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.

/pdf/using-satellite-imagery-to-understand-and-promote-4lb71b16ri.pdf

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 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 models' predictive performance. We quantify recent model performance across multiple sustainable development domains, discuss research and policy applications, explore constraints to future progress, and highlight key research directions for the field.

Mapping crop types and land cover in smallholder farming systems in sub-Saharan Africa remains a challenge due to data costs, high cloud cover, and poor temporal resolution of satellite data. With improvement in satellite technology and image processing techniques, there is a potential for integrating data from sensors with different spectral characteristics and temporal resolutions to effectively map crop types and land cover. In our Malawi study area, it is common that there are no cloud-free images available for the entire crop growth season. The goal of this experiment is to produce detailed crop type and land cover maps in agricultural landscapes using the Sentinel-1 (S-1) radar data, Sentinel-2 (S-2) optical data, S-2 and PlanetScope data fusion, and S-1 C2 matrix and S-1 H/α polarimetric decomposition. We evaluated the ability to combine these data to map crop types and land cover in two smallholder farming locations. The random forest algorithm, trained with crop and land cover type data collected in the field, complemented with samples digitized from Google Earth Pro and DigitalGlobe, was used for the classification experiments. The results show that the S-2 and PlanetScope fused image + S-1 covariance (C2) matrix + H/α polarimetric decomposition (an entropy-based decomposition method) fusion outperformed all other image combinations, producing higher overall accuracies (OAs) (>85%) and Kappa coefficients (>0.80). These OAs represent a 13.53% and 11.7% improvement on the Sentinel-2-only (OAs < 80%) experiment for Thimalala and Edundu, respectively. The experiment also provided accurate insights into the distribution of crop and land cover types in the area. The findings suggest that in cloud-dense and resource-poor locations, fusing high temporal resolution radar data with available optical data presents an opportunity for operational mapping of crop types and land cover to support food security and environmental management decision-making.

Crop Type and Land Cover Mapping in Northern Malawi Using the Integration of Sentinel-1, Sentinel-2, and PlanetScope Satellite Data

Paddy rice is a staple food of three billion people in the world. Timely and accurate estimation of the paddy rice planting area and paddy rice yield can provide valuable information for the government, planners and decision makers to formulate policies. This article reviews the existing paddy rice mapping methods presented in the literature since 2010, classifies these methods, and analyzes and summarizes the basic principles, advantages and disadvantages of these methods. According to the data sources used, the methods are divided into three categories: (I) Optical mapping methods based on remote sensing; (II) Mapping methods based on microwave remote sensing; and (III) Mapping methods based on the integration of optical and microwave remote sensing. We found that the optical remote sensing data sources are mainly MODIS, Landsat, and Sentinel-2, and the emergence of Sentinel-1 data has promoted research on radar mapping methods for paddy rice. Multisource data integration further enhances the accuracy of paddy rice mapping. The best methods are phenology algorithms, paddy rice mapping combined with machine learning, and multisource data integration. Innovative methods include the time series similarity method, threshold method combined with mathematical models, and object-oriented image classification. With the development of computer technology and the establishment of cloud computing platforms, opportunities are provided for obtaining large-scale high-resolution rice maps. Multisource data integration, paddy rice mapping under different planting systems and the connection with global changes are the focus of future development priorities.

Mapping Paddy Rice with Satellite Remote Sensing: A Review

Ground reference data are an essential prerequisite for supervised crop mapping. The lack of a low-cost and efficient ground referencing method results in pervasively limited reference data and hinders crop classification. In this study, we apply a convolutional neural network (CNN) model to explore the efficacy of automatic ground truthing via Google Street View (GSV) images in two distinct farming regions: Illinois and the Central Valley in California. We demonstrate the feasibility and reliability of our new ground referencing technique by performing pixel-based crop mapping at the state level using the cloud-based Google Earth Engine platform. The mapping results are evaluated using the United States Department of Agriculture (USDA) crop data layer (CDL) products. From ~ 130,000 GSV images, the CNN model identified ~ 9,400 target crop images. These images are well classified into crop types, including alfalfa, almond, corn, cotton, grape, rice, soybean, and pistachio. The overall GSV image classification accuracy is 92% for the Central Valley and 97% for Illinois. Subsequently, we shifted the image geographical coordinates 2–3 times in a certain direction to produce 31,829 crop reference points: 17,358 in Illinois, and 14,471 in the Central Valley. Evaluation of the mapping results with CDL products revealed satisfactory coherence. GSV-derived mapping results capture the general pattern of crop type distributions for 2011–2019. The overall agreement between CDL products and our mapping results is indicated by R2 values of 0.44–0.99 for the Central Valley and 0.81–0.98 for Illinois. To show the applicational value of the proposed method in other countries, we further mapped rice paddy (2014–2018) in South Korea which yielded fairly well outcomes (R2 = 0.91). These results indicate that GSV images used with a deep learning model offer an efficient and cost-effective alternative method for ground referencing, in many regions of the world.

Exploring Google Street View with deep learning for crop type mapping

High resolution satellite imagery and modern machine learning methods hold the potential to fill existing data gaps in where crops are grown around the world at a sub-field level. However, high resolution crop type maps have remained challenging to create in developing regions due to a lack of ground truth labels for model development. In this work, we explore the use of crowdsourced data, Sentinel-2 and DigitalGlobe imagery, and convolutional neural networks (CNNs) for crop type mapping in India. Plantix, a free app that uses image recognition to help farmers diagnose crop diseases, logged 9 million geolocated photos from 2017–2019 in India, 2 million of which are in the states of Andhra Pradesh and Telangana in India. Crop type labels based on farmer-submitted images were added by domain experts and deep CNNs. The resulting dataset of crop type at coordinates is high in volume, but also high in noise due to location inaccuracies, submissions from out-of-field, and labeling errors. We employed a number of steps to clean the dataset, which included training a CNN on very high resolution DigitalGlobe imagery to filter for points that are within a crop field. With this cleaned dataset, we extracted Sentinel time series at each point and trained another CNN to predict the crop type at each pixel. When evaluated on the highest quality subset of crowdsourced data, the CNN distinguishes rice, cotton, and “other” crops with 74% accuracy in a 3-way classification and outperforms a random forest trained on harmonic regression features. Furthermore, model performance remains stable when low quality points are introduced into the training set. Our results illustrate the potential of non-traditional, high-volume/high-noise datasets for crop type mapping, some improvements that neural networks can achieve over random forests, and the robustness of such methods against moderate levels of training set noise. Lastly, we caution that obstacles like the lack of good Sentinel-2 cloud mask, imperfect mobile device location accuracy, and preservation of privacy while improving data access will need to be addressed before crowdsourcing can widely and reliably be used to map crops in smallholder systems.

Rob Strey

Papers

Mapping Crop Types in Southeast India with Smartphone Crowdsourcing and Deep Learning