Lievin Kasanda Nkuba

Lievin Kasanda Nkuba
Université du Québec en Abitibi-Témiscamingue · Forest Research Institute (IRF)

Master of Science

About

9
Publications
2,967
Reads
How we measure 'reads'
A 'read' is counted each time someone views a publication summary (such as the title, abstract, and list of authors), clicks on a figure, or views or downloads the full-text. Learn more
19
Citations
Introduction
I am a PhD student at Laval University. My research focuses on the creation of a new approach to mapping the rivers of forested Quebec by the automated processing of massive high-resolution geospatial data produced by lidar. This study provides a geographic database of the forest hydrographic network of Quebec. My career is focused on geomatics, remote sensing and digital image processing by developing machine learning models to achieve sustainable forest management.
Additional affiliations
October 2017 - March 2018
French National Institute for Agriculture, Food, and Environment (INRAE)
Position
  • Research intern in analysis and modeling of spatial organizations and changes in soil conditions and cultivated landscapes
Description
  • The objective is to test possible strategies for estimating the uncertainty inherent in the spatial prediction of soil properties, using quantile regression forests, which are widely used in digital soil mapping, as an example of data mining algorithms.

Publications

Publications (9)
Article
It has long been acknowledged that the soil spatial samplings used as inputs to DSM models are strong drivers – and often limiting factors – of the performances of such models. However, few studies have focused on evaluating this impact and identifying the related spatial sampling characteristics. In this study, a numerical experiment was conducted...
Poster
Full-text available
L'objectif est de développer d'une part, à l’aide du lidar et de capteurs multiples, un système d’identification et de caractérisation (classification fonctionnelle) de routes forestières préexistantes à partir des variables extraites des modèles numériques de terrain (MNT) de très haute résolution spatiale et d’indices de végétation. Et de l'autre...
Conference Paper
Full-text available
It has been acknowledged for a long time that the soil spatial sampling used as input of DSM models is a strong driver-and often a limiting factor-of the performances of such models. However, few studies focused on evaluating this impact and on identifying the related spatial sampling characteristics. In this paper, a numerical experiment was condu...
Poster
Full-text available
Les routes forestières sont essentielles pour l’aménagement forestier durable, il est important pour les gestionnaires des forêts de détenir l’information adéquate du réseau routier pour une meilleure gestion dans leur prise de décision. Ce projet vise à élaborer une méthode de détection et de caractérisation automatisée du réseau routier préexista...
Presentation
Full-text available
La stratégie de communication écrite : C’est rédiger un document en fonction de la responsabilité dédiée au rédacteur et des besoins des lecteurs (Larose, 1992).
Thesis
Full-text available
Réalisée en lien avec une activité de recherche au sein de l’unité mixte de recherche (UMR) du laboratoire d’études des interactions entre sols agrosystème et hydrosystème (LISAH), cette étude porte sur l’application des algorithmes de fouille de données dans la cartographie numérique des sols sur la mise en place et tests des stratégies robustes d...
Article
Dans leur étude, Mahsa Saleh et al abordent la thématique des systèmes de prédiction de risque de feux de forêt par approche de l’apprentissage non-supervisé. L’objectif visé par leur étude est d’examiner la dynamique spatio-temporelle des feux de forêts afin de prédire le risque de feux de forêt tenant compte des variations météorologiques en se b...

Network

Cited By

Projects

Projects (2)
Archived project
This study aims to develop an automated or semi-automated detection method to identify the presence of gravel forest roads in the boreal forest. The method developed will make it possible to identify and extract the physical properties of forest roads in the boreal environment in Quebec.
Project
The aim of the course is to test possible strategies for estimating the uncertainty inherent in prediction in the geographical space of soil properties, taking as an example data mining algorithms for quantile regression (Quantile Regression). Forest) which are very widely used in digital mapping of soils. Two types of strategies will be tested: - A priori quantification of uncertainty through random forest features: (Random Forest "out of bag" error, local estimates of confidence intervals (Quantile Regression Forest). - Quantification a posteriori from a sampling of validation sites that did not participate in the calibration of the algorithm. This is done either in cross-validation or validation from new sites. A very sensitive aspect concerning the application of these two strategies is the number and the reasoned location or not of the sites used to calibrate (for the first strategy) or to validate (for the second strategy) the predictions obtained. Too few sites that are poorly positioned in space can generate significant uncertainty estimation biases. These data-limited situations are often encountered in operational digital mapping applications. We will test these aspects from a ground truth consisting of a soil surface clay rate image obtained during previous work by processing a VisPIR hyperspectral remote sensing image located in the Cap Bon region (North Tunisia ).