Potential of global vegetation maps in capturing xerophytic vegetation cover: insights from Madagascar's arid ecosystems

IF 2.6 3区环境科学与生态学 Q2 ECOLOGY

Journal of Arid Environments Pub Date : 2025-05-27 DOI:10.1016/j.jaridenv.2025.105406

Felana Nantenaina Ramalason , Olivia Lovanirina Rakotondrasoa , Arthur Vander Linden , Guillaume Renard , Harifidy Rakoto Ratsimba , Jan Bogaert , Jean-François Bastin

{"title":"Potential of global vegetation maps in capturing xerophytic vegetation cover: insights from Madagascar's arid ecosystems","authors":"Felana Nantenaina Ramalason , Olivia Lovanirina Rakotondrasoa , Arthur Vander Linden , Guillaume Renard , Harifidy Rakoto Ratsimba , Jan Bogaert , Jean-François Bastin","doi":"10.1016/j.jaridenv.2025.105406","DOIUrl":null,"url":null,"abstract":"<div><div>Global vegetation maps are essential tools for ecosystem monitoring, yet they often fail to accurately represent vegetation in arid regions. This study evaluates the performance of four global products—GFCC, GFW, CGLS-LC100, and DW—in the xerophytic shrublands of southwestern Madagascar, a region undergoing rapid deforestation. Satellite-derived estimates were compared with field data collected in 2022 from 41plots (900 m<sup>2</sup> each) using the point-intercept method. Results revealed substantial discrepancies among products, with a coefficient of variation reaching 78.94 % for tree cover, primarily due to differences in vegetation definitions and classification methods. Aggregating tree and shrub cover into a single woody cover layer significantly reduced this variability by 58.19 %. Among the products tested, DW provided the most accurate estimates of woody cover (R<sup>2</sup> = 0.70), while CGLS-LC100 was the most responsive to floristic composition. Product biases varied by vegetation type: DW and CGLS-LC100 tended to overestimate tree cover in structurally developed plots, while GFCC and GFW underestimated it in multi-stemmed thickets. Shrub cover was consistently underestimated across all products. In this context of active forest loss, overestimations may reflect past landscape conditions, while underestimations expose structural limitations. These findings highlight the need for mapping strategies that better capture the complexity of dryland vegetation.</div></div>","PeriodicalId":51080,"journal":{"name":"Journal of Arid Environments","volume":"229 ","pages":"Article 105406"},"PeriodicalIF":2.6000,"publicationDate":"2025-05-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Arid Environments","FirstCategoryId":"93","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0140196325000904","RegionNum":3,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"ECOLOGY","Score":null,"Total":0}

引用次数: 0

Abstract

Global vegetation maps are essential tools for ecosystem monitoring, yet they often fail to accurately represent vegetation in arid regions. This study evaluates the performance of four global products—GFCC, GFW, CGLS-LC100, and DW—in the xerophytic shrublands of southwestern Madagascar, a region undergoing rapid deforestation. Satellite-derived estimates were compared with field data collected in 2022 from 41plots (900 m² each) using the point-intercept method. Results revealed substantial discrepancies among products, with a coefficient of variation reaching 78.94 % for tree cover, primarily due to differences in vegetation definitions and classification methods. Aggregating tree and shrub cover into a single woody cover layer significantly reduced this variability by 58.19 %. Among the products tested, DW provided the most accurate estimates of woody cover (R² = 0.70), while CGLS-LC100 was the most responsive to floristic composition. Product biases varied by vegetation type: DW and CGLS-LC100 tended to overestimate tree cover in structurally developed plots, while GFCC and GFW underestimated it in multi-stemmed thickets. Shrub cover was consistently underestimated across all products. In this context of active forest loss, overestimations may reflect past landscape conditions, while underestimations expose structural limitations. These findings highlight the need for mapping strategies that better capture the complexity of dryland vegetation.

查看原文本刊更多论文

全球植被图在捕捉旱生植被覆盖方面的潜力：来自马达加斯加干旱生态系统的见解

全球植被图是生态系统监测的重要工具，但它们往往不能准确地反映干旱地区的植被。本研究评估了四种全球产品gfcc、GFW、CGLS-LC100和dw4在马达加斯加西南部干生灌丛中的表现，该地区正经历着快速的森林砍伐。利用点截法将卫星估算值与2022年从41个地块（每个地块900平方米）收集的实地数据进行了比较。结果表明，不同产品间存在较大差异，树木覆盖度变异系数达78.94%，主要是由于植被定义和分类方法的差异。将乔灌木覆盖层聚集成单一木本覆盖层，显著降低了这一变异率58.19%。DW对树木覆盖度的估计最准确（R2 = 0.70），而CGLS-LC100对植物区系组成的反应最灵敏。不同植被类型的产品偏差不同：DW和CGLS-LC100在结构发达的样地倾向于高估树木覆盖度，而GFCC和GFW在多茎灌丛中倾向于低估树木覆盖度。在所有产品中，灌木覆盖一直被低估。在森林活跃损失的背景下，高估可能反映了过去的景观条件，而低估则暴露了结构限制。这些发现强调了更好地捕捉旱地植被复杂性的制图策略的必要性。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Journal of Arid Environments 环境科学-环境科学

CiteScore

5.70

自引率

3.70%

发文量

144

审稿时长

55 days

期刊介绍： The Journal of Arid Environments is an international journal publishing original scientific and technical research articles on physical, biological and cultural aspects of arid, semi-arid, and desert environments. As a forum of multi-disciplinary and interdisciplinary dialogue it addresses research on all aspects of arid environments and their past, present and future use.