Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Les écosystèmes forestiers d’Afrique sont en proie à la déforestation et subissent des dégradations, ayant pour conséquences une pression importante sur les services écosystémiques et une contribution à l’émission de gaz à effet de serre dans l’atmosphère. L’accessibilité et le faible coût des technologies liées à la télédétection en font des outils valorisables et prometteurs à la caractérisation de forêts claires de type miombos, première étape vers le suivi temporel des stocks de biomasse végétale et de carbone. L’objectif général de ce travail de fin d’étude est de caractériser la structure forestière d’un miombo en périphérie de la ville de Lubumbashi (RDC) à l’aide d’imageries drone et satellitaires. Une approche arbre au départ du Modèle Numérique de Hauteur de la zone d’étude de 10 hectares construit sur base des imageries acquises par drone a été mise en place, d’une part pour évaluer la performance de l’algorithme de détection des arbres individuels, et d’autre part pour valider les hauteurs extraites du Modèle Numérique de Hauteur en les comparant avec les mesures faites sur le terrain. L’identification des arbres individuels est jugée satisfaisante (score F = 0.76) de même que les hauteurs extraites du Modèle Numérique de Hauteur (R² = 0.835, RMSE% = 10.99%). Ensuite, une approche surface a permis de mettre en évidence un modèle d’estimation de la biomasse aérienne au départ du volume sous le Modèle Numérique de Hauteur volha et du coefficient de variation de la hauteur hcv (variables surfaciques extraites du Modèle Numérique de Hauteur) pour une taille de parcelle de 0.25 ha (R² aj = 0.65, RMSE% = 14.43%). Finalement, un test de corrélation linéaire de Pearson a été fait entre la biomasse aérienne et, d’un côté, cinq indices de végétation calculés au départ des bandes spectrales des imageries Sentinel-2 et, d’un autre côté, l’intensité de rétrodiffusion en polarisations VV et VH d’imageries Sentinel-1. Les hautes valeurs de biomasse aérienne ainsi que leur faibles gammes n’ont pas permis de mettre en évidence une relation linéaire probante. De ce fait, aucun modèle d’estimation de la biomasse aérienne au départ d’imageries satellitaires n’a été construit. Les outils de télédétection appliqués aux miombos humides sont prometteurs au vu des résultats obtenus, ayant néanmoins permis de mettre en évidence l’importance d’un échantillonnage adéquat. Il est dès lors recommandé de combiner les technologies satellitaires, d’étendre la gamme de biomasse aérienne et d’augmenter le nombre de parcelles afin d’optimiser la représentativité de l’échantillonnage et de pouvoir construire des modèles d’estimation de la biomasse aérienne valides. Africa’s forest ecosystems are suffering from deforestation and degradation, resulting in significant pressure on ecosystem services and a contribution to the emission of greenhouse gases into the atmosphere. The accessibility and low cost of remote sensing technologies make them valuable and promising tools for characterizing miombo woodlands, a first step towards temporal monitoring of plant biomass and carbon stocks. The general objective of this master thesis is to characterize the forest structure of a miombo on the outskirts of Lubumbashi (DRC) using UAV and satellite imagery. A tree-based approach based on the Canopy Height Model of the 10 hectares study area derived from UAV images was implemented, on the one hand to evaluate the performance of the algorithm for detecting individual trees, and on the other hand to validate the heights extracted from the Canopy Height Model by comparing them with the measurements made in the field. The identification of individual trees is considered satisfactory (score F = 0.76) as well as the heights extracted from the Canopy Height Model (R² = 0.835, RMSE% = 10.99%). Then, an area-based approach made possible the development of a model for estimating above-ground biomass from the volume under the Canopy Height Model volha and the coefficient of variation of height hcv (metrics derived from the Canopy Height Model) for a plot size of 0.25 ha (R² adj = 0.65, RMSE% = 14.43%). Finally, a Pearson linear correlation test was performed between the above-ground biomass and, on the one hand, five vegetation indices computed from the spectral bands of Sentinel-2 imageries and, on the other hand, the backscatter intensity in VV and VH polarizations of Sentinel-1 imageries. The high values of above-ground biomass and their low ranges did not reveal a convincing linear relationship. As a result, no model for estimating above-ground biomass from satellite imagery has been developed. Remote sensing tools applied to wet miombos are promising in view of the results obtained, but have nevertheless highlighted the importance of adequate sampling. It is therefore recommended to combine satellite technologies, extend the range of above-ground biomass and
increase the number of plots in order to optimise the representativeness of the sampling and to be able to build valid above-ground biomass estimation models.

forêt claire, miombo, biomasse aérienne, AGB, drone, Sentinel-1, Sentinel-2, Modèle Numérique de Hauteur --- miombo woodland, above-ground biomass, AGB, UAV, Sentinel-1, Sentinel-2, Canopy Height Model, Individual Tree Detection, Area-based approach --- Sciences du vivant > Multidisciplinaire, généralités & autres --- Ingénierie, informatique & technologie > Multidisciplinaire, généralités & autres

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Past research has comprehensively assessed the capabilities of satellite sensors operating at microwave frequencies, both active (SAR, scatterometers) and passive (radiometers), for the remote sensing of Earth’s surface. Besides brightness temperature and backscattering coefficient, microwave indices, defined as a combination of data collected at different frequencies and polarizations, revealed a good sensitivity to hydrological cycle parameters such as surface soil moisture, vegetation water content, and snow depth and its water equivalent. The differences between microwave backscattering and emission at more frequencies and polarizations have been well established in relation to these parameters, enabling operational retrieval algorithms based on microwave indices to be developed. This Special Issue aims at providing an overview of microwave signal capabilities in estimating the main land parameters of the hydrological cycle, e.g., soil moisture, vegetation water content, and snow water equivalent, on both local and global scales, with a particular focus on the applications of microwave indices.

time series analysis --- passive microwave soil moisture --- Sentinel-1 and Sentinel-2 --- Snow Depth and Snow Water Equivalent --- snow cover characteristics --- vegetation biomass --- roughness --- sea ice --- SMOS --- microwave radiometry --- soil moisture downscaling --- Vegetation Biomass --- vegetation index --- Terra MODIS --- Sentinel-1 --- Microwave Indices --- soil moisture content --- dual-frequency ratios --- SMAP --- passive microwave --- water-cloud model --- snow --- Sentinel-1 backscatter --- AMSR2 --- data fusion --- microwaves --- mountain region --- SAR --- start of season --- crops --- NDVI --- scatterometer --- Radarsat-2 --- polarization --- vegetation water content --- co-pol ratio --- active microwaves --- microwave indices --- harvest --- Microwave Radiometry --- soil moisture --- Soil Moisture Content --- snow correlation length --- radiometer --- radar --- soil scattering --- vegetation descriptor --- scale gap --- snow water equivalent

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

As computer and space technologies have been developed, geoscience information systems (GIS) and remote sensing (RS) technologies, which deal with the geospatial information, have been rapidly maturing. Moreover, over the last few decades, machine learning techniques including artificial neural network (ANN), deep learning, decision tree, and support vector machine (SVM) have been successfully applied to geospatial science and engineering research fields. The machine learning techniques have been widely applied to GIS and RS research fields and have recently produced valuable results in the areas of geoscience, environment, natural hazards, and natural resources. This book is a collection representing novel contributions detailing machine learning techniques as applied to geoscience information systems and remote sensing.

artificial neural network --- n/a --- model switching --- sensitivity analysis --- neural networks --- logit boost --- Qaidam Basin --- land subsidence --- land use/land cover (LULC) --- naïve Bayes --- multilayer perceptron --- convolutional neural networks --- single-class data descriptors --- logistic regression --- feature selection --- mapping --- particulate matter 10 (PM10) --- Bayes net --- gray-level co-occurrence matrix --- multi-scale --- Logistic Model Trees --- classification --- Panax notoginseng --- large scene --- coarse particle --- grayscale aerial image --- Gaofen-2 --- environmental variables --- variable selection --- spatial predictive models --- weights of evidence --- landslide prediction --- random forest --- boosted regression tree --- convolutional network --- Vietnam --- model validation --- colorization --- data mining techniques --- spatial predictions --- SCAI --- unmanned aerial vehicle --- high-resolution --- texture --- spatial sparse recovery --- landslide susceptibility map --- machine learning --- reproducible research --- constrained spatial smoothing --- support vector machine --- random forest regression --- model assessment --- information gain --- ALS point cloud --- bagging ensemble --- one-class classifiers --- leaf area index (LAI) --- landslide susceptibility --- landsat image --- ionospheric delay constraints --- spatial spline regression --- remote sensing image segmentation --- panchromatic --- Sentinel-2 --- remote sensing --- optical remote sensing --- materia medica resource --- GIS --- precise weighting --- change detection --- TRMM --- traffic CO --- crop --- training sample size --- convergence time --- object detection --- gully erosion --- deep learning --- classification-based learning --- transfer learning --- landslide --- traffic CO prediction --- hybrid model --- winter wheat spatial distribution --- logistic --- alternating direction method of multipliers --- hybrid structure convolutional neural networks --- geoherb --- predictive accuracy --- real-time precise point positioning --- spectral bands --- naïve Bayes