Search

SIELETERS is an airborne hyperspectral imaging system that covers the mid- and long-wavelength infrared bands. It is part of the SYSIPHE system that covers bands from 0.4 μm to 11.8 μm. The ground sampling distance of the SIELETERS instrument is 0.5 m when 2000 m above the ground and the spectral resolution is between 6.7 cm-1 and 14.4 cm-1. In this article, we introduce the SYSIPHE project, the SIELETERS system and we show the latest results that we have obtained with the system. Our recent studies led to significant improvements on the quality of the hyperspectral images and of the spectra. The hyperspectral images from the previous airborne campaigns are now compatible with atmospheric compensations, data classifications and anomaly detection. The system is opened to external users. [ABSTRACT FROM AUTHOR]

Accurate and consistent reporting on forest cover and land use/cover changes is crucial in the post-Paris agreement context where countries are obliged to regularly submit their reports on the greenhouse gas balance of emissions and removals. This reporting of changes can have a direct impact on payments through comparisons with national reference (emission) levels in the context of reducing emissions from deforestation and forest degradation, particularly as part of the REDD+ process (Reducing emissions from deforestation and forest degradation). However, before doing so, countries are invited to strengthen their national forest monitoring system and also to produce map products that comply with robust methods for assessing the uncertainty of forest cover change estimates as prescribed by UNFCCC (United Nations Framework Convention on Climate Change). In this study, we present the results of the accuracy of the numerous forest maps available in Congo using a new method developed by SIRS and adapted to the context of the Republic of Congo. The results of the study show an underestimation of nearly 50% of forest losses in the period between 2000 and 2014 either by the direct estimation method or by regression estimates. Comparison of the two methods shows that the estimates of forest cover losses by the sample point method underestimates the estimate of losses over the period 2000-2014 as well as over the period 2014-2016. We conclude that there is a need to strengthen the national teams in charge of forest mapping, but also that the method by Sannier et al. 2014 should be adopted for the evaluation of precision. [ABSTRACT FROM AUTHOR]

This work deals with the evaluation of the potential of multi-temporal and multi-sensor (optical and radar) images of the satellites Sentinel-1 and Sentinel-2 satellites for land cover mapping and coastline evolution monitoring in a tropical ecosystem over the period 2015 to 2020. The study area chosen is the Cameroon Estuary. This area represents a major ecological transition in the sub-region with the presence of protected natural reserves (mangroves, dense forest, wetlands) but also a strong anthropic activity (construction, agriculture, degraded forest). The methodological approach consisted of a chain of pre-processing and visual image analysis, followed by a combination of the bands from each sensor, a supervised Random Forest classification to draft a land use map and finally a digitization of the coastline. Overall, the results show that the classification with the Sentinel- 2 images using 10 bands and adding 4 vegetation indices is slightly more accurate (95.75 %) than the classification with the 13 initial bands (91.78 %). The classification performed with the Sentinel-1A multi-temporal image series (dual polarization (VV, VH) and VV/VH ratio) leads to an accuracy of 78.44 %. The combination of Sentinel-2 A&B bands (10 bands and 4 indices) and Sentinel-1 improves the results and leads to an accuracy of 98.76 %. Our results also show that the use of multi-temporal image time series significantly improves the classification accuracy compared to the use of a single image (mono-date), and this for both sensors, i.e. an additional gain of 13 % and 10 % respectively for Sentinel-2 and Sentinel-1. However, this gain remains small for temporally stable classes. The results of the coastline evolution analysis show that the Cameroon estuary is disturbed by different levels of erosion (Cape Cameroon, North Manoka Island, Sanaga River mouth), and accretion (Limbe, and South Manoka Island) and also by weak internal variations. The fusion of optical and radar remote sensing data in the discrimination of land use classes, has shown that the construction zones are the most vulnerable to coastal erosion. On the other hand, the presence of vegetation (mangrove, forest) stabilizes and protects the coast from possible risks of this nature. [ABSTRACT FROM AUTHOR]

The mapping of the land use of Gabon in 2010 and 2015 and the changes between these two dates was carried out as part of the OSFACO program (Space observation of Central and Western African Forests) by the Gabonese Agency for Space Studies and Observations (AGEOS). This is the first reference land use map at the national level, including the legend resulting from a consultation of major players in the Gabonese forest sector. This map was obtained by refining the forest cover map of Gabon (forest/non-forest) obtained in the framework of OSFT (Spatial Observation of Tropical Forests) and GEOFORAFRI project (Capacity building and access to earth observation satellite data for monitoring forests in Central and Western Africa). The mapping was based on the use of SPOT 5/7, Sentinel 2 satellite images for the year 2015 and the SPOT 4, ASTER and ALOS 1 images for the year 2010. Semi-automated classification method and Manual enhancement were combined for better accuracy of land use classes. The results obtained highlight ten land use classes representing the country's large landscape groups dominated by forests which occupied 89% of the total area of the country in 2015. The most dynamic land use classes are forests, savannas, agricultural land and artificial surfaces. The main forest losses are linked to the conversion of forests to artificial surfaces, cultivated land and bare soils. The most significant gains in forest are observed at the level of the closure of forest roads. The validation of the product, carried out by an independent expertise from the map production team, was based on a reference data resulting from a sampling plan combining a systematic and random component according to the methodology of Sannier et al. (2016). The analysis of the correspondence between the land use map produced and the reference data gave an overall accuracy of 95%. [ABSTRACT FROM AUTHOR]

Despite the increasing openness and democratisation of satellite data and associated processing tools, the use of these data and the information derived from them remains relatively limited in the Southern countries. For several years now, initiatives have been launched to improve access and use by local actors. This paper presents the results of 3 successive projects funded by the French Development Agency in Central and West Africa between 2010 and 2020. These projects, led by IGN FI and the French National Research Institute for Sustainable Development (IRD), have provided some answers to the main challenges identified by local partners such as: access to data, improving data processing skills, raising awareness among decision-makers and land managers of the information and services that can be derived from it, reference land cover maps production by local teams, and the development and structuring of user networks. This paper presents the actions and results in three areas of intervention, namely (i) the free provision of recent high-resolution data, (ii) the local production of forest cover and land use change maps and (iii) the promotion of the use of images through training and scientific activities. The results of several projects supported within the framework of these actions are published in this special Africa issue. [ABSTRACT FROM AUTHOR]

How can sustainable agricultural development be reconciled with the fight against climate change and the preservation of biodiversity? To reach such a challenge, political, technical, and socio-economic decision-makers now have access to accurate and valuable environmental information. These include satellite images and environmental data produced from these images. In Guinea, in order to carry out public policies for the sustainable management of territories and natural resources, the government must be in a position to have reliable, consistent, precise, and as recent as possible information on the state and dynamics of its territory (soil artificialisation, agricultural extension, deforestation, mining...) and on the evolution of certain features of the biosphere (forests, savannas, wetlands, mangroves...). To this end, the Guinean Ministry of Agriculture has relied on the French technical expertise to draw up an agroecological zoning for the country, aimed at identifying and quantifying current agricultural areas, assessing agricultural potential, and mapping land cover and land cover change over a 10-year period (2005-2015). This project was implemented over two years in a partnership framework between a team of French thematic experts and a team of Guinean technicians and experts involving a specific organisation and procedure, to ensure a complete transfer of skills on all project activities, from the realisation of a land cover layer to relevant applications in the field of agriculture and the environment, with the aim of providing decision support tools to multiple Guinean actors. [ABSTRACT FROM AUTHOR]

The role of forests in carbon storage in a climate change context is relevance. In this paper, we will assess the carbon stock in the Libreville area using remote sensing data over a 2001-2018 period. This assessment of carbon stocks in forests is mainly based on Landsat optical imagery. A first analysis aimed at producing a land use mapping between 2001 and 2018 was carried out. These preliminary results were then used as a proxy for the assessment of carbon stocks using terrestrial and mangrove forests. The average value of carbon stock contained in each class comes from the bibliographical work carried out mainly in the same study area. The total value of the carbon stock was estimated at 70 913 018 ± 896 and 68 340 383 ± 927 tons in 2001 and 2018 respectively. [ABSTRACT FROM AUTHOR]

In remote sensing, there is a large number of algorithms to classify a satellite image. Among these classification algorithms, the Random Forest appears to be particularly powerful. The objectives of this study are to evaluate (1) the importance of image selection for the level of accuracy of the training model and (2) the nature of the relationship between the level of accuracy of the model and the overall accuracy of the thematic map resulting from the classification of the satellite image with this algorithm. Based on a Landsat 8 OLI image taken over a tropical mountain area : the West Cameroon region, 35 models were built and tested. Analysis shows that the level of overall accuracy of the Random Forest results are closely dependent on the accuracy of the training model used to classify the satellite image, and on the choice of the images used to train this model. Moreover, the selection of these images is itself dependent on the quality of the Regions of interest (ROI) that will be used to build the model. It is therefore important to place particular emphasis on the quality of the input data in order to guarantee satisfactory results with this algorithm. Otherwise, the performance of this algorithm could be inferior to that of other classification algorithms. [ABSTRACT FROM AUTHOR]

Remote sensing is a particularly suitable tool for monitoring land use but also particularly suited for déforestation monito-ring. By launching the Setinel-1 and Sentinel-2 satellites in the Copernicus program, the community now has free data with important time-revisits allowing the greatest number of people to effectively monitor land cover of a study area. This paper présents a land use monitoring method based on the combination of two approaches computed using Open Source tools (QGIS, Orfeo ToolBox, python). First, we focus on land use monitoring at an annual time scale using Sentinel-1 and Sentinel-2 data, and then, by the use f entienl-1 data, we detect changes in the forest cover due to déforestation at bi-monthly time scale. The results obtained show a very good synergy of thèse approaches allowing the complementary of optical and radar data. In order to make accessible the proposed method, ail the used open source tools are available on this link http ://remotesensing4all.net/index.php/2018/09/11/kit-use-des-donnees-radar-sentinel-1-in-de-latelier-radar-du-FOSS4G-en-2018-2/. [ABSTRACT FROM AUTHOR]

The organic matter (OM) is an important element of the fertility of cultivated soils. Reflectance measurements have been successfully used to predict soil organic carbon (OC) content of agricultural soils, measured in the laboratory on samples of dried soil or directly in the field using field spectro-radiometer or satellite imagery. This study proposes a method for spatial prediction of soil OC, from UAV imagery. An agricultural plot of 13 ha "Grande Borne" located in the plain of Versailles (78) was analysed in April 2013, shortly before sowing while it was still bare soil. The AIRINOV® UAV used was equipped with a four channels multispectral camera MultiSPEC 4C® (550nm, 660nm, 735 nm and 790 nm) and flying at an altitude of 150m. Twenty three ground control points, distributed in the plot were delineated by target located with a centimetric DGPS and different measures have been made synchronously with the drone flight: Spectral Measurements (spectroradiometer ASD FieldSpec 3®), roughness measurements by photogrammetric method, samples of soil samples for analysis of chemical composition and water content. The images obtained with a resolution of 15cm have been georeferenced with an infra-pixel accuracy. The field spectra were aggregated on the spectral bands of the Multispec 4C for validation by linear regression. The obtained R2, between 0.8 and 0.9 by band, with an error between 1 and 3%, have shown good reliability of the Multispec 4C® sensor. The OC content was then estimated by partial least squares regression (PLSR) from spectral measurements of the 23 plots. The mean squared error of cross validation (RMSECV) by LOO (Leave One Out) method was 1.97 g of OC per kg of soil. A second correction of the model incorporating the effects of moisture and roughness on reflectance, has improved the quality of the prediction by 18% and a RMSECV of 1.61 g / kg. The model was finally spatialized in GIS and compared to simple spatial interpolations of soil analyses. [ABSTRACT FROM AUTHOR]

The recent democratization of drones creates a methodological break for archaeological remote sensing. The first point concerns the ease of use of these vectors, which provide secure flight conditions, thanks to the GPS and the ability to set waypoints for semi-automated flight navigation systems. This flexibility also allows repeated flights over the same area. The second interesting point for archaeological remote sensing is the low flight altitude of drones, allowing documentation of an intermediate level between the site and the micro-region. Low altitude is also a way to gain better information and higher resolution. The last point is the low cost of using, even if the purchase price of the equipment is quite high. The development of UAV for archaeological remote sensing implies to define new uses for traditional sensors. We describe those developed in the framework of the Archeodrones program on airborne photogrammetry and thermography New sensors will probably be adapted on drones in the future, allowing greater access to non-destructive methods of survey. [ABSTRACT FROM AUTHOR]

Monitoring forest state is required to estimate the levels of greenhouse gases emissions from deforestation. The objective of this study was to estimate deforestation trends for 2005-2010 and 2010-2013 periods in the eastern rainforests ecoregion of Madagascar. A supervised classification approach of multi-date Landsat images using the Random Forest algorithm was used. A complete 30 m map of the forest cover change was produced on 20.5 million hectares. The result was validated by photo interpretation of nearly 11 000 points from a regular sampling grid on the basis of SPOTand Landsat satellite images acquired in 2010-2013. The overall mapping accuracy was estimated to be about 90%. The area of Madagascar rainforests has been estimated at 4.5 million hectares in 2005, 4.4 million hectares in 2010 and 4.3 million hectares in 2013. The annual rate of deforestation has increased from 0.5%. to 0.9 % between the two periods. This study could be considered as a reliable cost-effective and reproductible approach for monitoring forests in Madagascar. [ABSTRACT FROM AUTHOR]

Large-scale characterization of tropical rainforest is a challenge for their conservation and management. Very high spatial resolution images as provided by the Pléiades satellites offer new opportunities to study the structural organization of heterogeneous rainforests with limited accessibility. In this study, we have evaluated the potential of Pléiades images to map structural parameters of New Caledonian rainforests by analyzing texture of forest canopies. We have applied the Fourier transform textural ordination (FOTO) method to very high spatial resolution images to compute texture indices of canopy grain (i.e. a combination of size distribution and spatial pattern of tree crowns). The results have showed that this method was promising to isolate the rainforest from other types of vegetation, and to highlight their structural diversity on a large scale. Finally, this case study showed that the use of Pléiades images is promising to predict the structure of rainforests. [ABSTRACT FROM AUTHOR]

A statistical model for detecting changes in remote sensing images has recently been proposed in (Prendes et al., 2014, 2015). This model is sufficiently general to be used for homogeneous images acquired by the same kind of sensors (e.g., two optical images from Pléiades satellites, possibly with different acquisition conditions), and for heterogeneous images acquired by different sensors (e.g., an optical image acquired from a Pléiades satellite and a synthetic aperture radar (SAR) image acquired from a TerraSAR-X satellite). This model assumes that each pixel is distributed according to a mixture of distributions depending on the noise properties and on the sensor intensity responses to the actual scene. The parameters of the resulting statistical model can be estimated by using the classical expectation-maximization (EM) algorithm. The estimated parameters are finally used to learn the relationships between the images of interest, via a manifold learning strategy. These relationships are relevant for many image processing applications, particularly those requiring a similarity measure (e.g., image change detection and image registration). The main objective of this paper is to evaluate the performance of a change detection method based on this manifold learning strategy initially introduced in (Prendes et al., 2014, 2015). This performance is evaluated by using results obtained with pairs of real optical images acquired from Pléiades satellites and pairs of optical and SAR images. [ABSTRACT FROM AUTHOR]

Knowing the distribution of vectors of zoonoses is fundamental to understand epidemiological dynamics and characterize risk areas. This distribution can be assessed at different scales: regionally, the range of these species defines the limits of occurrence of pathogen transmission and, locally, niches determine the hazard. This latter scale requires a very precise knowledge of the areas that can be gained through the analysis of very high spatial resolution satellite images. As part of a research project on leptospirosis in La Réunion (LeptOI), this study proposes to measure the potential of a Pleiades image (product resampled to to 50 cm) to characterize the local distribution of various small terrestrial mammals in Saint-Paul swamp area. Three trapping campaigns have led to the capture of 134 animals belonging to four of the five species occurring in Reunion Island (Rattus rattus, R. norvegicus, Suncus murinus and Tenrec ecaudatus). An objectbased image analysis of the Pleiades image allowed discriminating ten land use classes with a total accuracy of 83.6%. Landscape indices were calculated from the classification around each trap (shortest distances to each land use class, edge densities and proportion of each patch within buffers defined by a 50 meters or a 100 meters radius). Three areas ("wet", "urbanized" and "dry") are clearly discriminated by these variables, which can be used as ecological determinants of different epidemiological profiles. The wetland area is known as suitable to leptospirosis whereas the urbanized area is a place of co-occurrence of the four species and thus potentially favorable to the exchange of pathogens. [ABSTRACT FROM AUTHOR]

Segmentation is a widely used operation in very high resolution remote sensing processing such as object based image analysis. Since the available memory resource might be limited, it is often impossible to process a whole satellite image without using piece-wise processing, which in the case of segmentation introduces a huge amount of artifacts. The work presented in this paper introduce a solution to this problem in the case of the Mean-Shift algorithm, which guarantees that piece-wise segmentation results matches exactly those of full image processing at once. First, we define a new property of segmentation algorithms called stability. After proposing a methodology to measure the stability of segmentation algorithms, we demonstrate that among the Mean-Shift, watershed and connected components algorithms, only the latter is stable. Then, we propose a stabilized version of the Mean-Shift algorithm and use it to build a rigorous and exact solution for piece-wise processing of this algorithm. Last, we present some examples demonstrating the usefulness of the proposed method. This method is available in the Orfeo ToolBox free software, and documented in the software guide. [ABSTRACT FROM AUTHOR]

In 2009 a preliminary feasability study on the fourth ring road of Alger was carried out based on SPOT5 images. A comparative study was also conducted with panchromatic (PAN) and multispectral (MS) Pléiades images. The focus of the studies has distinguished six different geological formations. The diversity of terrain encountered has enabled the comparison of parameters such as the resolution and the spectral band. From a geological point of view at macroscopic scale, both satellites enable an aerial sketch mapping of the lineaments. However, a more precise structural mapping was observed with Pléiades PAN images. For SPOT5 and Pléiades MS images, even though their resolution is quite similar (respectively 2,5 m and 2,8 m), a stronger spectral contrast was observed with Pléiades MS which have enabled to highlight the different lithologies. For smaller geological details (1 m), Pléiades PAN images have given good results. Hence, because preliminary studies have to be the most advanced possible, it is clear that Pléiades images are the one that can lead to an exhaustive mapping of the geomorphology encountered. [ABSTRACT FROM AUTHOR]