Performance of the ecosystem demography model (EDv2.2) in simulating gross primary production capacity and activity in a dryland study area
Introduction
Dryland vegetation plays an important role in the global carbon budget, including regulating the global land carbon sink (Ahlström et al., 2015; Metcalfe, 2014; Poulter et al., 2014). Vegetation dynamics in drylands are largely a function of spatial and temporal variability in climate. For example, characteristic vegetation structure, composition, and function (e.g. photosynthesis) can differ markedly between lower and higher elevations in drylands, largely in response to corresponding differences in precipitation and temperature. Climate variability can also drive long-term changes in vegetation dynamics in dryland ecosystems. For example, climate change in the drylands of the Great Basin, US, may lead to a shift from winter snow to rain-dominated precipitation regimes, which in turn may favor fire-prone invasive species, such as cheatgrass (Bromus tectorum), that can convert native shrub-steppe communities to exotic annual grasslands (Concilio et al., 2013; Polley et al., 2013; Scott et al., 2015). These changes in structure and function ultimately affect ecosystem-scale vegetation productivity. Indeed, studies have shown that up to 60 percent of the global carbon sink anomaly can be explained by vegetation dynamics in dryland ecosystems (Ahlström et al., 2015; Poulter et al., 2014). This carbon sink variability is mostly associated with changes in gross primary production (GPP) (Yao et al., 2020). Thus, modelling the spatial and temporal dynamics of GPP in drylands is essential for global-scale studies on carbon balance and atmospheric CO2.
GPP represents ecosystem-scale apparent photosynthesis and is a primary indicator of the vegetation state of an ecosystem. GPP can be assessed in terms of capacity (i.e., amount) and activity (i.e., dynamics) (Medvigy et al., 2013; Smith et al., 2018). Estimation of GPP at the ecosystem scale in drylands is helpful in understanding food and fiber availability, livestock grazing resources, and long-term processes related to the global carbon cycle (Ryu et al., 2019). However, there is a poor understanding of GPP in drylands due to variable hydrometeorological conditions (e.g., along an elevation gradient) and different plant functional types and photosynthetic pathways (e.g., C3 vs. C4) (Yan et al., 2019), among other factors. Direct measurement of GPP is a challenging task (Ryu et al., 2019; Yan et al., 2019) and commonly used products are based on data from eddy covariance (EC) towers, remote sensing, and dynamic global vegetation models (DGVMs) (for the latest review refer to Ryu et al., 2019). EC data provide the most reliable estimates of GPP capacity; however, in drylands the spatial distribution of EC towers is limited, and their time series may not be long enough to represent GPP dynamics. Remote sensing-based GPP derived from space-based sensors, such as MODIS (Running et al., 2004), provides long-term estimates that can be used for analysis of photosynthetic activity; however, GPP derived from remotely sensed data may be under- or overestimated depending on the ecosystem (Stocker et al., 2019; Verma et al., 2014), thus limiting our understanding of GPP capacity.
Process-based DGVMs are important tools to study GPP capacity and activity that can provide complementary observations to EC tower and remote sensing data. These models can provide simulations of photosynthesis at the leaf scale, as well as at the canopy and ecosystem scales. A wealth of studies have implemented DGVMs in different ecosystem types (see review in Fisher et al., 2018) to estimate GPP from local (EC towers) to regional scales. An evaluation of the DGVMs’ performance prior to implementation should take place and include model parameterization, sensitivity analysis, calibration, and validation (Fer et al., 2018; Keenan et al., 2013; Kuppel et al., 2012; Pandit et al., 2019a; Post et al., 2017; Renwick et al., 2019; Santaren et al., 2007; Wang et al., 2001). However, there is an information gap in the evaluation of DGVMs in drylands and, more specifically, in dryland regions where vegetation productivity rapidly changes, especially across elevation gradients. Many studies have investigated the correlation between simulated GPP and GPP estimated from EC towers or remote sensing (Antonarakis et al., 2014; Pandit et al., 2019a; Renwick et al., 2019; Trugman et al., 2016). Comparing simulated GPP of drylands with EC towers is largely applicable to assessing GPP capacity rather than activity due to limited years of EC data. To fully capture GPP activity more specific metrics are required. Phenometrics, such as start of season (SOS) and end of season (EOS), and long-term trend analyses are examples of criteria that can be used for studying GPP activity (Chen et al., 2016; Forkel et al., 2015; Zhao et al., 2019).
The focus of this paper is on the Ecosystem Demography model version 2 (EDv2.2; Medvigy et al., 2009; Moorcroft et al., 2001). This model has been implemented in a variety of ecosystems (Antonarakis et al., 2014; Davidson et al., 2011; Kim et al., 2012; Levine et al., 2016; Lokupitiya et al., 2016; Medvigy et al., 2013, 2012; Trugman et al., 2016; Xu et al., 2016; Zhang et al., 2015) and, unlike most DGVMs, EDv2.2 represents the vertical and horizontal heterogeneity of terrestrial ecosystems (Longo et al., 2019b, 2019a). A further strength is that EDv2.2 accommodates the local variability of vegetation composition and structure. However, a thorough evaluation of this model in drylands is lacking. The heterogeneity and in many areas, the sparsity of vegetation in dryland communities, makes modeling these ecosystems challenging.
The objective of this study is to explore the ability of EDv2.2 to predict GPP capacity and activity in a dryland study area. To address this objective we (i) perform a sensitivity analysis and a state-of-the-art calibration method; (ii) assess GPP capacity by comparing the model output with EC tower data with varying vegetation productivity; and (iii) assess GPP activity by comparing long-term trends and phenometrics (SOS and EOS) between model GPP simulations and remote sensing data.
Section snippets
Study area and data
The study area is Reynolds Creek Experimental Watershed (RCEW), located in the northern Great Basin of the western US (Fig. 1). In this study, we used three EC towers on the RCEW spanning an elevation range of ~1,000 m (Fig. 1, Table 1). With increasing elevation, mean annual precipitation increases and temperature decreases (Table 1; Flerchinger et al., 2019). The dominant vegetation cover of each EC site is a different species of sagebrush (Artemisia spp.) including Wyoming big sagebrush (
Results
The results of the Morris sensitivity analysis (SA) are shown in Fig. 2. Among all parameters, the specific leaf area (SLA), stomatal slope (STO_S), cuticular conductance (CUT_C), and maximum carboxylation rate (VM0) show the highest individual influence (µ*). These parameters also show the largest non-linear influence and interaction effect at all sites (σ). The results from the other eight parameters indicate that EDv2.2 is less sensitive to them in simulating GPP (e.g. clustering near the
Model calibration and sensitivity analysis
We compared parameters identified here with other studies in RCEW (Pandit et al., 2019a; Renwick et al., 2019). Three of the four influential parameters in this study (SLA, STO_S and VM0) are similar to those identified by Pandit et al., 2019. However, only SLA was identified as important by Renwick et al. (2019). These parameter discrepancies among studies may be due to the SA method or the DGVM structure. For example, in Pandit et al. (2019), a local SA was used in comparison to the Morris SA
Conclusion
In summary, our main conclusion is that at lower elevations, precipitation drives the general trend of GPP which is captured by both MODIS and EDv2.2; however, the model generally exaggerates this trend in comparison to MODIS. Introducing additional PFTs, making structural modification to the model (e.g. phenology scheme), and incorporating land surface processes should increase model applicability at higher elevations. Adding more dryland PFTs will not only contribute to GPP capacity (e.g.
Funding
This research was funded by NASA Terrestrial Ecology NNX14AD81G, Department of the Interior Northwest Climate Adaptation Science Center graduate fellowship, US Forest Service Western Wildlands Environmental Threat Assessment Center (WWETAC) to Boise State University through Joint Venture Agreement (17-JV-11221633-130), and the Joint Fire Science Program Project ID: 15-1-03-23.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
Field access and support were provided by USDA ARS Northwest Watershed Research Center, NSF EAR 1331872, and NSF EAR 1665519. We would like to acknowledge the high-performance computing support of the R2 compute cluster (DOI: 10.18122/B2S41H) provided by Boise State University's Research Computing Department. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.
References (90)
- et al.
Sensitivity analysis of an environmental model: an application of different analysis methods
Reliab. Eng. Syst. Saf.
(1997) - et al.
The role of sensitivity analysis in ecological modelling
Ecol. Modell.
(2007) - et al.
Ten iterative steps in development and evaluation of environmental models
Environ. Model. Softw.
(2006) - et al.
Estimating the critical phosphorus loading of shallow lakes with the ecosystem model PCLake: sensitivity, calibration and uncertainty
Ecol. Modell.
(2010) - et al.
Representing the root water uptake process in the Common Land Model for better simulating the energy and water vapour fluxes in a Central Asian desert ecosystem
J. Hydrol.
(2013) - et al.
Process refinement contributed more than parameter optimization to improve the CoLM's performance in simulating the carbon and water fluxes in a grassland
Agric. For. Meteorol.
(2020) - et al.
Estimation of big sagebrush leaf area index with terrestrial laser scanning
Ecol. Indic.
(2016) - et al.
Climate change and North American rangelands: trends, projections, and implications
Rangel. Ecol. Manag.
(2013) - et al.
Modeling phenological controls on carbon dynamics in dryland sagebrush ecosystems
Agric. For. Meteorol.
(2019) - et al.
Climate change, phenology, and phenological control of vegetation feedbacks to the climate system
Agric. For. Meteorol.
(2013)