Editing IPCC:AR6/WGII/Chapter-2 (section)

==== 2.5.2.1 Global Overview, Assessment of Ecosystem-Level Models and Sources of Uncertainties ====

<div id="h3-33-siblings" class="h3-siblings"></div>

Shifts in terrestrial biome and changes in ecosystem processes in response to climate change are most frequently projected with dynamic global vegetation models (DGVMs) or land-surface models that form part of ESMs, which use gridded climate variables, atmospheric CO 2 concentration and information on soil properties as input variables. Since AR5, most DGVMs have been upgraded to capture carbon–nitrogen cycle interactions (e.g., ( [[#Le%20Quéré--2018|Le Quéré et al., 2018]] ), many also include a representation of wildfire and fire–vegetation interactions ( [[#Rabin--2017|Rabin et al., 2017]] ) and a small number now also account for land management (e.g., wood removal from forests and crop fertilisation harvest of irrigation ( [[#Arneth--2017|Arneth et al., 2017]] ). Other forms of disturbance, such as tree mortality, in response to, for example, episodic weather extremes or insect pest outbreaks, are relatively poorly represented or not at all, although they demonstrably impact calculated carbon cycling ( [[#Pugh--2019a|Pugh et al., 2019a]] ). Simulated biome shifts are generally in agreement in projecting broad patterns on a global scale but vary greatly regarding the simulated trends in historical and future carbon uptake or losses, both regionally and globally ( [[#Chang--2017|Chang et al., 2017]] ; [[#Canadell--2021|Canadell et al., 2021]] ).

Similar to other models, models to project large-scale changes in vegetation and ecosystem processes have to deal with structural uncertainty (associated with the choice and the representation of processes in models), input-data uncertainty (associated with variability in initial conditions and parameter values) and error propagation (associated with coupling models) ( [[#Rounsevell--2019|Rounsevell et al., 2019]] ). The IPBES methodological assessment report on scenarios and models of biodiversity and ecosystem services provides a comprehensive overview over the relevant issues ( [[#Ferrier--2016|Ferrier et al., 2016]] ).

In order to assess performance, most models have been individually evaluated against a range of observations. Moreover, in the annual updates of the global carbon budgets, a model has to meet a small set of basic criteria to have its output included ( [[#Le%20Quéré--2018|Le Quéré et al., 2018]] ). More systematic benchmarking approaches have also been proposed that utilise a range of different datasets ( [[#Kelley--2013|Kelley et al., 2013]] ; [[#Chang--2017|Chang et al., 2017]] ) to assess multiple simulated processes. These methods, in principle, facilitate assigning quality scores to models based on their overall performance ( [[#Kelley--2013|Kelley et al., 2013]] ). So far, this scoring does not yet allow a clear quality ranking of models, since individual DGVMs tend to score well for some variables and badly for others. A recent comparison of global fire–vegetation model outputs was also able to clearly identify outliers when using a formalised benchmarking and scoring approach ( [[#Hantson--2020|Hantson et al., 2020]] ). However, benchmarking does not address sources of uncertainty and it would be advisable to perform ‘perturbed-physics’ experiments, in which multiple model parameters are varied in parallel more frequently as a means to test parameter-value uncertainty ( [[#Wramneby--2008|Wramneby et al., 2008]] ; [[#Booth--2012|Booth et al., 2012]] ; [[#Lienert--2018|Lienert and Joos, 2018]] ).

Species diversity impacts ecosystem functioning and hence ecosystem services ( [[#Hooper--2012|Hooper et al., 2012]] ; [[#Mokany--2016|Mokany et al., 2016]] ). So far, however, integrated modelling of ecosystem processes and biodiversity across multiple trophic levels and food webs is in its infancy ( [[#Harfoot--2014|Harfoot et al., 2014]] ). Whether or not the enhanced integration of state, function and functional diversity across multiple trophic levels in models will markedly alter projections of how ecosystems respond to climate change thus remains an open research question.

Beyond dynamic simulation of biome shifts and carbon cycling, which are important aspects of climate regulation, DGVMs can also provide information on a number of variables closely linked to other ecosystem services such as water availability, air quality or food provisioning ( [[#Krause--2017|Krause et al., 2017]] ; [[#Rabin--2020|Rabin et al., 2020]] ). However, they are not intended to provide a comprehensive assessment of ecosystem services. For these, other approaches applied but, to date, these are mostly applied on regional scales and are only weakly dynamic ( [[#Ferrier--2016|Ferrier et al., 2016]] ).

<div id="2.5.2.2" class="h3-container"></div>

<span id="projected-changes-globally-at-the-biome-level"></span>