offres d'emploi formations actualités contact accès annuaire intranet
Séminaires >

Contribution to the completion of univariate and multivariate times series based on Dynamic Time Warping. Following by : From the Amazon Jungle to Calais : spatial modelling of deforestation in the Guiana Shield.

Camille Dezecache, LISIC, Imap Team

jeudi 5 juillet 2018 à 13h30

B014


This talk is composed into 2 parts :

Title (1rt Part, approx. 10min) : Contribution to the completion of univariate and multivariate times series based on Dynamic Time Warping.

Abstract : I will present my contribution to the creation of two R-packages elaborated at LISIC, and within the MARCO project. DTWBI and DTWUMI enable the completion of univariate and multivariate times series respectively, based on Dynamic Time Warping.

Title (2nd part, approx. 20min) : Previous work - From the Amazon Jungle to Calais : spatial modelling of deforestation in the Guiana Shield.

Abstract : Since its appearance at the beginning of the 2000’s, the Reduction of Emissions from Deforestation and forest Degradation, known as REDD+, has gradually been admitted as an indispensable option for global policies aiming at tackling climate change. Relying on the concept of « avoided deforestation », it requires the formulation of a reference scenario aiming at evaluating efforts made by the countries involved in controlling their level of deforestation. Those scenarios are facing numerous political and scientific debates. This presentation aims at putting into perspective the principles of REDD+, based on a study site covering the Guiana Shield, one of the last large and continuous forested areas in the Tropics. In a first step, the formulation of a spatialiazed deforestation model over French Guiana will be presented, insisting on the necessity to focus on the socio-economic drivers of deforestation. This statement will be reinforced, in a second step, by the study of deforestation due to gold mining in the Guiana Shield, showing a very strong effect of gold prices and national political contexts on deforestation. These results finally contribute to a critical discussion on REDD+ mechanism, based on the formulation of very variable reference scenarios, relying more on political than on scientific determinants, and potentially enabling the emergence of environmental blackmailing.

Keywords : Spatial modelling, random forest, REDD+, deforestation