ESD Reviews: Evidence of multiple inconsistencies between representations of terrestrial and marine ecosystems in Earth System Models

Terrestrial and marine ecosystems interact with other Earth system components through different biosphere-climate feedbacks that are very similar among ecosystem types. Despite these similarities, terrestrial and marine systems are often treated relatively separately in Earth System Models (ESM). In these ESM, the ecosystems are represented by a set of biological processes that are able to influence the climate system by affecting the chemical and physical properties of the environment. While most of the climate-relevant processes are shared between ecosystem types, model representations of terrestrial and 5 marine ecosystems often differ. This raises the question whether inconsistencies between terrestrial and marine ecosystem models exist and potentially skew our perception of the relative influence of each ecosystem on climate. Here we compared the terrestrial and marine modules of 17 Earth System Models in order to identify inconsistencies between the two ecosystem types. We sorted out the biological processes included in ESM regarding their influence on climate into three types of biosphereclimate feedbacks (i.e. the biogeochemical pumps, the biogeophysical mechanisms and the gas and particle shuttles), and 10 critically compare their representation in the different ecosystem modules. Overall, we found multiple evidences of unjustified differences in process representations between terrestrial and marine ecosystem models within ESM. These inconsistencies may lead to wrong predictions about the role of biosphere in the climate system. We believe that the present comparison can be used by the Earth system modeling community to increase consistency between ecosystem models. We further call for the development of a common framework allowing the uniform representation of climate-relevant processes in ecosystem modules 15


Introduction
Terrestrial and marine ecosystems have been mostly independently studied (Steele, 1991;Raffaelli et al., 2005;Menge et al., 2009). The separate development of terrestrial and marine ecosystem models led to important differences in their conceptual-20 The three mechanisms described above are able to affect climate in different directions by either buffering or accentuating changes in the climatic conditions. For instance, contemporary climate change affects biological processes which can reciprocally accelerate (i.e positive feedback) or dampen (i.e. negative feedback) climate change. A good example of a negative feedback mechanism is enhanced vegetation carbon uptake following the increase in atmospheric CO 2 concentration that may buffer human induced CO 2 emissions and thus surface temperature (Canadell and Raupach, 2008). On the other hand, the 130 increase in temperature is expected to enhance soil bacterial metabolism and may therefore increase natural CO 2 and methane production by micro-organisms (Montzka et al., 2011;Dean et al., 2018), increasing further more atmospheric temperature (i.e. positive feedback). These two examples highlight the central role of the biosphere on climate regulation and the necessity to accurately represent biosphere-climate feedbacks on both land and ocean to predict future (and past) climate change.

135
To compare terrestrial and marine ecosystem models, we listed all the climate-relevant processes that are involved in the biosphere-climate feedbacks described in the previous section. We focused on the processes that were present at least in one of the 15 terrestrial ecosystem models and 16 marine ecosystem models we reviewed. Based on this list, we compared how the different biosphere-climate feedbacks are considered in terrestrial and marine ecosystem modules of current ESM (Table 2).
Overall, inconsistencies exist among terrestrial and marine ecosystem models of ESM regarding the biogeochemical, the 140 biogeophysical and, to a lesser extent, the gas and particle shuttle mechanisms (Table 2, Fig. 2). In the next paragraphs, we describe these inconsistencies among ecosystem modules in ESM by providing a detailed comparison of each process, as well as its potential influence on -and its response to -climate.

Biogeochemical processes
The biogeochemical pump, including the carbon cycle, is the mechanism represented with the most diverse processes in ter-145 restrial and marine modules of Earth System Models. However, terrestrial and marine models show important differences (Table 2, Fig. 2), due to the way they were initially developed. Because terrestrial ecosystem models were initially based on the carbon cycle, the processes of photosynthesis, respiration, phenology, mortality and soil respiration (i.e. C remineralization in Table 2) are represented in the 15 terrestrial modules. The net primary production is calculated in each model from the balance of photosynthesis and respiration, each of these processes being represented by a set of physiological traits interacting with the 150 environment ( Fig. 1).
Besides carbon, the representation of other important elements such as nitrogen, phosphorus and iron in terrestrial ecosystem models is rather sparse compared to marine models, while the different ecosystem types are similar in terms of N and P limitation (Elser et al., 2007). Nitrogen is an essential element of protein, including the photosynthetic enzymes such as Rubisco, the enzyme that fixes carbon from the atmosphere into carbohydrates. Phosphorus is also important for plant physiology as it 155 5 https://doi.org/10.5194/esd-2020-55 Preprint. Discussion started: 3 August 2020 c Author(s) 2020. CC BY 4.0 License. is part of the composition of nucleid acid, lipid and bioenergetic molecules such as ATP (Wright et al., 2004). For that reason, the terrestrial modules used for the CMIP6 experiments include nitrogen limited carboxylation capacity of the plants. The last generation of land surface models incorporate a representation of the phosphorous cycle as well (Fisher and Koven, 2020).
However, these land surface models are not integrated in ESM and the representation of the phosphorus cycle is still needed in most of the terrestrial ecosystem models within ESM.
In contrast, the development of marine ecosystem models was initially based on the cycles of limiting nutrients for phytoplankton growth. In the simplest models, only one element was explicitly modeled and the concentrations of others such as carbon were calculated using a fixed ratio (i.e. the Redfield ratio, Redfield, 1934). However, the ratios of the main nutrients and elements (N, C, P, Si, Fe) vary among phytoplankton groups and with the environmental conditions (Rhee.G.Yull, 1978;Goldman et al., 1979;Geider and La Roche, 2002;Quigg et al., 2003). Therefore, the use of a fixed ratio among elements 165 constrains the predictions within a narrow range of potential ecosystem responses that may not represent the current states of marine ecosystems. Some recent marine ecosystem models (e.g. BFMv5.2, Vichi et al., 2015) thus included variable ratios that are influenced by the environmental conditions (i.e. the availability of the different elements in the environment). Nevertheless, the explicit carbon cycle representation is still lacking in most of the models and the processes of photosynthesis and cell respiration are ignored (Table 2, Fig. 1).

170
Another difference between terrestrial and marine models in Earth System Models is the representation of trophic interactions (Table 2, Fig. 2). In terrestrial models, only plants and soil microorganisms are represented while higher trophic levels are ignored. Although it has been traditionally assumed that trophic interactions had a limited effect on large scale climate, recent studies underline the potential strong influence of grazers and higher trophic levels on ecosystem carbon uptake and storage (e.g. Schmitz et al., 2018). For instance, the disturbance induced by elephants enhance aboveground biomass, and thus carbon 175 storage, in African tropical forest (Berzaghi et al., 2019). Grazers may also influence climate by changing the biogeophysical properties of the ecosystem. For instance, larger animals grazing on boreal ecosystem limit shrub height and density, cooling down air temperature by increasing summer albedo (Te Beest et al., 2016), or protect the permafrost from thawing (Beer et al., 2020).
Most of the marine models reviewed here considered zooplankton, heterotrophic organisms that feed on primary producers.

180
Zooplankton has been implemented in marine ecosystem models because they exert a strong grazing pressure on phytoplankton, impeding or buffering phytoplankton bloom formation (Prowe et al., 2012). Zooplankton and higher trophic levels in general, play also an important role in carbon removal (Davison et al., 2013;Steinberg and Landry, 2017), nutrient distribution and recycling (Vanni, 2002;Schmitz et al., 2010). However, the representation of the trophic chains in marine models remains very simple. Complex trophic interactions can strongly influence carbon storage by marine ecosystems. For instance, Wilmers et al. 185 (2012) showed that the presence of sea otters increases the carbon fixation by kelp by a factor of 10 due to their predation pressure on kelp grazers. A better understanding of the importance of trophic complexity and length on carbon cycle and on climate is needed to properly judge the necessity to include them in Earth System Models.
There are also several justified differences between terrestrial and marine ecosystem which are represented in ESM (Table 2).
Terrestrial and marine ecosystem models differ in the representation of processes involved in the water cycle and silicate cycle.

190
The role of the biosphere on the carbon cycle is central and has been considered in terrestrial and marine models of ESM. Both ecosystem models are based on the growth of photosynthetic primary producers (i.e. net primary production, NPP). However, important differences exist in the representation of growth and its implication for the carbon cycle among ecosystem types.
In terrestrial ecosystem models, NPP corresponds to the difference between the carbon fixed by photosynthesis and the carbon released by respiration.
Photosynthesis has been modeled using three different approaches in terrestrial ecosystem models (Arora, 2002;Fisher et al., 2014): the biochemical approach, the light use efficiency approach and the carbon assimilation approach. Most of the modules reviewed here followed the biochemical approach to represent photosynthesis (a). In this approach, the rate of CO 2 assimilation is limited by i) the rate of carboxylation by Rubisco, ii) electron transport which depends on light and iii) the transport of photosynthetic products (Farquhar et al., 1980;Collatz et al., 1992). Note that the transport of photosynthetic products is sometimes ignored in the present modules (a). In the biochemical approach, the assimilation of carbon by photosynthesis is closely linked to stomatal conductance that control intercellular CO 2 concentration and water exchange with the atmosphere. A part of the carbon fixed by photosynthesis is re-emitted to the atmosphere through respiration. In all the models, respiration is divided into maintenance respiration and growth respiration. The growth respiration is a fixed proportion of the NPP, while the maintenance respiration, considered at the organ level (i.e. leaf, stem, root), can depend on temperature, nitrogen content and the rate of carboxylation (a).
In marine models (b), the growth of primary producers ignore the processes of photosynthesis and respiration. NPP depends on light, nutrients and temperature (b). Most of the modules do not explicitly consider carbon, and thus deduce carbon assimilation by applying a stochiometric ratio. For that reason, respiration cannot be properly estimated because the ratio of elements in phytoplankton is considered constant. Only one model that explicitly represents the carbon cycle accounts for respiration by phytoplankton (BFMv5.2).  These differences rely on biological singularity of terrestrial and marine ecosystems. Water is an essential and limiting element for terrestrial plant growth and is tightly linked to carbon uptake. Similarly, silicon is an important constituent of the shell of an abundant phytoplankton group (i.e. the diatoms). Diatoms are often considered in marine ecosystem models because their silicate shell influence particle sinking from the surface to the deep ocean, and thus influence carbon storage in the ocean.
For these reasons, these differences among ecosystem models are not part of the inconsistencies we identified in the previous 195 paragraphs.

Biogeophysical processes
The biogeophysical influence of the biosphere is unequally considered in terrestrial and marine models of Earth system models (  in the ocean (i.e. surface albedo, air-sea gas exchange, turbulent viscosity and vertical mixing) are more difficult to evaluate because they are less studied and will thus not be discussed further.

Gas and particles emission processes
Terrestrial and marine ecosystem models are rather similar in their representation of gas and particle shuttles. Both ecosystem modules only sparsely consider the role of the biosphere on greenhouse gas and aerosol atmospheric concentration, except 220 CO 2 and water vapor in terrestrial ecosystem models (Table 2, Fig. 2).
The ecosystem models mainly differ with regard to CO 2 emissions. While CO 2 emission back to the atmosphere through plant and soil respiration is well considered in terrestrial ecosystem models, the influence of marine biota on dissolved and  atmospheric CO 2 is poorly represented (Table 2, Fig. 2). The great majority of the models do not explicitly consider the processes of photosynthesis and respiration (Fig. 1). We now know that the reaction norm of photosynthesis and respiration to 225 9 https://doi.org/10.5194/esd-2020-55 Preprint. Discussion started: 3 August 2020 c Author(s) 2020. CC BY 4.0 License. temperature are different and that organisms adapt and both rates change with higher temperature (Padfield et al., 2016;Schaum et al., 2017;Barton et al., 2020). The resulting net exchange of carbon between atmosphere and marine biota is thus sensitive to temperature. By ignoring photosynthesis and respiration processes, marine models may fail to properly predict the influence of phytoplankton on the carbon cycle, and thus on atmospheric CO 2 concentration under changing climatic conditions. CH 4 emissions are considered only in 5 of the 15 terrestrial modules and in none of the marine modules (Table 2, Fig. 2).

230
Knowing that the large majority of CH 4 is produced by terrestrial ecosystems (Dean2018) and that soil production is predicted to increase under future conditions (Montzka et al., 2011;Dean et al., 2018), CH 4 emissions in terrestrial modules of ESM are still underrepresented. Ocean methane production is assumed to be small compared to terrestrial ones (Valentine, 2011), thus the absence of methane emission in marine ecosystem models of current ESM may be justified. Nevertheless, the CH 4 emission by coastal ecosystems might also grow under future climatic conditions (Al-Haj and Fulweiler, 2020) and further 235 evaluation of their role in global CH 4 emissions will be needed to judge the necessity to include them in ESM.
N 2 O production is represented in a quarter of the terrestrial and marine modules of current ESM (Table 2, Fig. 2). While the estimations of N 2 O emissions are still rather uncertain (from 3,3 to 9 Tg N y −1 for terrestrial ecosystems and from 1,8 to 9,45 Tg N y −1 for marine ecosystems, Ciais et al., 2014), it has been shown that N 2 O plays an important role in the past and ongoing climate change (Schilt et al., 2010;Stocker et al., 2013). Furthermore, climate change is affecting biologically mediated N 2 O 240 emissions (Stocker et al., 2013;Martinez-Rey et al., 2014) and also the overall impact of N 2 O on climate. Increasing temperature may enhance the transport of N 2 O from its source location (Earth surface) to its sink location (stratosphere), reducing both the lifetime of N 2 O and its global warming potential (Kracher et al., 2016). ESM represent the adequate tools to study such complex feedbacks between biosphere, atmosphere and climate. However, it necessitates a good representation of the processes behind N 2 O emissions that is currently missing in many current ESM (in both terrestrial and marine modules and 245 despite their consistency regarding this particular process).
The emission of aerosols is represented in the form of biogenic volatile organic compound (BVOC, 3/15 modules) in terrestrial ecosystems and dimethyl sulphide (DMS, 6/16 modules) in marine ones (Table 2, Fig. 2). The fact that there is consistent evidence that DMS decreases atmospheric temperature (e.g. Charlson et al. (1987);McCoy et al. (2015) but see Quinn and Bates (2011)) may explain the slightly higher number of marine modules representing volatile production than terrestrial ones Similarly, BVOC production is sparsely presented in terrestrial modules of ESM (Table 2, Fig. 2). Current knowledge indicates that the influence of terrestrial aerosols on climate is highly variable and depends on the type of emitted substances 255 (see Peñuelas and Staudt, 2010). Nevertheless, a recent study estimated that aerosol-climate feedbacks could be strong enough to moderate the CO 2 -related atmospheric temperature increase . Further studies are needed to identify the most important aerosols emitted by terrestrial ecosystems in order to facilitate their integration in ESM. The future inclusion of BVOC emission in ESM may be further considered knowing that the production of BVOC by vegetation is predicted to increase under climate change (Laothawornkitkul et al., 2009;Peñuelas and Staudt, 2010;Zhao et al., 2017).

260
The present review highlights important inconsistencies in the representation of the main biosphere-climate feedbacks between terrestrial and marine ecosystem models within ESM. The main processes related to the carbon cycle (i.e. photosynthesis and respiration) are still implicitly represented in marine models, while nutrient limitation of growth is only scarcely considered in terrestrial ones. Major differences also occur in the representation of biophysical mechanisms. Light absorption only is partially 265 considered in marine ecosystem models while terrestrial ecosystem models represent adequately the influence of vegetation on land surface biogeophysics (i.e. albedo, roughness length, evapotranspiration and light absorption). Conversely, emission of non-CO 2 gases and particles by the biosphere into the atmosphere is consistently represented, even if scarcely, in terrestrial and marine modules of ESM.
Inconsistencies in ecosystem representation can lead to wrong predictions about the role of the biosphere in the climate 270 system. The relative importance of terrestrial and marine ecosystems on climate regulation might be inaccurately perceived because predictions result from different models representing various processes with different complexity levels. As an example, terrestrial ecosystems models currently include a wider range of biosphere-climate feedbacks than marine models (Fig. 2), leading to the potential overweighting of the influence of terrestrial ecosystem on climate compared to marine ones. Future predictions could be further biased by the lack of mechanistic representation of particular processes in one or the other ecosystem 275 model. While human activities are predicted to bring atmospheric green-house gas concentration above 750ppm CO 2 equivalent and temperature above +4 • C in the worst-case scenario (RCP8.5, IPCC, 2014), the response of biosphere to these changes and the subsequent feedbacks on climate can be underestimated in marine models due to the lack of explicit representation of photosynthesis and respiration (Fig. 1). Conversely, the predicted increase of CO 2 uptake by terrestrial ecosystem following the increase in atmospheric CO 2 and temperature might be overestimated due to the lack of representation of nutrient limitation 280 on plant growth (Zaehle et al., 2015;Wieder et al., 2015;Terrer et al., 2019).
Nevertheless, there is the potential to reduce inconsistencies in the representation of terrestrial and marine ecosystems in ESM in the near future. We believe that more collaboration among terrestrial and marine scientific communities can strongly improve the representation of the biosphere in ESM. By identifying the major inconsistencies that currently exist among ecosystem modules in ESM, the present work provides a solid basis toward future consistency in biosphere representation. Altogether, we argue that increasing consistency between ecosystem modules of ESM enables to extend their ability to predict future climate. While human activities strongly impact the biosphere, improving our understanding and representation 295 of the biosphere-climate feedbacks in both terrestrial and marine ecosystem models is crucial to make reliable predictions and build efficient management policies.
Author contributions. All authors designed the study. FP reviewed the ESM and wrote the first draft of the manuscript. All authors contributed substantially to revisions.  Table 1. List of the 17 earth system models reviewed in this study. For each ESM, the names of the terrestrial ecosystem module and the marine ecosystem module are indicated as well as the CMIP phase for which the model has been applied and the main references. A total of 15 different terrestrial ecosystem models and 16 marine ecosystem models are considered.