Demonstrate in the Coupled DOE-E3SM Model, the Importance of Environmental Factors
in Affecting Ecosystem Productivity and Surface Energy Exchanges
OVERALL PERFORMANCE MEASURES
1ST QUARTER METRIC COMPLETED:
Soils contain the largest terrestrial pool of organic carbon (C), storing at least twice as much C as earth’s atmosphere (Köchy et al., 2015; Scharlemann et al., 2014). Uncertainties surrounding the response of soils to climatic and other changes contribute substantial uncertainty to C cycle and climate projections in the Earth system (Arora et al., 2013; Friedlingstein et al., 2014; Todd-Brown et al., 2013): the magnitude of their uncertainty is comparable to that of cloud feedbacks, traditionally regarded as the most significant unknown in climate modeling (Gregory et al., 2009). For example, Jones and Falloon (2009) reported a strong relationship between changes in soil organic C (SOC) and the strength of simulated C‐climate feedbacks within ESMs, while Riley et al. (2018) and Gaudio et al. (2015) found that model representation of nitrogen biogeochemistry and uptake patterns had significant climatic effects at larger spatial scales. At the same time, models’ structural uncertainty (the uncertainty deriving from how various models represent particular processes differently) is an unknown factor (Tebaldi and Knutti, 2007); there have been few attempts to examine how structural uncertainty within a single model–as opposed to model-to-model variability in, e.g., CMIP5 (Friedlingstein et al., 2014; Knutti and Sedláček, 2012)–affects model behavior and performance (Ricciuto et al., 2008). The investigation here indicates that the structural uncertainty deriving from models’ biogeochemical process representation is significant, although not as large as other sources such as parametric uncertainty (uncertainty deriving from the model inputs such as field-based data).
The U.S. Department of Energy’s Energy Exascale Earth System Model (E3SM) is unusual among ESMs in that it has two approaches to terrestrial biogeochemistry in its land model, the E3SM Land Model (ELM): the primary approach ELMv1-CTC-CNP (led by a team at Oak Ridge National Laboratory) and the alternative ELMv1-ECA-CNP (led by Lawrence Berkeley National Laboratory group). These differ in three key aspects of biogeochemistry–stoichiometry, allocation, and nutrient competition–and represent distinct approaches to the overall problem, as described below. To evaluate the effects of uncertainty in biogeochemistry methodology, we performed a series of site- and global-scale uncoupled simulations using both CTC and ECA. The models’ outputs were compared against a variety of observational reference datasets. This work will allow the model structural uncertainty in this area to be assessed, for the first time, against other sources of uncertainty, e.g. parametric and ensemble sources.
2ND QUARTER METRIC COMPLETED:
Stream temperature is a key water quality measure for water management, thermoelectric power production, and conservation activities. Currently, 85% of electricity generation in the United States comes from power plants that require cooling. Changes in stream temperature and water availability directly affect thermoelectric power generation capacity. High water temperature and low streamflow can increase cooling water requirements and restrict cooling water availability, which constrains the usable capacity of thermoelectric power plants. This constraint may become more acute in the future, as droughts are projected to be more widespread and prolonged in a warmer climate. These motivate the need to model stream temperature and understand the relative impacts of air temperature and water management on stream temperature.
Stream temperature is mainly controlled by the heat exchanges between river water body and air and river banks, and heat transportation along the river networks. Notably, 98% of the rivers and streams in the United States have been dammed, diverted, or developed. Reservoirs regulate flows for various purposes, such as flood control, irrigation, hydropower production, and navigation. This alters the flow regime by storing water during high-flow periods and enhancing the low flows during the dry season. Changes in the flow regime have important impacts on stream temperature, as they alter the heat exchanges between the rivers and air and river banks.
Through an effort supported by the MultiSector Dynamics activity within the Earth and Environmental Systems Modeling program, a stream temperature module has been developed as part of the Model for Scale Adaptive River Transport (MOSART) and coupled with a water management model (WM). This metric report describes (1) the implementation of the stream temperature module and its global testing and evaluation within the Department of Energy’s Energy Exascale Earth System Model (E3SM) and (2) analysis of simulations to understand the relative impacts of air temperature and water management on stream temperature in basins around the world. Simulations with and without water management show that water management has large impacts on the seasonal cycle of stream temperature in arid and semi-arid regions (western U.S., central Asia, northeastern China) where water management alters the flow regime to provide irrigation water supply and in India where groundwater pumping for irrigation is prominent.
MOSART-heat is a stream temperature module (Li et al. 2015a) developed on top of MOSART (Li et al. 2013; 2015b), which is the river component of E3SM v1. MOSART-heat has been implemented in E3SM through coupling with the E3SM Land Model (ELM). MOSART-heat mainly represents natural thermodynamic processes that control stream temperature, including the lateral heat fluxes from hillslope and soils (along with surface and subsurface runoff) into tributary channels, heat balance in tributary channels and main channels respectively. The surface runoff temperature is estimated as the average soil temperature of the top three soil layers in ELM, and the subsurface runoff temperature is estimated as the average soil temperature within the saturated soil layers, which vary with time due to changes in the water table level.
3RD QUARTER METRIC COMPLETED:
A paper published in Nature in 2000 showed that a climate system model with prognostic atmospheric CO2 concentration was sensitive to the representation of carbon cycle processes in land and ocean ecosystems (Cox et al. 2000). Since that time a multitude of studies has been performed with increasingly sophisticated climate and Earth system models, examining the interactions or feedbacks, among atmospheric CO2 concentration, near-surface air temperature, and physical and biological processes on land and in the oceans.
More recently, studies have begun investigating the interactions of the carbon cycle with other biogeochemical cycles. Nitrogen and phosphorous are two of the most important nutrients for plant growth, and the availability of these nutrients plays important roles in the carbon cycle dynamics of land ecosystems. Several modeling studies have examined the interactions of carbon and nitrogen cycles. Those studies reported that the overall effect of land ecosystem nitrogen limitation on the carbon cycle was to reduce the rate of carbon uptake as fertilized by rising atmospheric CO2 concentration, and to significantly reduce or eliminate the loss of carbon to the atmosphere caused by warming, as compared with earlier modeling studies that ignored nitrogen cycle dynamics. New model development and evaluation have started to produce initial estimates of the influence of phosphorus limitations on global-scale carbon cycle dynamics. Here, the results of those studies are examined and summarized, to reflect the best current understanding of how the additional process-level complexity of phosphorus limitations modifies climate and carbon cycle interactions.
Evaluation of the influence of phosphorus limitation on the carbon cycle and carbon-climate feedbacks is presented at two spatial scales. First, the impact of carbon-nitrogen-phosphorus (CNP) coupling is examined relative to coupling of carbon-nitrogen (CN) coupling at the continental scale, focusing on the tropical forests of the Amazon region in South America. Simulation results presented here for the Amazon region are extracted from offline land model simulations using the Community Land Model version 4 (CLM4) in its CN and CNP configurations (Yang et al. 2016).
Second, the influence of CNP coupling is evaluated at the global scale, focusing on quantification of climate-carbon cycle feedbacks over the period 1850-present for simulations carried out within a coupled earth system model. Simulations presented here for the global-scale feedback analysis are extracted from fully coupled atmosphere-ocean-sea ice-land simulations performed with the Energy Exascale Earth System Model (E3SM), version 1.1, in its coupled biogeochemistry configuration (E3SMv1.1-BGC, Burrows et al. in review). The influence of coupled CNP biogeochemistry on carbon-climate feedbacks is evaluated in a quantitative climate-carbon cycle feedbacks framework (Friedlingstein et al. 2001).
Taken together, these initial regional and global results suggest that while there are certain to be regional differences, especially in tropical forest regions, the global scale implications of phosphorus limitation on carbon cycling may emerge as a second-order modification to the prior estimates associated with the introduction of nitrogen limitation.
4TH QUARTER METRIC COMPLETED:
Representations of terrestrial ecological processes have been overly simplified in the current generation of models used for climate and ecosystem projections, such as the land surface model (LSM) components of Earth System Models (ESMs). This has implications for the ability of these models to adequately capture climate-ecosystem feedbacks. For example, most LSMs use prescribed, static maps of plant functional types (PFTs), and thus cannot capture shifts in PFT composition. Some LSMs do simulate simplistic PFT shifts, but are constricted to current bioclimatic zones (i.e., zones where the residing vegetation exists depending on temperature gradients, both latitudinal and elevation, and precipitation gradients) instead of emerging from the physiology- and competition-based demographic rates that determine resource competition and plant distributions in real ecosystems. Thus, current ESMs are limited in ecological detail and realism, for example, important factors such as ecosystem structure and demography. Omission of mechanisms, such as mortality and disturbances that influence biomass turnover and carbon allocation response to changing climates, limits the ability of these models to realistically forecast ecosystem responses to anomalous temperature and precipitation conditions.
Therefore, a Vegetation Demographic Model (VDM) called The Functionally Assembled Terrestrial Ecosystem Simulator (FATES) that incorporates dynamic vegetation processes has been implemented within the Department of Energy’s Energy Exascale Earth System Model (E3SM). The benefits of simulating vegetation processes by FATES is that it provides 1) height-structured heterogeneity in light availability, 2) competition for water and nutrient resources that could lead to exclusion or coexistence, 3) ecosystems that assembly mechanistically, 4) representing major disturbances (e.g., extreme drought) and recovery, 5) plant distribution that emerges from trait filtering, and 6) abandons bioclimatic envelopes.
ELM-FATES has been successfully implemented with promising model evaluation and predictive capabilities for carbon cycling, and surface energy and water fluxes. Among multiple advances to ELM-FATES, we have improved a previous high latitude bias in leaf area index and net primary productivity, described in more detail below. Additionally, as a result of dynamic vegetation and ability for plants to assemble in a manner that optimized current conditions, the improved predictions of productive tropical plant functional types increased the evapotranspiration (ET) fluxes from the tropics, a critical water flux for global circulations of water and energy. This analysis showed that improved shifts in global plant distributions lead to large differences in land-atmosphere exchanges.
ELM-FATES is a cutting edge VDM coupled within E3SM’s land surface model, ELM, thus representing demography at global scales and is an alternative to the “big-leaf” vegetation processes in ELM. ELM-FATES represents vegetation in a more highly resolved manner through demography. For example, demography provides finer scales of plant heterogeneity, by vegetation being separated into many ‘cohorts’ (i.e. plants of similar height and functional types) across various ‘patches’ (i.e. area of similar time since a disturbance event), therefore allowing for the land surface to be divided into different successional stages. Cohorts of trees are each competing in different growth phases, which can be informed directly by field observations.
For each cohort the dynamic process of recruitment, growth, and mortality are mechanistically represented and emergent properties due to competition. ELM-FATES is a critical addition for representing disturbance-partitioned landscapes in ESMs. For documentation, see https://fates-docs.readthedocs.io/en/latest/index.html.
The underlying model structure and concepts in ELM-FATES are based on the ED model (Moorcroft et al., 2001), introducing individual plants being represented as cohorts, patch ‘time-since-disturbance’ concept, and trait filtering, but with multiple updates, and coupled to the physical processes inherit to a land surface model scheme (Fisher et al., 2015). A major addition to ELM-FATES is the adoption of the Perfect Plasticity Approximation (PPA) (Purves et. al., 2008) used for the accounting of canopy crown spatial arrangements. Further updates include multi-layer multi-PFT radiation transfer, carbon allocation to storage, and a plant hydrodynamic module (Christoffersen et. al. 2016). The hydrodynamic model simulates different diurnal and seasonal cycles of photosynthetic response to water stress depending on the strategies related to diverse hydraulic traits, which are difficult for traditional ESMs to capture. Plant hydraulics in ELM-FATES enables model-based experiments on how temperature and drought influence plant stress and mortality due to hydraulic failure.
Vegetation demography integrated with plant hydraulics, enhanced representations of plant trait variation, and explicit treatments of resource competition have all been identified as critical areas for advancing current models, and constitute necessary advances for realistically representing future ecosystem states (Choat et. al., 2018, Fisher et. al., 2015, 2018, Scheiter et. al., 2013, Weng et. al., 2015). Demography also provides different sensitivities to rising CO2 via finer-scale ecological processes, ultimately altering responses of land-atmosphere interactions (Levine et al., 2016; Purves and Pacala, 2008), which cannot be achieved in models with aggregated vegetation.