TY - JOUR

T1 - Allometric equations to estimate the dry mass of Sahel woody plants mapped with very-high resolution satellite imagery

AU - Hiernaux, Pierre

AU - Issoufou, Hassane Bil Assanou

AU - Igel, Christian

AU - Kariryaa, Ankit

AU - Kourouma, Moussa

AU - Chave, Jérôme

AU - Mougin, Eric

AU - Savadogo, Patrice

N1 - Funding Information: The authors thank Jorge Pinzon and Robert Kaufmann for the fruitful discussions about data heteroskedasticity and the generalized least square (GLS) application that allowed to discuss the uncertainty estimates and facilitate the choice among allometry models. The authors thank Compton J. Tucker for his advices and help in editing. The authors are also grateful to late Mohamed Idrissa Cissé and ILRI staff in Niono, in the Gourma and in Niger who contributed to the field data collection under research projects of the International Livestock Research Institute, ILRI. More recent observations were funded by the AMMA-CATCH observatory (http://www.amma-catch.org). C.I. and A.K. acknowledge support by the Villum Foundation through the project Deep Learning and Remote Sensing for Unlocking Global Ecosystem Resource Dynamics (DeReEco). Funding Information: The authors thank Jorge Pinzon and Robert Kaufmann for the fruitful discussions about data heteroskedasticity and the generalized least square (GLS) application that allowed to discuss the uncertainty estimates and facilitate the choice among allometry models. The authors thank Compton J. Tucker for his advices and help in editing. The authors are also grateful to late Mohamed Idrissa Cissé and ILRI staff in Niono, in the Gourma and in Niger who contributed to the field data collection under research projects of the International Livestock Research Institute, ILRI. More recent observations were funded by the AMMA-CATCH observatory (http://www.amma-catch.org). C.I. and A.K. acknowledge support by the Villum Foundation through the project Deep Learning and Remote Sensing for Unlocking Global Ecosystem Resource Dynamics (DeReEco). P.H, B.-A. H. I. M.K.: contributed to data collection; P.H. C.I. A.K.: participated to data analysis; P.H. wrote the first draft all authors contributing to revisions and editing. Publisher Copyright: © 2022 Elsevier B.V.

PY - 2023/2/1

Y1 - 2023/2/1

N2 - Very high-resolution satellite images and deep learning are achieving the mapping of individual trees and shrubs over large areas in Africa. Each woody plant is precisely georeferenced and defined by its crown area and, sometimes, its height. The challenge is to build allometric equations for foliage, wood and root dry masses based either on crown area alone, or the product crown area × tree height as independent variables regardless of species. This was met by reanalyzing existing Sahel woody plant data from destructive sampling. Overall, the foliage (seasonal maximum), wood and root dry masses were measured on 900, 698 and 26 trees or shrubs from 27, 26 and 5 species respectively. The regression models tested for foliage, wood or root masses were linear or power functions using a range of different approaches: Ordinary least square (OLS) regression after log–log transform with or without Baskerville correction (1972), and non-linear regressions (NLR). All the models outputs were intercompared for fit indicators and prediction uncertainty, as well as the resulting trends in root to wood ratio, and foliage to wood ratio over the range of crown areas. This process selected a set of OLS log–log equations with crown area as independent variable. When the Baskerville correction is applied the approximation errors are 0.4 kg Dry Matter (DM) for foliage, 12.2 for wood and 6.3 for roots, aggregating at 13.8 kg DM per tree. These allometry equations were compared to published equations for tropical trees, most from the more humid tropics that were generally based on stem diameter, tree height and wood density. This paper's allometry predictions were within the range of these other allometry equations, reinforcing the confidence in their use even beyond the Sahel domain towards sub-humid savannas.

AB - Very high-resolution satellite images and deep learning are achieving the mapping of individual trees and shrubs over large areas in Africa. Each woody plant is precisely georeferenced and defined by its crown area and, sometimes, its height. The challenge is to build allometric equations for foliage, wood and root dry masses based either on crown area alone, or the product crown area × tree height as independent variables regardless of species. This was met by reanalyzing existing Sahel woody plant data from destructive sampling. Overall, the foliage (seasonal maximum), wood and root dry masses were measured on 900, 698 and 26 trees or shrubs from 27, 26 and 5 species respectively. The regression models tested for foliage, wood or root masses were linear or power functions using a range of different approaches: Ordinary least square (OLS) regression after log–log transform with or without Baskerville correction (1972), and non-linear regressions (NLR). All the models outputs were intercompared for fit indicators and prediction uncertainty, as well as the resulting trends in root to wood ratio, and foliage to wood ratio over the range of crown areas. This process selected a set of OLS log–log equations with crown area as independent variable. When the Baskerville correction is applied the approximation errors are 0.4 kg Dry Matter (DM) for foliage, 12.2 for wood and 6.3 for roots, aggregating at 13.8 kg DM per tree. These allometry equations were compared to published equations for tropical trees, most from the more humid tropics that were generally based on stem diameter, tree height and wood density. This paper's allometry predictions were within the range of these other allometry equations, reinforcing the confidence in their use even beyond the Sahel domain towards sub-humid savannas.

KW - Africa drylands

KW - Allometry

KW - Foliage

KW - Root

KW - Sahel

KW - Sudan

KW - Wood

UR - http://www.scopus.com/inward/record.url?scp=85142708771&partnerID=8YFLogxK

U2 - 10.1016/j.foreco.2022.120653

DO - 10.1016/j.foreco.2022.120653

M3 - Journal article

AN - SCOPUS:85142708771

VL - 529

JO - Forest Ecology and Management

JF - Forest Ecology and Management

SN - 0378-1127

M1 - 120653

ER -