FY 2017 Performance Metrics

Extend the capabilities of the DOE’s Advanced Climate Model for Energy (ACME) to simulate and evaluate human-natural interdependencies for the carbon and water cycles.

OVERALL PERFORMANCE MEASURES

1ST QUARTER METRIC – COMPLETED

Implement and evaluate the impacts of irrigation on the hydrologic cycle in the ACME model

Product Definition

Globally, around 70% and 90% of freshwater withdrawals and consumptions are used for irrigation purposes, respectively [Döll, 2009]. The large-scale water withdrawals from rivers, lakes, reservoirs, and aquifers have directly and substantially altered the terrestrial water and energy cycles [Haddeland et al., 2006; Kustu et al., 2010; Wada et al., 2010; Famiglietti et al., 2011; Leng et al., 2014; Pohkrel et al., 2015]. Irrigation impacts are expected to increase with growing world population and food demand, while production of bioenergy crops may further exacerbate irrigation water demands. Hence understanding the role of irrigation in human-Earth system interactions is important for adaptive planning of water, land, and energy use. During the last two decades, noteworthy progress has been made to represent irrigation in global hydrological models for assessing water resource availability and use, but there have been limited efforts in modeling irrigation in global land surface models and Earth system models (ESMs). From a modeling perspective, estimation of irrigation amount and choices of irrigation water sources and methods are key aspects in parameterizing irrigation water use and modeling its impacts. The most under-represented aspect of irrigation modeling is the irrigation method that determines how the extracted water is applied to the irrigated areas. Given the various approaches used to model irrigation, considerable discrepancy exists in simulating irrigation effects, especially at local/regional scales. To reduce such uncertainty, an interactive irrigation scheme has been developed and incorporated into the Accelerated Climate Modeling for Energy (ACME) Land Model (ALM) with consideration of the irrigation water sources (i.e., surface water and groundwater) and methods (i.e., drip, sprinkler and flood irrigation). The model has been used to investigate how the simulated irrigation effects are influenced by the water sources and methods for irrigation. By designing different irrigation scenarios, the pathways through which irrigation affects land surface water balance and how differences in irrigation water sources and methods affect irrigation water use efficiency are studied using numerical experiments. A set of numerical experiments with the new irrigation scheme in ALM showed that both different water sources (e.g., groundwater versus surface water) and methods of irrigation (e.g., sprinkler, flood, or drip) could lead to large differences in simulating the hydrological impacts of irrigation, suggesting that modeling of irrigation must consider those two factors to reduce uncertainty.

Product Documentation

The Accelerated Climate Modeling for Energy (ACME) is an ESM developed by the U.S. Department of Energy (DOE). To represent irrigation in the ACME Land Model (ALM), a groundwater pumping module has been added following Leng et al., 2014 so that the estimated irrigation water demand can be met by withdrawing water from surface water storage (i.e., accumulated total runoff) and/or groundwater. The fraction of surface water withdrawal to the total water withdrawal is constrained by a global data set. The water table depth is updated after water is extracted for irrigation. In addition to the direct impacts on groundwater resources, changes in water table depth lead to subsequent changes in the subsurface drainage and recharge from the bottom soil layer to the aquifer, thus indirectly influencing soil water content through soil-aquifer interactions.

After the irrigation water is extracted, it can be applied to the irrigated areas using three distinct irrigation methods, i.e., drip, sprinkler, and flood irrigation. In practice, drip irrigation applies water directly to the root zone to reduce the wetted area, so it features the highest water use efficiency. Sprinkler irrigation is less efficient because a large fraction of the water is sprayed into the air, which can result in wind drift and evaporation losses. In flood irrigation, less water is lost to evaporation than in sprinkler irrigation, but more water can be lost as runoff in the fields. The irrigation methods are parameterized mainly following the guidance of reflecting the distinct water use efficiencies among the three methods. For sprinkler irrigation, water is added directly to the canopy as precipitation. In drip irrigation, water is applied directly and slowly to the soil layers in the root zone to facilitate plant water use. In flood irrigation, water is poured directly and quickly to the ground surface in a period of 30 min. as precipitation falling on the ground surface, bypassing the canopy. The approach implemented in ALM can reflect the distinct irrigation water use efficiencies among the three irrigation methods, which is important for evaluating water use and hydrologic impacts driven by irrigation.

A series of sensitivity experiments were performed with ALM at 1-degree resolution driven by observed atmospheric forcing [Qian et al., 2006]. After cycling the atmospheric forcing for model spin-up, results for 1971-2005 were compared to contrast simulations with different irrigation water sources (surface water only, surface water and groundwater) and different irrigation methods (sprinkler with irrigation water applied to the canopy versus directly to the ground, flood, and drip) with water sources from both surface and groundwater. The impacts of water sources and irrigation methods were evaluated by comparing the effects of irrigation on water fluxes (e.g., evapotranspiration and runoff), soil moisture, groundwater table depth, and irrigation water use efficiency.

2ND QUARTER METRIC – COMPLETED

Develop and Implement Global Crop-Planting Date Triggers for the ACME Model

Product Definition

Climate and weather conditions are among the key drivers of crop suitability and productivity in a region. The influence of climate and weather on the growing season determine the amount of time crops spend in each growth phase, which in turn impacts productivity and, more importantly, yields. Precipitation and temperature are responsible for 30% of the variability in crop yields (Lobell and Field, 2007). Planting date can also have a strong influence on yields (Deryng et al., 2011; Drewniak et al., 2013; Waongo et al., 2015), with earlier planting generally resulting in higher yields (Choi et al., 2017). Crop models are sensitive to planting date: for example, early planting results in higher yield for soybean crops because cooler spring temperatures slow crop development and allow an extended growing season (Drewniak et al., 2013). Furthermore, planting date is changing. Maize planting date in the Midwest U.S. is now two weeks earlier than just a few decades ago (Kucharik, 2006), which has resulted in a 19-53% increase in yields across many U.S. states (Kucharik, 2008). Although this trend is the result of improved genotypes (Kucharik, 2006) rather than early spring warming, longer growing seasons caused by future climate change could also drive early planting decisions. The Accelerated Climate Model for Energy (ACME) Land Model (ALM) does include temperature thresholds to define crop planting dates, but the application is limited to mid-latitudes. A more robust metric that includes other climate factors (i.e., precipitation) is added to improve the predicted timing of crop planting and will also allow global crop production.

Product Documentation

Crop models need an accurate method to predict plant date to allow these models to: 1) capture changes in crop management to adapt to climate change, 2) accurately model the timing of crop phenology, and 3) improve crop simulated influences on carbon, nutrient, energy, and water cycles. Climate plays a strong role governing planting date (Choi et al., 2017). Previous studies have used climate as a predictor for planting date (Deryng et al., 2011; Waha et al., 2012; Waongo et al., 2014), although other factors besides climate influence a farmer’s decision to plant (e.g. workability, economics, culture) (Iizumi and Ramankutty, 2014). Climate as a plant date predictor has more advantages than the use of fixed plant dates. For example, crop expansion and other changes in land use due to changing temperature conditions, new drought or cold tolerant crop hybrids and cultivars, and increased food demands can be accommodated without additional inputs. As such, a new methodology to implement predictive planting date based on climate inputs is added to the Accelerated Climate Model for Energy (ACME) Land Model (ALM). The model considers two main sources of climate data important for planting: precipitation and temperature. The new ALM model is compared with observations to test its ability to capture planting dates globally.

3RD QUARTER METRIC – COMPLETED

Implement and evaluate the impacts of water management on the hydrologic cycle in the ACME model

Product Definition

Human activities associated with water use, reservoir operations, and groundwater pumping can induce pronounced effects on water resources availability and streamflows. Postel et al. [1996] estimated that 26% of global evapotranspiration is associated with human activities, and 56% of spatially and temporally available global runoff is withdrawn from water bodies. Also, unsustainable groundwater depletion to supplement surface water supply is a widespread practice across the globe, so about a third of Earth’s largest groundwater aquifers are being rapidly depleted by human consumption [Wada et al. 2012, 2014; Richey et al. 2015]. Recognizing the critical role of water in the Earth system, human activities such as irrigation, reservoir operations, and groundwater pumping have been represented in some land surface models. Modeling experiments with Earth system models informed by modeling of the socio-economic drivers of human activities are only emerging [Hejazi et al. 2015]. Two products of research are described here.

First, building on previous efforts to better understand and represent the human-natural system feedbacks, new development has been implemented in an integrated water model based on the Model for Scale-Adaptive River Transport (MOSART) [Li et al. 2013, 2015] coupled to a water management (WM) model [Voisin et al. 2013a&b] to represent groundwater use and return flow. Return flow refers to unconsumed water such as that withdrawn for thermoelectric power generation and then returned to the streams. Driven by runoff simulated by a climate model and spatially distributed sectoral withdrawal and consumptive water demands simulated by the Global Change Assessment Model (GCAM) [Hejazi et al. 2015], MOSART-WM was applied to the US to simulate the spatial redistribution of water resources and water deficit by water use and flow regulation [Voisin et al. 2017].

Second, MOSART-WM has been coupled with the ACME Land Model (ALM) as a new model feature to represent irrigation and water management in ACME. This new capability will enable ACME to more realistically simulate water-cycle processes in the Earth system and understand water availability and its natural and human drivers. Global simulations have been performed with ALM-MOSART-WM to evaluate the impacts of irrigation and reservoir operations on streamflow and reservoir storage.

Water supply deficit, which should be negligible over the historical period, is reduced in simulations that included groundwater use and return flow. This demonstrates that the new development in MOSART-WM provides a more realistic representation of water management. Simulations using the coupled ALM-MOSART-WM demonstrated successful coupling in the ACME framework and a new capability to address science questions related to the interactions between human and natural processes that influence the regional and global water cycles.

Product Documentation

In the integrated water model, MOSART-WM, river transport is simulated by MOSART while reservoir operations and flow regulation are simulated by WM. WM is a large-scale water management model [Voisin et al. 2013a] that allocates water demands and manages reservoir releases and storage. WM relies on generic operating rules that mimic monthly releases and storage patterns based on the objectives of the reservoirs. The reservoir model is coupled to MOSART, which routes the regulated flow from reservoirs to downstream channels as well as the local runoff in the hillslope and tributaries within each grid cell. The GCAM irrigation and non-irrigation demand are used as inputs to MOSART-WM.

In the previous implementation described in Voisin et al. [2013a,b], MOSART-WM only considers consumptive demand, although in reality a significantly larger amount of water can be withdrawn for non-consumptive use such as water for cooling thermoelectric power plants that is returned to rivers. In addition, the model limits demands to be provided by the surface-water systems only, which tends to exaggerate water stress. In more recent developments, Voisin et al. [2017] have addressed the limitations of MOSART-WM by considering groundwater supply as well as distinguishing consumptive and withdrawal demands and the associated return flow. Representing return flow in MOSART-WM requires both withdrawals and consumptive use for both irrigation and non-irrigation sectors. These are provided by GCAM and allocated to groundwater and surface-water systems. The unconsumed water is returned to the grid cells where withdrawals are distributed.

In the offline implementation of MOSART-WM, runoff is provided by an offline land surface model simulation; irrigation and non-irrigation water demands are provided by an offline GCAM simulation. This implementation lacks consistency between the irrigation demand simulated by GCAM and that simulated by the land surface model due to differences in spatial and temporal scales, process representations, and land use-land cover used in the two models. For more realistic simulations of irrigation effects and water supply deficit, a one-way coupling of MOSART-WM with the ACME Land Model (ALM) has been implemented. During the crop growing season, ALM calculates the irrigation demand at 6am local time each day based on the soil moisture status [Leng et al. 2013]. Irrigation is applied at a uniform rate over a 6-hour period by adding it to the precipitation input to the ALM grids. The runoff and irrigation demand are passed from ALM to MOSART-WM through the flux coupler used in ACME. Water balance checks are implemented in ALM and MOSART-WM to ensure water conservation in the coupled model. This simple one-way coupling of ALM with MOSART-WM provides a framework for future implementations to allow groundwater use and return flow, and two-way coupling in which unmet demand is passed from MOSART-WM to ALM through the flux coupler to constrain irrigation water supply in ALM.

The new developments in MOSART-WM to account for groundwater use and return flow are used to understand their contrasting local effects: more available supply due to the use of both surface water and groundwater versus more stress due to the consideration of withdrawals in addition to consumptive demand. A benchmark simulation and three numerical experiments were performed to understand and quantify the relative impacts of accounting for groundwater use and consumptive versus withdrawal demands and the associated return flow.

4TH QUARTER METRIC – COMPLETED

Decompose and evaluate the effects of changes to CO2, climate, and land-use/land-cover on the carbon cycle for the E3SM model

Product Definition

Changes in temperature, precipitation, atmospheric CO2 concentrations, and land use/land cover have implications for the carbon cycle. Increases in atmospheric CO2 concentrations lead to more carbon stored in terrestrial ecosystems, while warming tends to reduce terrestrial carbon storage [P. Friedlingstein et al., 2006; Pierre Friedlingstein et al., 2014], although models diverge widely in the strength of these responses. Land-use and land-cover change can result in more or less carbon storage in the land, depending on the change; for example, carbon release to the atmosphere from deforestation accounts for approximately 15% of anthropogenic greenhouse gas emissions on an annual basis [Le Quere et al., 2009], but the regrowth of previously cleared land is a significant carbon sink. Changes in temperature and CO2 concentrations also influence crop yields [Rosenzweig et al., 2014], which can in turn result in changes in land use and land cover [Nelson et al., 2014]. However, the direction and magnitude of the effect varies across crop and Earth system model. Previous studies have examined the role of changing atmospheric carbon and climate (temperature and precipitation) on the carbon cycle [P. Friedlingstein et al., 2006; Pierre Friedlingstein et al., 2014; Jones et al., 2013], but these studies do not specifically attribute changes to land use/land cover or the feedbacks between changing climate and land use/land cover. To address this problem, we performed a series of experiments, and developed novel analytical metrics, to decompose the effects of carbon, climate, and land use on the carbon cycle for the Energy Exascale Earth System Model (E3SM) project. Higher temperatures tend to reduce carbon storage, while increases in CO2 concentrations increases carbon storage. Land-use feedbacks increase terrestrial carbon storage, resulting in more carbon stored per unit of temperature change and more carbon stored per unit of CO2 increase than without feedbacks.

Product Documentation

The integrated Earth System Model (iESM; [Collins et al., 2015]) couples the energy and land use components of the GCAM integrated assessment (IA) model with the atmosphere, ocean, and land components of an Earth System Model (ESM). Within this modeling framework, carbon, climate, land use, and feedbacks to land use can be isolated and the contribution of each to the global carbon cycle quantified. A new set of simulations and methodology to estimate the contribution of carbon, climate, land, and feedbacks to the carbon cycle is developed for the Energy Exascale Earth System Model (E3SM). 

The Integrated Earth System Model (iESM) couples the human economic and energy components of the Global Change Assessment Model (GCAM, www.globalchange.umd.edu/gcam) with the physical, hydrological, and biogeochemical components of the Community Earth System Model (CESM, http://www.cesm.ucar.edu/). The iESM represents a major new model capability that permits the exploration of process-level interactions among human and Earth systems that were previously not represented in the existing suite of computational tools and procedures. The initial version of the iESM focuses on carbon cycle interactions (see Figure 1, [Collins et al., 2015]). Code and input sets are available at: www.github.com/ACME-Climate/iESM.

Additionally, since iESM is developed within an ESM framework, the model retains all of the capabilities of that ESM. As such, the effect of carbon and climate on the carbon cycle can be isolated, replicating experiments like those described in P. Friedlingstein et al. [2006]. Those experiments compare simulations where CO2 is treated as a nonradiatively active gas (i.e., its effect on climate is excluded) but only affects vegetation growth and exchange with the terrestrial system, with simulations where CO2 influences climate.

To isolate these factors, a suite of six experiments was developed (see Table 1). Each experiment uses the results of an RCP8.5 simulation (which is based on high-estimated CO2 emissions) and is run with three different initial conditions, each taken from different points in a long pre-industrial “steady” simulation. In each experiment, some aspects of the experiment were permitted to change, or not, as indicated in Table 1.

Using these simulations, parameters quantifying the contribution of CO and temperature on the carbon cycle can be calculated following the equations described in P. Friedlingstein et al. [2006]. These calculations can be expanded to quantify the contribution of land use and land-use feedbacks on the carbon cycle.