Mapping plant traits on the Tibetan Plateau: towards a robust upscaling framework for diverse vegetation landscapes

Abstract

Upscaling traits from plant-level measurements to grid-scale predictions is crucial for accounting for biodiversity when simulating and predicting the impacts of climate change and human activities on ecosystems at large scales. However, current trait upscaling frameworks face limitations, particularly the scarcity of trait observations. Based on a dense sampling strategy on the Tibetan Plateau, this study aims to develop a robust upscaling framework that (1) provides reliable trait predictions for this region and (2) enables analysis of how sampling density affects trait prediction.The Tibetan Plateau, known as the Roof of the World with an average elevation above 4,000 m, supports diverse zonal vegetation, both horizontally and vertically. This significant environmental and vegetation heterogeneity, combined with sparse in situ trait measurements, currently leads to high prediction uncertainty in existing global and Chinese trait maps for this region, limiting their ecological accuracy for spatial scaling on the Tibetan Plateau.Our approach toward a more robust trait upscaling includes: 1) performing standardized trait measurements on 3,961 species-level leaf samples and 504 site-level fine root samples collected from 650 sites between 2018 and 2024, covering 12 morphological and chemical traits; 2) constructing predictor sets that include bioclimate, soil, topography, and vegetation indices; 3) training machine learning models (such as random forest, boosted regression trees, and generalized additive models), using cross-validation to evaluate performance and select optimal parameters for each trait; 4) refining plant functional type (PFT) based on regional vegetation characteristics and aligning them with a detailed 10 m resolution land cover map of the Tibetan Plateau; 5) predicting traits for each PFT and aggregating them into grid-level values using PFT abundance weighting; and 6) generating a suite of 1 km resolution trait maps. We expect this work to establish a reproducible methodological framework for trait upscaling in heterogeneous landscapes, yielding more reliable trait maps for the Tibetan Plateau and providing further insight into how sampling density influences trait upscaling.

Plant phenotyping relevance

植物の形態・化学形質を測定し、機械学習とPFT加重により広域推定する再現可能な形質アップスケーリング手法の開発が中心であり、単なる生態学的測定ではない。

Code and data availability

公開論文であることは確認できましたが、現在の公式API・許可済み取得経路では本文を自動取得できませんでした。

No evidence-backed public reproduction asset is currently recorded.