort to combine multi-source remote sensed and environmental data for large-scale yield prediction. The proposed methodology provides a paradigm for other crop yield predictions and in other regions.

Cotton Yield Estimate Using Sentinel-2 Data and an Ecosystem Model over the Southern US

High-resolution data with nearly global coverage from Sentinel-2 mission open a new era for crop growth monitoring and yield estimation from remote sensing. The objective of this study is to demonstrate the potential of using Sentinel-2 biophysical data combined with an ecosystem modeling approach for estimation of cotton yield in the southern United States (US). The Boreal Ecosystems Productivity Simulator (BEPS) ecosystem model was used to simulate the cotton gross primary production (GPP) over three Sentinel-2 tiles located in Mississippi, Georgia, and Texas in 2017. Leaf area index (LAI) derived from Sentinel-2 measurements and hourly meteorological data from Modern-Era Retrospective Analysis for Research and Applications, Version 2 (MERRA-2) reanalysis were used to drive the ecosystem model. The simulated GPP values at 20-m grid spacing were aggregated to the county level (17 counties in total) and compared to the cotton lint yield estimates at the county level which are available from National Agricultural Statistics Service in the United States Department of Agriculture. The results of the comparison show that the BEPS-simulated cotton GPP explains 85% of variation in cotton yield. Our study suggests that the integration of Sentinel-2 LAI time series into the ecosystem model results in reliable estimates of cotton yield.

Real-Time Prediction of Crop Yields From MODIS Relative Vegetation Health: A Continent-Wide Analysis of Africa

Developing countries often have poor monitoring and reporting of weather and crop health, leading to slow responses to droughts and food shortages. Here, I develop satellite analysis methods and software tools to predict crop yields two to four months before the harvest. This method measures relative vegetation health based on pixel-level monthly anomalies of NDVI, EVI and NDWI indices. Because no crop mask, tuning, or subnational ground truth data are required, this method can be applied to any location, crop, or climate, making it ideal for African countries with small fields and poor ground observations. Testing began in Illinois where there is reliable county-level crop data. Correlations were computed between corn, soybean, and sorghum yields and monthly vegetation health anomalies for every county and year. A multivariate regression using every index and month (up to 1600 values) produced a correlation of 0.86 with corn, 0.74 for soybeans, and 0.65 for sorghum, all with p-values less than 10 − 6 . The high correlations in Illinois show that this model has good forecasting skill for crop yields. Next, the method was applied to every country in Africa for each country’s main crops. Crop production was then predicted for the 2018 harvest and compared to actual production values. Twenty percent of the predictions had less than 2% error, and 40% had less than 5% error. This method is unique because of its simplicity and versatility: it shows that a single user on a laptop computer can produce reasonable real-time estimates of crop yields across an entire continent.