Develop capabilities to extend temporal resolution to sub-decadal for earth system models.
Overall Performance Measures
1st Quarter Metric – COMPLETED
New model developments for dynamically representing changes in river flow have led to important advances in capturing seasonal to decadal changes in surface water dynamics and will be critical for associated changes in nutrients and regional climate. Through freshwater discharge, sediment, carbon, and nitrogen fluxes to the ocean, and outgassing of CO2 to the atmosphere, rivers play an important role in the water and biogeochemical cycles of the coupled earth system. Surface water transport is also closely linked to human activities as water resources are managed to balance supply and demand through regulations of streamflow. Understanding and predicting natural hazards such as flooding and its hydrological and ecological consequences have required improved understanding and modeling of surface water dynamics.
In response to these requirements, the Model for Scale Adaptive River Transport (MOSART) has now been coupled with the Community Land Model (CLM) and implemented and evaluated globally. Decadal simulations of streamflow are shown to reproduce reasonably well the observed daily and monthly streamflow at over 1,600 world’s major river stations in terms of annual, seasonal, and daily flow statistics. In contrast to the River Transport Model (RTM) currently used in the standard version of CLM that assumes a constant river velocity in space and time, numerical experiments show that the spatial and temporal variability of river velocity simulated by MOSART is necessary for capturing streamflow seasonality and the annual maximum flood.
MOSART is a physically based runoff-routing model designed for applications across watershed, regional, and global scales with relatively consistent performance at different resolutions (Li et al. 2013). All model parameters are physically based and only a small subset of them requires calibration. MOSART has been coupled with CLM (Lawrence et al. 2011) in the same manner as standard runoff routing module, River Transport Model (RTM) (Branstetter and Erickson, 2003). CLM simulates the surface runoff and baseflow for each grid cell at each time step. The gridded CLM-simulated surface runoff and baseflow is transferred to MOSART at the end of the time step and MOSART routes the runoff across hillslope and through sub-network and main channels. While RTM uses globally uniform and constant river velocity, MOSART explicitly simulates both spatial and temporal variability of flow velocity.
The coupled CLM-MOSART model has been applied globally using land surface parameters and atmospheric forcing (Qian et al. 2006) provided with the NCAR-I2000 configuration for 1995-2004. CLM simulation is performed on a 0.9o×1.25o grid at 30-minute time step. MOSART is applied at 0.5 degree resolution for runoff routing. To assess the impacts of the added model complexity in MOSART, five successive simulations are performed by turning off the subgrid routing and removing the temporal and spatial variability of channel flow velocities. Results show that representing the spatial and temporal variations of flow velocities has important effects on simulating seasonality of streamflow and magnitude of annual maximum flood. Each level of complexity enabled by MOSART compared to a simpler model can lead to statistically significant differences in simulating streamflow. The more process-oriented MOSART overall captures the dynamics of surface water and provides a framework for modeling stream temperature, river biogeochemistry, and inundation dynamics that provide key linkages with other Earth system and human system components for a more holistic representation of global and regional water and carbon cycles.
2ND QUARTER METRIC – COMPLETED
New land model developments improve upon the exchange of carbon between plants and atmosphere, while also accommodating field measurements that are difficult to understand. The timing of plant growth observed in temperate forests does not align well with photosynthetic potentials: leaf-scale measurements indicate that more carbon should be taken up by the forest canopy over the course of the growing season than actually appears as new growth, based on repeat measurements. Based on measurements of the ratios of carbon-to-nutrients, such as nitrogen and phosphorus (vegetation stoichiometry), and measurements of the relative growth of various tissues such as leaves, stems, and roots (vegetation allometry), it appears that availability of nutrients has a strong control on actual growth and accumulation of biomass without having a strong influence on leaf-scale photosynthetic potential. Another newer hypothesis suggests that plant storage of carbohydrates allows high rates of photosynthesis without immediate evidence of growth.
Recent field experimentation in several temperate forest systems supports this idea, and non-structural carbohydrate pools have been measured in many plant types. Terrestrial ecosystem models use a variety of approaches to reconcile the apparent contradiction. Some reduce concentrations or activity of photosynthetic enzymes to match observed growth under conditions of limited nutrient availability. Other models shift allocation toward roots to improve access to limiting resources, while others, including the Community Land Model, assume a down-regulation of photosynthetic rate without imputing a physiological mechanism. A series of fully-coupled climate-biogeochemistry simulations is carried out to investigate the new hypothesis and the influence of a temporary carbon storage mechanism on atmospheric carbon dioxide concentrations (CO2,atm) at global and regional scales. In particular, the seasonal cycle of CO2,atm is investigated as an integrative metric of predicted carbon cycle dynamics. Temporary carbon storage by plants is found to have a profound influence on the seasonal cycle amplitude of CO2,atm without necessarily changing other aspects of the coupled climate-biogeochemistry simulation.
The influence of various parameterizations of temporary non-structural carbohydrate (NSC) storage on global and regional scale seasonal cycle of CO2,atm was studied and quantified through a series of global-scale fully coupled climate-biogeochemistry simulations using the Community Earth System Model (CESM). Simulations were performed on a nominal 1° x 1° horizontal grid. A global synthesis of observed CO2,atm data for the period 1990-2000 was obtained from the GlobalView data set (GLOBALVIEW-CO2, 2013). In order to compare simulation results with the GlobalView product, transient historical simulations were performed starting in simulated year 1850 and continuing through 2000, using standard historical forcing data sets for fossil fuel emissions of CO2 and other greenhouse gases, and observation-based estimates of atmospheric nitrogen deposition and land use and land cover change. Simulations branched at year 1990 with the specification of several different parameterizations for NSC storage, with different turnover times. We also performed a control simulation using the default approach of assumed instantaneous down-regulation of photosynthetic rate in the face of nutrient limitation. Since measurements of CO2,atm are made at many locations and with high precision, it may be possible to place strong constraints on the physiological mechanism of temporary carbon storage by optimizing against existing the observed seasonal cycle in the GLOBALVIEW data set.
3RD QUARTER METRIC – COMPLETED
The regionally refined capability in the Accelerated Climate Model for Energy (ACME) was used to show improvement of precipitation over mountainous regions in the United States at high-resolution without having to use a global high-resolution atmospheric model. The refined region was centered over the continental United States (CONUS). A regionally refined model can display the signatures of a global high-resolution model at a tenth of the computational cost of a global high-resolution simulation. A dry precipitation bias off the coast of the eastern United States in the global 1° low-resolution model was improved in the regionally refined simulation when compared to observations. This improvement in the regional refinement is also comparable to the global ¼° high-resolution simulation.
To demonstrate the improvement of precipitation over the refined region, the climatologies of a regionally refined configuration, a global low-resolution 1° configuration, and a high-resolution ¼° configuration of the Community Atmospheric Model (CAM5.3), version 5.3, in the ACME model, version 0.1, are compared with a Tropical Rainfall Measuring Mission (TRMM) high-resolution ¼° precipitation dataset. Five-year pre-industrial simulations of the climatological year of 1850 of the CONUS grid, the 1° low-resolution grid, and the ¼° high-resolution grid were performed. The case for using regional refinement to reduce the computational costs is strong. The low-resolution 1° simulation has 5400 elements and can produce 3.31 Simulated Years per Day (SYPD) on an institutional cluster with 400 cores. The CONUS simulation has 9905 elements and can produce 0.90 SYPD on the same cluster. The reduction in speed is due to the smaller time step needed by the higher resolution region. For comparison, the number of elements in the globally uniform ¼° simulation is 86,400. This high-resolution simulation was not run on the institutional cluster because it would take months to produce a 5-year simulation, and therefore must be run on a Leadership Class Facility computer.
4TH QUARTER METRIC – COMPLETED
The Los Alamos Sea Ice Model (CICE) includes complex parameterizations of sea-ice physics (such as ice rheology treatment or biogeochemistry), with a large number of parameters for which accurate values are still not well established. To enhance the credibility of sea-ice predictions, it is necessary to understand the relative sensitivity of model results to choices of model physics parameters and external forcing. We describe here studies undertaken to evaluate the relative importance of physical representations in the CICE model to its prediction of ice extent, area, and volume.
A large ensemble of over 500 global CICE (v5.1) simulations for the period 1980-2009 was performed with observed atmospheric fields and a mixed-layer ocean. Such a large ensemble allowed 10 to 20 simulations for each input parameter with the parameters sampled within their acceptable range. Ensemble results were evaluated against observed sea-ice extent, area, and volume to quantify the sensitivity of each parameter, and screen for the most important parameters for later studies.