2.4 Sub-models and parameterizations 2.4.8 Heterotrophic respiration Heterotrophic respiration comes from the decomposition of carbon from the three soil/litter car- bon pools. For each carbon pool i2(1,2,3), we determine the maximum carbon loss based on the characteristic decay rate ˙Rz = 3.49·10 7;1.43·10 7;6.34·10 9 [s 1], which roughly cor- responds to the typical half-life for metabolic litter and microbes; structural litter; and slow soil organic matter determined from Bolker et al. (1998), respectively: ˙ Czi,a=R˙zi Ci MCz(Tg20, ˆ J20), (2.128) 0 10 20 30 40 50 60 70 80 90 100 −10 0 10 20 30 40 50 60 Heterotrophic respiration: relative response function ED−2.2 ϑ^ [%] Temper ature [ ° C ] Figure 2.8: Decomposition rate reduction factor (z(T,Jˆ)) as a function of temperature and soil moisture. the relative soil moisture for each layer being defined as: ˆ Jk=JJk JRe Po JRe; (2.129) andz(T,Jˆ)is a function similar to Eqn. (2.119) that reduces the decomposition rate due to tem- perature or soil moisture under extreme conditions: z(T,Jˆ) = h 1 1+e 0.24(T 291.15)i h1+e+0.60(T 318.15)i ⇥h 1 1+e 0.18(Jˆ 0.36)i h1+e+0.36(Jˆ 0.96)i . (2.130) The combined effect of moisture and temperature limiting the decomposition rates is shown in Fig. 2.8. Chapter 3 Evaluation of ED-2.2 at multiple time scales for South America 3.1 Introduction Increase in CO2 concentration and associated changes in climate are likely to have profound changes in plant community structure, dynamics and ecosystem ability to store carbon, and despite major advances in model representation of dynamic vegetations within the Earth System (Levis, 2010), there is still great uncertainty on how tropical ecosystems may respond to such changes. For example, while various synthesis studies of future climate for the tropics using dynamic global vegetation models (DGVMs) predict that CO2fertilization may be the dominant effect in large ar- eas of the tropics and subtropics (Rammig et al., 2010; Cox et al., 2013; Huntingford et al., 2013), recent results from long-term measurements of above-ground net primary productivity at La Selva (Costa Rica) did not detect such dominant effect of CO2fertilization (Clark et al., 2013). In ad- dition, several dynamic vegetation models show reasonable equilibrium conditions within forests, but results in these models tend to present sharper transitions in ecological measurements than similar observed values (e.g. see Fig. 3 of Good et al., 2011). Most DGVMs have thorough and mechanistic representation of biophysical and biogeochemical cycles, which increases realism by mechanistically representing such processes; however, they tend to describe ecosystems based on very simple characteristics, often from average properties of the ecosystem as a whole, which precludes a proper representation of the heterogeneous environment in which individuals live. In reality, the dynamics of an ecosystem is the emerging property that integrates the contribution of a system of individuals with different strategies and ability to access resources needed for their growth, survival, and reproduction (Moorcroft, 2003, 2006; Evans, 2012). In addition to the impact of individual contribution to long term dynamics, life cycles of in- dividuals, disturbance history, and anthropogenic land use change often create a heterogeneous environment, and the response of such heterogeneous landscape to relatively similar forcings can vary substantially. For example, during the LITFASS-2003 experiment in a 400 km2 area at a heterogenous rural landscape near Berlin, Beyrich et al. (2006) found large variations of sensi- ble and latent heat flux between different landscapes that included forests, croplands and lakes, but even croplands with similar characteristics showed significant differences. Also, during the EBEX-2000 experiment in California, Oncley et al. (2007) found significant variations in the av- erage sensible and latent heat fluxes over a irrigated cotton field area. In this sense, the Ecosystem Demography Model structure is advantageous because it has a mechanistic representation of the different micro-environments and the distribution of individuals living in it (Moorcroft et al., 2001, Sec. 2.2.1). Since the inception of the second generation of the ED model, there has been a signif- icant development of the biophysical and biogeochemical modules (Medvigy, 2006; Knox, 2012, and Chap. 2), and earlier versions of the model have been optimized for reproducing the biophysics for forest ecosystems at mid-latitudes (Medvigy et al., 2009) and one region in the Amazon (Kim et al., 2012). However, to this point no detailed regional evaluation of both short-term biophys- ical and physiology modules has been made for the tropics, covering a variety of regions with significantly different characteristics. Therefore, the aim of this chapter is to evaluate the ability of ED-2.2 to describe both the short- term biophysical and physiology at tropical South America, and whether this improvement in the ent levels of complexity of size, age, and functional diversity may affect the equilibrium (potential) vegetation and its trajectories towards equilibrium under the current version of the model. The eval- uation presented here is deliberately extensive and cover several different process, using different data sources or published values for references: as pointed out by Vanclay and Skovsgaard (1997), extensive evaluations are important to provide as much information on what are the main strengths and shortcomings of the current model, so new users can decide whether the model is suitable for their applications and needs. In addition to its informative value, extensive evaluation of results provides valuable information to guide future developments, particularly in process-based models, where results are outcomes of various interacting processes. Furthermore, the evaluation intended to use as much biophysical data from multiple sites at both short and long term scales. Short- term comparisons at multiple sites are important because the magnitude of different processes may vary substantially amongst sites, whereas long-term evaluations are necessary to assess whether the accumulated inaccuracies of the model processes could cause large errors in the development of plant communities under a wide range of biomes observed in the tropics, as exemplified by the distribution of above-ground biomass in South America. Unlike Medvigy et al. (2009) and Kim et al. (2012), I have not carried out any formal optimiza- tion at this point. As pointed out by Medlyn et al. (2005), in models that have too many parameters as it is the case of ED-2.2, optimizing some parameters can compensate for other nearly collinear parameters, resulting either in the model low sensitivity. In addition, in case of a process-based model with too many parameters, a full optimization of all parameters simultaneously is not viable, and choosing a subset of parameters brings the risk of drifting away parameters that directly affect a process to compensate parameters that only indirectly affect a process but were not included in the partial optimization, but nevertheless were far from optimal. One practical example would be to tune the photosynthetic capacity and stomatal conductance to improve the model accuracy of gross primary productivity, when in reality a detailed analysis would point that the issue is that the model underestimates absorption of photosynthetically active radiation by leaves. Therefore it is necessary to first carry out a thorough assessment of which processes and parameters present the main differences between model and observations, and how discrepancies observed in different processes may be intertwined, and thus identify which processes and parameters must be included in any optimization. In Sec. 3.2, I describe the simulation design for the model evaluation at short and long term dynamics, in addition to the overview of the selected sites. In Sec. 3.3, I present the main results of the multiple site comparison with site level measurements of eddy covariance fluxes, radiation, and additional site level measurements and published values from previous studies. In Sec. 3.4 I present the results from long-term simulations at multiple sites covering three different biomes in South America using both the default ED-2.2 configuration and simulations without size, age, or functional diversity structures (and combinations). In Sec. 3.5 I discuss the main results from this suite of simulations, and in Sec. 3.6 I summarize the main concluding remarks. In document Amazon Forest Response to Changes in Rainfall Regime: Results from an Individual-Based Dynamic Vegetation Model (Page 83-88)