Sophie Bontemps

Assessment of an Operational System for Crop Type Map Production Using High Temporal and Spatial Resolution Satellite Optical Imagery

Crop area extent estimates and crop type maps provide crucial information for agricultural monitoring and management. Remote sensing imagery in general and, more specifically, high temporal and high spatial resolution data as the ones which will be available with upcoming systems, such as Sentinel-2, constitute a major asset for this kind of application. The goal of this paper is to assess to what extent state-of-the-art supervised classification methods can be applied to high resolution multi-temporal optical imagery to produce accurate crop type maps at the global scale. Five concurrent strategies for automatic crop type map production have been selected and benchmarked using SPOT4 (Take5) and Landsat 8 data over 12 test sites spread all over the globe (four in Europe, four in Africa, two in America and two in Asia). This variety of tests sites allows one to draw conclusions applicable to a wide variety of landscapes and crop systems. The results show that a random forest classifier operating on linearly temporally gap-filled images can achieve overall accuracies above 80% for most sites. Only two sites showed low performances: Madagascar due to the presence of fields smaller than the pixel size and Burkina Faso due to a mix of trees and crops in the fields. The approach is based on supervised machine learning techniques, which need in situ data collection for the training step, but the map production is fully automatic.

An Automated Method for Annual Cropland Mapping along the Season for Various Globally-Distributed Agrosystems Using High Spatial and Temporal Resolution Time Series

Cropland mapping relies heavily on field data for algorithm calibration, making it, in many cases, applicable only at the field campaign scale. While the recently launched Sentinel-2 satellite will be able to deliver time series over large regions, it will not really be compatible with the current mapping approach or the available in situ data. This research introduces a generic methodology for mapping annual cropland along the season at high spatial resolution with the use of globally available baseline land cover and no need for field data. The methodology is based on cropland-specific temporal features, which are able to cope with the diversity of agricultural systems, prior information from which mislabeled pixels have been removed and a cost-effective classifier. Thanks to the JECAM network, eight sites across the world were selected for global cropland mapping benchmarking. Accurate cropland maps were produced at the end of the season, showing an overall accuracy of more than 85%. Early cropland maps were also obtained at three-month intervals after the beginning of the growing season, and these showed reasonable accuracy at the three-month stage (>70% overall accuracy) and progressive improvement along the season. The trimming-based method was found to be key for using spatially coarse baseline land cover information and, thus, avoiding costly field campaigns for prior information retrieval. The accuracy and timeliness of the proposed approach shows that it has substantial potential for operational agriculture monitoring programs.

Automated Training Sample Extraction for Global Land Cover Mapping

Land cover is one of the essential climate variables of the ESA Climate Change Initiative (CCI). In this context, the Land Cover CCI (LC CCI) project aims at building global land cover maps suitable for climate modeling based on Earth observation by satellite sensors. The challenge is to generate a set of successive maps that are both accurate and consistent over time. To do so, operational methods for the automated classification of optical images are investigated. The proposed approach consists of a locally trained classification using an automated selection of training samples from existing, but outdated land cover information. Combinations of local extraction (based on spatial criteria) and self-cleaning of training samples (based on spectral criteria) are quantitatively assessed. Two large study areas, one in Eurasia and the other in South America, are considered. The proposed morphological cleaning of the training samples leads to higher accuracies than the statistical outlier removal in the spectral domain. An optimal neighborhood has been identified for the local sample extraction. The results are coherent for the two test areas, showing an improvement of the overall accuracy compared with the original reference datasets and a significant reduction of macroscopic errors. More importantly, the proposed method partly controls the reliability of existing land cover maps as sources of training samples for supervised classification.

Production of a Dynamic Cropland Mask by Processing Remote Sensing Image Series at High Temporal and Spatial Resolutions

The exploitation of new high revisit frequency satellite observations is an important opportunity for agricultural applications. The Sentinel-2 for Agriculture project S2Agri (http://www.esa-sen2agri.org/SitePages/Home.aspx) is designed to develop, demonstrate and facilitate the Sentinel-2 time series contribution to the satellite EO component of agriculture monitoring for many agricultural systems across the globe. In the framework of this project, this article studies the construction of a dynamic cropland mask. This mask consists of a binary “annual-cropland/no-annual-cropland” map produced several times during the season to serve as a mask for monitoring crop growing conditions over the growing season. The construction of the mask relies on two classical pattern recognition techniques: feature extraction and classification. One pixel- and two object-based strategies are proposed and compared. A set of 12 test sites are used to benchmark the methods and algorithms with regard to the diversity of the agro-ecological context, landscape patterns, agricultural practices and actual satellite observation conditions. The classification results yield promising accuracies of around 90% at the end of the agricultural season. Efforts will be made to transition this research into operational products once Sentinel-2 data become available.

Monitoring forest changes in Borneo on a yearly basis by an object-based change detection algorithm using SPOT-VEGETATION time series

Monitoring land cover over large areas on a yearly basis is challenging. The spatial and temporal consistency of an object-based change detection algorithm was tested through a multi-year application on the forest of Borneo, using SPOT-VEGETATION time series from 2000 to 2008. Continuous change thresholds allowed the tuning of the algorithm according to specific requirements in terms of omission and commission errors. The accuracy of the method was assessed using the ROC (relative operating characteristics) curves, which were found useful to evaluate the performance of the method independently of the selected threshold and to support the selection of an optimal threshold. The forest area that annually changed between 2000 and 2008 was detected and a cumulative change map was produced. The resulting change rates and the distribution of the forest change patterns were in agreement with other sources of information. These results demonstrated the very promising temporal consistency of the proposed approach. Further work aims at testing it at larger scales.

A global NDVI and EVI reference data set for land-surface phenology using 13 years of daily SPOT-VEGETATION observations

Time series of vegetation indices (VIs) obtained by remote sensing are widely used to study phenology on regional and global scales. The aim of the study is to design a method and to produce a reference data set describing the seasonal and inter-annual variability of the land-surface phenology on a global scale. Specific constraints are inherent in the design of such a global reference data set: (1) the high diversity of vegetation types and the heterogeneous conditions of observation, (2) a near-daily resolution is needed to follow the rapid changes in phenology, (3) the time series used to depict the baseline vegetation cycle must be long enough to be representative of the current vegetation dynamic and encompass anomalies, and (4) a spatial resolution consistent with a land-cover-specific analysis should be privileged. This study focuses on the SPOT (Satellite Pour l’Observation de la Terre)-VEGETATION sensor and its 13-year time series of reflectance values. Five steps addressing the noise and the missing data in the reflectance time series were selected to process the daily multispectral reflectance observations. The final product provides, for every pixel, three profiles for 52 × 7-day periods: a mean, a median, and a standard deviation profile. The mean and median profiles represent the reference seasonal pattern for variation of the vegetation at a specific location whereas the standard deviation profile expresses the inter-annual variability of VIs. A quality flag at the pixel level demonstrated that the reference data set can be considered as a reliable representation of the vegetation phenology in most parts of the Earth.

New global land cover mapping exercise in the framework of the ESA Climate Change Initiative

The ESA Climate Change Initiative land cover project focuses on the deriving land cover information driven by requirements for observing Essential Climate Variables. Consultation mechanisms were established with the climate modelling community in order to identify its specific needs in terms of satellite-based global land cover products. Key findings were the needs for successive land cover maps stable over time. As response, an innovative global land cover mapping approach, based on multi-year MERIS and SPOT-Vegetation datasets is proposed. Pre-processing and classification chains able to handle huge amount of data have been developed and a first global land cover map associated to the 2008-2010 epoch is being produced.

Compilation and Validation of SAR and Optical Data Products for a Complete and Global Map of Inland/Ocean Water Tailored to the Climate Modeling Community

Accurate maps of surface water extent are of paramount importance for water management, satellite data processing and climate modeling. Several maps of water bodies based on remote sensing data have been released during the last decade. Nonetheless, none has a truly (90 ∘ N/90 ∘ S) global coverage while being thoroughly validated. This paper describes a global, spatially-complete (void-free) and accurate mask of inland/ocean water for the 2000–2012 period, built in the framework of the European Space Agency (ESA) Climate Change Initiative (CCI). This map results from the synergistic combination of multiple individual SAR and optical water body and auxiliary datasets. A key aspect of this work is the original and rigorous stratified random sampling designed for the quality assessment of binary classifications where one class is marginally distributed. Input and consolidated products were assessed qualitatively and quantitatively against a reference validation database of 2110 samples spread throughout the globe. Using all samples, overall accuracy was always very high among all products, between 98 % and 100 % . The CCI global map of open water bodies provided the best water class representation (F-score of 89 % ) compared to its constitutive inputs. When focusing on the challenging areas for water bodies’ mapping, such as shorelines, lakes and river banks, all products yielded substantially lower accuracy figures with overall accuracies ranging between 74 % and 89 % . The inland water area of the CCI global map of open water bodies was estimated to be 3.17 million km 2 ± 0.24 million km 2 . The dataset is freely available through the ESA CCI Land Cover viewer.

“Sentinel-2 for agriculture”: Supporting global agriculture monitoring

Developing better agricultural monitoring capabilities based on Earth Observation data is critical for strengthening food production information and market transparency. In 2014, the European Space Agency launched the Sentinel-2 for Agriculture project which aims at preparing the exploitation of Sentinel-2 data for agriculture monitoring through the development of an open source system able to generate relevant agricultural products. In order to meet this objective, the project carried out a benchmarking exercise to identify the best algorithms that will be in this system. For each product, a minimum of five algorithms were tested over 12 sites globally distributed. This paper gives a general overview of the project and presents in detail the benchmarking.

Mapping Forest Change in Borneo in 2000-2006 by a Multispectral Statistically-Based Detection Technique with SPOT-VEGETATION

This research aims to develop a multispectral, object-oriented and statistically-based change detection method using SPOT-VEGETATION time series. The dataset consists in daily images spanning years 2000 to 2006 and covering Bornean forest ecosystems. Seasonal composites are processed and homogeneous objects are delineated through a multi-temporal segmentation procedure. Each object is defined by a signature describing its spectral behaviour during years to compare. Such signature is formed by reflectance values of analyzed composites and is statistically expressed allowing accounting for temporal correlations existing between and within time series. Each object is compared to an unchanged situation through a distance computation and a Chi-square test. Objects associated to high distances -indicative of change -are identified and associated to a change probability. The algorithm analyses red, NIR, SWIR and NDVI channels independently, resulting in four distinct sets of detections. These results are then combined to keep only objects detected as changed by the four channels. The overall accuracy ranges from 80% to 91% according to the threshold of probability used and the compared years. In terms of change temporal trajectories, we note a sharp increase of changed area in 2003. We also observe that from 2000 to 2006, fire is always the dominant change factor, with 62% to 82% of detections corresponding to burned areas. Our observations confirm that Borneo currently undergoes environmental degradation dynamics.