Automated cropland mapping of continental Africa using Google Earth Engine cloud computing

Jun Xiong; Prasad S. Thenkabail; Murali Krishna Gumma; Pardhasaradhi G. Teluguntla; Justin Poehnelt; Russell G. Congalton; Kamini Yadav; David Thau

doi:10.1016/j.isprsjprs.2017.01.019

USGS · 70192160

Automated cropland mapping of continental Africa using Google Earth Engine cloud computing

Jun Xiong·Prasad S. Thenkabail·Murali Krishna Gumma·Pardhasaradhi G. Teluguntla·Justin Poehnelt·Russell G. Congalton·Kamini Yadav·David Thau

View original source ↗Export metadata ↓

Abstract

The automation of agricultural mapping using satellite-derived remotely sensed data remains a challenge in Africa because of the heterogeneous and fragmental landscape, complex crop cycles, and limited access to local knowledge. Currently, consistent, continent-wide routine cropland mapping of Africa does not exist, with most studies focused either on certain portions of the continent or at most a one-time effort at mapping the continent at coarse resolution remote sensing. In this research, we addressed these limitations by applying an automated cropland mapping algorithm (ACMA) that captures extensive knowledge on the croplands of Africa available through: (a) ground-based training samples, (b) very high (sub-meter to five-meter) resolution imagery (VHRI), and (c) local knowledge captured during field visits and/or sourced from country reports and literature. The study used 16-day time-series of Moderate Resolution Imaging Spectroradiometer (MODIS) normalized difference vegetation index (NDVI) composited data at 250-m resolution for the entire African continent. Based on these data, the study first produced accurate reference cropland layers or RCLs (cropland extent/areas, irrigation versus rainfed, cropping intensities, crop dominance, and croplands versus cropland fallows) for the year 2014 that provided an overall accuracy of around 90% for crop extent in different agro-ecological zones (AEZs). The RCLs for the year 2014 (RCL2014) were then used in the development of the ACMA algorithm to create ACMA-derived cropland layers for 2014 (ACL2014). ACL2014 when compared pixel-by-pixel with the RCL2014 had an overall similarity greater than 95%. Based on the ACL2014, the African continent had 296 Mha of net cropland areas (260 Mha cultivated plus 36 Mha fallows) and 330 Mha of gross cropland areas. Of the 260 Mha of net cropland areas cultivated during 2014, 90.6% (236 Mha) was rainfed and just 9.4% (24 Mha) was irrigated. Africa has about 15% of the world’s population, but only about 6% of world’s irrigation. Net cropland area distribution was 95 Mha during season 1, 117 Mha during season 2, and 84 Mha continuous. About 58% of the rainfed and 39% of the irrigated were single crops (net cropland area without cropland fallows) cropped during either season 1 (January-May) or season 2 (June-September). The ACMA algorithm was deployed on Google Earth Engine (GEE) cloud computing platform and applied on MODIS time-series data from 2003 through 2014 to obtain ACMA-derived cropland layers for these years (ACL2003 to ACL2014). The results indicated that over these twelve years, on average: (a) croplands increased by 1 Mha/yr, and (b) cropland fallows decreased by 1 Mha/year. Cropland areas computed from ACL2014 for the 55 African countries were largely underestimated when compared with an independent source of census-based cropland data, with a root-mean-square error (RMSE) of 3.5 Mha. ACMA demonstrated the ability to hind-cast (past years), now-cast (present year), and forecast (future years) cropland products using MODIS 250-m time-series data rapidly, but currently, insufficient reference data exist to rigorously report trends from these results.

Explore related subjects

ISPRS Journal of Photogrammetry and Remote Sensing Article

Source-reported bounding extent: -36.03133177633187° to 37.579412513438385° latitude; -18.80859375° to 52.03125° longitude. This indicates report coverage, not an exact sampling location. View area on OpenStreetMap.

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Jun Xiong, Prasad S. Thenkabail, Murali Krishna Gumma, Pardhasaradhi G. Teluguntla, Justin Poehnelt, Russell G. Congalton, Kamini Yadav, David Thau. 2017. Automated cropland mapping of continental Africa using Google Earth Engine cloud computing. https://doi.org/10.1016/j.isprsjprs.2017.01.019

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related USGS reports

USGS2023

National Land Cover Database 2019: A new strategy for creating clean leaf-on and leaf-off Landsat composite images

National Land Cover Database (NLCD) 2019 is a new epoch of national land cover products for the conterminous United States. Image quality is fundamental to the quality of any land cover product. Image preprocessing has often taken a considerable proportion of overall time and effort for this kind of national project. An approach to prepare image inputs for NLCD 2019 production was developed to ensure efficiency and quality of operational production. Here, we introduce a new and comprehensive strategy to produce clear Landsat composite images for NLCD 2019 production. First, we developed a new median-value compositing method. Second, we designed parameter settings for selecting images and pixels to generate 4 composite images (leaf-on, leaf-off, primary reference, and complementary reference) for a target year based on the US Landsat Analysis Ready Data surface reflectance dataset. Third, we developed a method, referred to as Detection and Filling with Simulated Image, to detect and replace clouds and cloud shadow pixels to produce the final clean leaf-on and leaf-off image composites. This image compositing and processing strategy was implemented for the entire conterminous United States to produce images for NLCD 2019. Our image results and NLCD 2019 change detection and land cover products, which were released in July 2021, showed this new strategy to be effective and efficient.

ISPRS Journal of Photogrammetry and Remote Sensing↗

USGS2022

A novel regression method for harmonic analysis of time series

Harmonic analysis of time series is an important technique in remote sensing to reveal seasonal land surface dynamics. However, frequency selection in the harmonic analysis is often difficult because high-frequency components are useful for delineating seasonal dynamics but sensitive to noise and gaps in time series. On the other hand, it is challenging to obtain temporally continuous satellite data with high quality because of atmospheric contamination. We developed a novel regression method named Harmonic Adaptive Penalty Operator (HAPO) for harmonic analysis of unevenly distributed time series. We introduced a new penalty function to minimize unexpected fluctuations in the model, which can substantially reduce the overfitting issue of regression in time series with temporal gaps. Specifically, the new penalty function minimizes the length of the model curve and the value range difference between the model and the time series observations. We compared HAPO with three widely used regression methods (OLS: Ordinary Least Squares; LASSO: Least Absolute Shrinkage and Selection Operator; and Ridge) in different scenarios using Landsat time series data across the United States. First, we evaluated methods using the Landsat surface reflectance time series within a single year. HAPO showed low and consistent monthly Root Mean Square Deviation (RMSD) values, in which most of the time RMSD of predicted reflectance were less than 0.04. More importantly, HAPO showed consistent and less bias given varying density and irregularity of time series. Second, we evaluated methods using multi-year time series. HAPO was a better predictor of relatively short time series (< 4 years) with steady low RMSD values. When a longer time series ( 4 years) was used, all four methods showed similar RMSD values, but HAPO outperformed the other methods if there were temporal gaps. Therefore, for places with large seasonal observation gaps or for time series that are relatively short (less than 4 years), HAPO can provide more consistent and accurate results in harmonic analysis of time series.

ISPRS Journal of Photogrammetry and Remote Sensing↗

USGS2021

Sentinel-2 and WorldView-3 atmospheric correction and signal normalization based on ground-truth spectroradiometric measurements

Remote sensing satellite Earth Observing Systems (EOS) provide a variety of products for monitoring Earth surface processes at varying spatial and spectral resolutions. Combining information from high and medium spatial resolution images is valuable for monitoring ground cover and vegetation status in cropland, grassland, forests, and other natural settings. However, coupling information from different EOS requires compensating for atmospheric and view angle effects before integrating comparable surface reflectance (SR) values. The objectives of this study were i) to assess how different atmospheric constituents affect the atmospheric correction results in Sentinel-2 and WorldView-3 imagery, ii) to establish a relationship with field spectra measurements, and iii) to develop an empirical approach to ensure that SR values extracted from different EOS can be normalized for use in monitoring vegetation and land cover status. We compared surface reflectance values derived from Sentinel-2 images corrected with Sen2Cor, MODTRAN or FLAASH atmospheric correction approaches for the visible-to-near infrared regions. Additionally, this information was compared to SR values extracted from WorldView-3 imagery acquired from the same dates and location (Central Spain) and corrected with MODTRAN and FLAASH approaches. Assessment of the atmospheric correction was conducted by comparing satellite image SR with ground-truth spectra acquired with a FieldSpec hand-held spectroradiometer. The results emphasized the importance of using common atmospheric parameters collected from ancillary data sources (i.e. MODIS Atmosphere & Land products) to ensure a reliable SR comparison. When compared to field-collected spectral data, SR from corrected Sentinel-2 push-broom imagery showed a reliable match (<4% difference in the visible bands and <0.52% difference in the near infrared bands). However, SR imagery from the pointable WorldView-3 instrument showed significant deviation, likely resulting from the effects of steep off-nadir acquisition angles (24.6° to 39.1°) combined with surface anisotropy. The magnitude and sign of the deviation in SR differed depending on the vegetation type, wavelength and sun-surface-sensor geometry. Therefore, it was necessary to account for angular effects to ensure reliable comparisons of imagery from the different EOS. In this study, an empirical angular correction approach was developed based on calibrating each WorldView-3 band against the ground-truth spectra. This correction allowed for the accurate signal normalization of WorldView-3 and Sentinel-2 imagery SR in the visible-to-near infrared regions.

ISPRS Journal of Photogrammetry and Remote Sensing↗