Spatial validation reveals poor predictive performance of large-scale ecological mapping models

Ploton Pierre, Mortier Frédéric, Rejou-Mechain Maxime, Barbier Nicolas, Picard Nicolas, Rossi Vivien, Dormann Carsten F., Cornu Guillaume, Viennois Gaëlle, Bayol Nicolas, Lyapustin Alexei I., Gourlet-Fleury Sylvie, Pélissier Raphaël. 2020. Spatial validation reveals poor predictive performance of large-scale ecological mapping models. Nature Communications, 11:4540, 11 p.

Url - éditeur : https://rdcu.be/b7ofW

Quartile : Outlier, Sujet : MULTIDISCIPLINARY SCIENCES

Résumé : Mapping aboveground forest biomass is central for assessing the global carbon balance. However, current large-scale maps show strong disparities, despite good validation statistics of their underlying models. Here, we attribute this contradiction to a flaw in the validation methods, which ignore spatial autocorrelation (SAC) in data, leading to overoptimistic assessment of model predictive power. To illustrate this issue, we reproduce the approach of large-scale mapping studies using a massive forest inventory dataset of 11.8 million trees in central Africa to train and validate a random forest model based on multispectral and environmental variables. A standard nonspatial validation method suggests that the model predicts more than half of the forest biomass variation, while spatial validation methods accounting for SAC reveal quasi-null predictive power. This study underscores how a common practice in big data mapping studies shows an apparent high predictive power, even when predictors have poor relationships with the ecological variable of interest, thus possibly leading to erroneous maps and interpretations.

Mots-clés Agrovoc : forêt tropicale, écologie, technique de prévision, cartographie, modèle mathématique, modèle de simulation, qualité, défaut

Mots-clés géographiques Agrovoc : Afrique centrale

Mots-clés complémentaires : automated cartographie (en)

Mots-clés libres : Ecosystem ecology, Environmental sciences, Forest ecology, Macroecology

Classification Agris : P01 - Conservation de la nature et ressources foncières
K01 - Foresterie - Considérations générales
U30 - Méthodes de recherche

Champ stratégique Cirad : CTS 1 (2019-) - Biodiversité

Agences de financement européennes : European Commission

Programme de financement européen : H2020

Projets sur financement : (EU) ERA-NET for Monitoring and Mitigation of Greenhouse Gases from Agri- and Silvi-Culture

Auteurs et affiliations

• Ploton Pierre, IRD (FRA) - auteur correspondant
• Mortier Frédéric, CIRAD-ES-UPR Forêts et sociétés (FRA)
• Rejou-Mechain Maxime, IRD (FRA)
• Barbier Nicolas, IRD (FRA)
• Picard Nicolas
• Rossi Vivien, CIRAD-ES-UPR Forêts et sociétés (CMR) ORCID: 0000-0001-5458-1523
• Dormann Carsten F., Universität Freiburg (DEU)
• Cornu Guillaume, CIRAD-ES-UPR Forêts et sociétés (FRA) ORCID: 0000-0002-7523-5176
• Viennois Gaëlle, CNRS (FRA)
• Bayol Nicolas, Forêt ressources management (FRA)
• Lyapustin Alexei I., NASA (USA)
• Gourlet-Fleury Sylvie, CIRAD-ES-UPR Forêts et sociétés (FRA) ORCID: 0000-0002-1136-4307
• Pélissier Raphaël, IRD (FRA)

Source : Cirad-Agritrop (https://agritrop.cirad.fr/596514/)

Métriques

[ Page générée et mise en cache le 2024-12-18 ]