Alexandre Bouvet

The K&C PALSAR Mosaic of the African Continent: Processing Issues and First Thematic Results

The Japan Space Exploration Agency Kyoto and Carbon (K&C) Initiative seeks to demonstrate the potential of the Advanced Land Observing Satellite (ALOS) Phased Arrayed L-band Synthetic Aperture Radar (PALSAR) data for addressing regional applications relating to climate change, carbon cycle science, and environmental conservation. This paper outlines the generation of a regional dual-polarization (HH and HV) mosaic for the entire African continent at spatial resolution on the order of 100 m. The main computational and radar science issues undertaken to generate a seamless mosaic with good radiometric and geometric accuracy are summarized. Preliminary investigations into the thematic information provided by the K&C Africa mosaic and comparisons with the JERS-1 SAR mosaic generated as part of the Global Rain Forest Mapping Project are reported, with emphasis placed on characterizing and detecting change in forests and savannas.

Forest Biomass From Radar Remote Sensing

Abstract: Forests play a primordial role for life on Earth. Beyond their contribution as a major source of raw materials and renewable energy, they also hold an inestimable treasure of biodiversity. They ensure the protection of arable land, are a continuous source of water and contribute to improved air quality. Whether for food or pharmacopoeia, forests are the principal source of subsistence for almost 2 billion people.

An End-to-End Error Model for Classification Methods Based on Temporal Change or Polarization Ratio of SAR Intensities

This paper aims at defining the expression of the probability of error of classification methods using a synthetic aperture radar (SAR) intensity ratio as a classification feature. The two SAR intensities involved in this ratio can be measurements from different dates, polarizations, or, also possibly, frequency bands. Previous works provided a baseline expression of the probability of error addressing the two-class problem with equal a priori class probabilities and no calibration error. This study brings up a novel expression of the error, providing the possibility to assess the effect of class probabilities and calibration errors. An extended expression is described for the n -class problem. The effect of calibration errors such as channel gain imbalance, radiometric stability, and crosstalk is assessed in the general case. The results indicate that, for the applications under study, channel gain imbalance is usually not a decisive parameter, but radiometric stability is more critical in methods based on the temporal change. Crosstalk has a negligible effect in the case of copolarizations. The impacts of other system parameters, such as ambiguity ratio, time-lapse between repeat-pass orbits, spatial resolution, and number of looks, are illustrated through a set of assumptions on the backscattering values of the considered classes. The model is validated by comparing some of its outputs to experimental results calculated from the application of rice fields mapping methods on real data. This error model constitutes a tool for the design of future SAR missions and for the development of robust classification methods using existing SAR instruments.

Mapping of Rice Varieties and Sowing Date Using X-Band SAR Data

Rice is a major staple food for nearly half of the world’s population and has a considerable contribution to the global agricultural economy. While spaceborne Synthetic Aperture Radar (SAR) data have proved to have great potential to provide rice cultivation area, few studies have been performed to provide practical information that meets the user requirements. In rice growing regions where the inter-field crop calendar is not uniform such as in the Mekong Delta in Vietnam, knowledge of the start of season on a field basis, along with the planted rice varieties, is very important for correct field management (timing of irrigation, fertilization, chemical treatment, harvest), and for market assessment of the rice production. The objective of this study is to develop methods using SAR data to retrieve in addition to the rice grown area, the sowing date, and the distinction between long and short cycle varieties. This study makes use of X-band SAR data from COSMO-SkyMed acquired from 19 August to 23 November 2013 covering the Chau Thanh and Thoai Son districts in An Giang province, Viet Nam, characterized by a complex cropping pattern. The SAR data have been analyzed as a function of rice parameters, and the temporal and polarization behaviors of the radar backscatter of different rice varieties have been interpreted physically. New backscatter indicators for the detection of rice paddy area, the estimation of the sowing date, and the mapping of the short cycle and long cycle rice varieties have been developed and assessed. Good accuracy has been found with 92% in rice grown area, 96% on rice long or short cycle, and a root mean square error of 4.3 days in sowing date. The results have been discussed regarding the generality of the methods with respect to the rice cultural practices and the SAR data characteristics.

Estimation of agricultural and biophysical parameters of rice fields in Vietnam using X-band dual-polarization SAR

In this study, we use time-series of X-band Synthetic Aperture Radar (SAR) data to monitor rice fields in the Mekong Delta, Vietnam. Ten dual-polarisation (HH&VV) images of the COSMO-SkyMed constellation are used in the third rice season of 2013, and 3 images in the first season of 2014. In 2013, the season-long comparison of backscatter and biophysical parameters measured in 40 fields allowed us to develop methods to map rice fields at both seasons and to derive the sowing date and the plant biomass at the vegetative stage in 2013.

Use of the SAR Shadowing Effect for Deforestation Detection with Sentinel-1 Time Series

To detect deforestation using Earth Observation (EO) data, widely used methods are based on the detection of temporal changes in the EO measurements within the deforested patches. In this paper, we introduce a new indicator of deforestation obtained from synthetic aperture radar (SAR) images, which relies on a geometric artifact that appears when deforestation happens, in the form of a shadow at the border of the deforested patch. The conditions for the appearance of these shadows are analyzed, as well as the methods that can be employed to exploit them to detect deforestation. The approach involves two steps: (1) detection of new shadows; (2) reconstruction of the deforested patch around the shadows. The launch of Sentinel-1 in 2014 has opened up opportunities for a potential exploitation of this approach in large-scale applications. A deforestation detection method based on this approach was tested in a 600,000 ha site in Peru. A detection rate of more than 95% is obtained for samples larger than 0.4 ha, and the method was found to perform better than the optical-based UMD-GLAD Forest Alert dataset both in terms of spatial and temporal detection. Further work needed to exploit this approach at operational levels is discussed.

Multistability of model and real dryland ecosystems through spatial self-organization

Significance Today, vast areas of drylands in semiarid climates face the dangers of desertification. To understand the driving mechanisms behind this effect, many theoretical models have been created. These models provide insight into the resilience of dryland ecosystems. However, until now, comparisons with reality were merely visual. In this article, a systematic comparison is performed using data on wavenumber, biomass, and migration speed of vegetation patterns in Somalia. In agreement with reaction–diffusion models, a wide distribution of regular pattern wavenumbers was found in the data. This highlights the potential for extrapolating predictions of those models to real ecosystems, including those that elucidate how spatial self-organization of vegetation enhances ecosystem resilience. Spatial self-organization of dryland vegetation constitutes one of the most promising indicators for an ecosystem’s proximity to desertification. This insight is based on studies of reaction–diffusion models that reproduce visual characteristics of vegetation patterns observed on aerial photographs. However, until now, the development of reliable early warning systems has been hampered by the lack of more in-depth comparisons between model predictions and real ecosystem patterns. In this paper, we combined topographical data, (remotely sensed) optical data, and in situ biomass measurements from two sites in Somalia to generate a multilevel description of dryland vegetation patterns. We performed an in-depth comparison between these observed vegetation pattern characteristics and predictions made by the extended-Klausmeier model for dryland vegetation patterning. Consistent with model predictions, we found that for a given topography, there is multistability of ecosystem states with different pattern wavenumbers. Furthermore, observations corroborated model predictions regarding the relationships between pattern wavenumber, total biomass, and maximum biomass. In contrast, model predictions regarding the role of slope angles were not corroborated by the empirical data, suggesting that inclusion of small-scale topographical heterogeneity is a promising avenue for future model development. Our findings suggest that patterned dryland ecosystems may be more resilient to environmental change than previously anticipated, but this enhanced resilience crucially depends on the adaptive capacity of vegetation patterns.

Two-point statistic of polarimetric sar data provided by a wavelet frame

The spatial correlation properties of the backscattered power in polarimetric SAR data are characterized by means of a two-point statistic provided by a wavelet frame that acts as a differential operator. In particular, the dependence of this statistic on a rotation of the linear polarization basis is considered. On the one hand, a theoretical model is proposed, which enables the numerical simulation of the wavelet variance of the crosspol and copol power synthesized in the rotated basis, given the power spectra of the feature vector in the H,V basis. On the other hand, experimental analysis of polarimetric SAR data is undertaken using wavelet variance estimates as a function of scale and polarization state. Results are encapsulated in a graphical form called the wavelet scaling polarimetric signature (WASPS).

A land cover map of Africa at 100 meters resolution using a mosaic of alos PALSAR dual-polarization data: Preliminary developments

In this paper, we investigate the potential of large-scale mosaics of Synthetic Aperture Radar (SAR) data for land cover mapping. The study is based on a wall-to-wall mosaic of double-polarization data (HH and HV) from the L-band sensor PALSAR, covering the whole African continent at a spatial resolution of about 100m. The GlobCover 2009 global land cover map is taken as a reference for the training of the classification algorithm. The joint use of the PALSAR mosaic, the GlobCover 2009 map and the Shuttle Radar Topography Mission (SRTM) Digital Elevation Model (DEM) allows retrieving backscatter signatures that take into account local topography (slope angle and slope orientation) for each of the 22 GlobCover classes. These signatures are then used to assign each pixel of the mosaic to one class. The 22 classes are finally combined to coarser classes: tree cover, shrub cover, herbaceous cover/bare soil, and water. Because of ecological and phenological variations over the continent, this land cover mapping scheme is actually applied to 5° by 5° tiles. The methodology has been developed so far on two tiles which illustrate different ecoregions. A visual assessment indicates that the classified maps are very satisfying, at least for the tree cover class, with a significantly improved spatial resolution compared to existing large-scale land cover products. Validation is ongoing before applying the method to the whole Africa in the future.

Characterization of singular structures in polarimetric SAR images by wavelet frames

Discontinuities in polarimetric SAR backscattered intensity (edges and point targets) can be characterized by a mathematical model. This paper presents a technique based on the Lipschitz regularity of an underlying singular function for detecting and interpreting such discontinuities. Numerical estimators of the Lipschitz parameters (exponent, swing and smoothing kernel variance) are implemented by a multi-voice wavelet frame transform. Local (one point in space) supervised estimates of the parameters and their dependence on polarization state are obtained by extracting the trajectory of the wavelet modulus maxima in the space-scale-polarization domain and by a non-linear regression with respect to a theoretical trend with scale. Global image-wide approximations of the Lipschitz parameters are also proposed, these aiming at a spatial description of the discontinuitys type, sharpness, and dependence on the polarization state. Examples are reported concerning experiments using simulated and real SAR polarimetric data.