OVERALL PERFORMANCE MEASURES
4th QUARTER METRIC COMPLETED:
Demonstrating Integrated Modeling of Irrigation and Urbanization in E3SM and Weather Research & Forecasting Model (WRF) to Evaluate Their Impact on Regional Precipitation and Temperature Extremes
Product Definition
Urbanization and irrigation, two of the major human activities altering land use and land cover, have expanded significantly over the last decades due to population growth, industrialization and modernization, rising food demand, and innovations in water pumping technologies. Through modifications of the surface energy and moisture budgets, irrigation has major impacts on regional surface temperature and heat stress (Sacks et al. 2009, Yao et al. 2025), cloud formation (Kawase et al. 2008), and precipitation (Lo et al. 2021), with potential remote impacts through changes in large-scale circulation (Zeng et al. 2022). While irrigation induces local cooling and moistening by increasing evapotranspiration, urbanization warms and dries the surface compared to the surrounding rural areas, which amplifies local heat stress and alters regional precipitation patterns. A recent assessment using observations finds that more than 60% of the global cities and their downwind regions receive more precipitation than the surrounding rural areas, and these anomalies are larger in cities with warm and humid compared to cold and dry background environments (Sui et al. 2024). Urbanization may also influence extreme precipitation associated with storms such as hurricanes through the modification of the surface roughness that alters the large-scale circulations and storm tracks (Zhang et al. 2018).
As irrigation and urbanization can influence temperature and precipitation through multiple pathways involving processes across a wide range of scales, models are indispensable tools for understanding, attributing, and predicting their local and remote impacts. However, representing how the local impacts of irrigation and urbanization on surface fluxes upscale to influence the boundary-layer turbulence, convection, clouds, precipitation, and atmospheric circulations, and how these changes provide feedback to alter the surface fluxes, requires kilometer-scale storm-resolving modeling over large domains. Storm-resolving regional simulations using the Weather Research and Forecasting (WRF) model have been demonstrated to improve the fidelity of modeling temperature and precipitation extremes (e.g., Prein et al. 2020). More recently, the Energy Exascale Earth System Model (E3SM) project has demonstrated a new storm-resolving modeling capability through the development of the Simple Cloud-Resolving E3SM Atmosphere Model (SCREAM) (Caldwell et al. 2021, Donahue et al. 2024). SCREAM can be used to perform kilometer-scale simulations globally or regionally through mesh refinement, providing improved capability to model the impacts of irrigation and urbanization.
This report documents efforts to demonstrate the integrated modeling of irrigation and urbanization in both the E3SM and WRF models for evaluating their impact on regional precipitation and temperature extremes. In Section 2, we briefly describe how irrigation and urbanization are represented in the land components of WRF and E3SM. This is followed by a description of modeling experiments designed to use WRF and SCREAM for coupled land-atmosphere simulations at cloud-resolving resolutions to evaluate the impacts of irrigation and urbanization on temperature and precipitation extremes over regions of the contiguous United States (CONUS). As described in Section 3 on the results of the modeling experiments, the analysis of temperature focuses on heat stress and urban heat island (UHI) effects. For precipitation, our analysis focuses on precipitation produced by large thunderstorms called mesoscale convective systems (MCSs) and their counterpart, isolated deep convection (IDC), both of which contribute substantially to extreme precipitation in the central-to-eastern United States. The impact of urbanization is also evaluated using SCREAM experiments that include projected urban expansion, focusing on precipitation patterns and extremes associated with Hurricane Isaias that influenced the eastern United States.
Analyses of the modeling experiments reveal that for temperature, irrigation significantly reduces daytime heat stress in urban clusters, lowering peak air temperatures and increasing humidity during the hottest hours. Urbanization, in contrast, mostly intensifies temperature extremes, particularly at night, with urban areas experiencing increased nighttime heat stress and higher daytime maximum heat index values. For storms and precipitation, irrigation enhances precipitation from both MCS and IDC, while urbanization reduces precipitation due to atmospheric drying. During Hurricane Isaias in August 2020, urban expansion generally enhanced extreme precipitation in expanded urban clusters and their original urban cores. Overall, the two models agree on the contrasting impacts of irrigation and urbanization on heat extremes and precipitation, except SCREAM produced more variable impacts, likely due to its shorter simulations. These results demonstrate the capability of WRF and SCREAM for integrated modeling of urbanization and irrigation effects and underscore the complex interplay between urbanization, irrigation, atmospheric dynamics, and thermodynamic environments in influencing heat and precipitation extremes across the United States.
Product Documentation
This report describes how urbanization and irrigation are represented in WRF and E3SM through their land model component, along with the modeling experiments using the coupled land-atmosphere configurations at storm-resolving resolutions to evaluate the impacts of urbanization and irrigation on regional precipitation and temperature extremes in the United States. It includes:
- Representation of Urbanization and Irrigation in WRF
- Representation of Urbanization and Irrigation in E3SM
- WRF and E3SM Modeling Experiments
- Defining Urban Clusters from Model Simulations
3rd QUARTER METRIC COMPLETED:
Demonstrating the Influence of Flooding on Coastal Development Patterns Using a DOE-Developed Agent-Based Model Driven by Projection of High-Resolution Flooding
Product Definition
Communities along the U.S. coastline, including the mid-Atlantic region, face risks of flooding from heavy rainfall, storm surge, river flooding, and compound events (Bates 2021, Mayo 2022). These hazards and homebuyers’ perception of hazard exposure (Kim 2020) can potentially shift the population among coastal communities (Shu 2023). Understanding time-evolving flood risk and its influence on coastal development patterns requires accounting for a variety of natural system processes, human activities, and complex interactions between the two.
Although a variety of tools are available for simulating flooding and the exposure of infrastructure and other human systems to flood hazards, far less attention has been paid to developing models capable of representing how human societies respond to flood risks. There are no tools capable of fully capturing the two-way feedbacks between human and natural systems that ultimately govern long-term vulnerability (for example, the so-called “levee effect” where flood risks actually increase after levees are built as a consequence of more homes being built behind them). The Coupled Human-Natural City Evolution – Coastal (CHANCE-C) model has been developed with support from the U.S. Department of Energy (DOE) to advance our overall understanding of flood risk, human response, and their interactions within human-Earth systems. CHANCE-C is an agent-based model (ABM) of urban development patterns that integrates representations of the urban environment, flood hazards, household agents, developers, and a housing market. CHANCE-C is unique, in part, because it imbues individual agents with different representations of risk perception and levels of risk tolerance. It is also capable of assessing relative market-driven shifts in real estate value that result from the interplay between risk, perception, and response.
Two case studies are included in this report to demonstrate the ability of CHANCE-C to simulate the influence of flooding and risk perception on coastal development. These initial applications of CHANCE-C use ultra-high-resolution (10-m) simulations of flooding produced by the Rapid Infrastructure Flood Tool (RIFT). They focus on the mid-Atlantic region and, in particular, Baltimore, Maryland, which is susceptible to both pluvial and coastal-driven flood risks. The first case explores how structural and parametric uncertainty in modeling homebuyer behavior influences projections of urban development patterns, including overall population distributions, demographic shifts, and home price dynamics. The second case examines the feedback between homebuyer risk perception and flood risk reduction strategies, using levee development as a representative case to demonstrate how the interplay between the two can affect the housing market and shift both near- and longer-term flood risks. Case studies such as these can leverage a range of flood information, such as projected 100- or 500-year flood hazard maps or flood depth distributions derived from historical tide and river gauge data. CHANCE-C is designed to be extensible so future case studies would integrate high-resolution simulations from the Energy Exascale Earth System Model (E3SM).
This report demonstrates how flooding influences coastal development patterns through the application of the DOE-developed CHANCE-C agent-based model. The case study results highlight how CHANCE-C captures homebuyer risk perception and the feedbacks between these perceptions and flood risk reduction strategies, which can potentially lead to changes in population distribution, demographic profiles, household income distribution, and home prices. With long-term simulations, CHANCE-C provides a robust means to examine the impact of different flood risk reduction strategies on urban development patterns and risk outcomes over time, linking urban and earth systems at local to regional scales. While the demonstration in this report focuses on the mid-Atlantic, the CHANCE-C capability is readily transferable to other geographic regions and hazard types to examine the interplay between risk and urban response.
Product Documentation
This report features the CHANCE-C model, driven by an ultra-high-resolution (10-m) simulation of flooding produced by RIFT for a design storm over the Mid-Atlantic region. This section provides an overview of the CHANCE-C agent-based model, the structural formulations of homebuyer risk perception, and behavioral responses. We then describe the experimental design of two case studies to demonstrate CHANCE-C’s ability to simulate the influence of flooding on coastal development patterns.
2nd QUARTER METRIC COMPLETED:
Demonstrating Modeling of Urban Hydrology and Flooding Using the E3SM Land-River Model with a New Urban Hydrology Module
Product Definition
Rapid global urbanization has fundamentally modified hydrologic processes by altering hydrologic conditions (e.g., imperviousness, soil compaction) and intensifying extreme weather events. Combined with the evolving environments, the impact of urbanization may be amplified (Grimm et al. 2008). These changes significantly impact urban hydrologic cycles and often lead to increased storm runoff and flood risks. Below-ground urban stormwater networks (BUSNs) are essential for mitigating urban floods, underscoring why their modification—primarily through upgrades or duplication, such as replacing pipes with larger ones—remains a predominant, albeit costly and disruptive, flood mitigation strategy (Burns et al. 2015; Chocat et al. 2021; Argue and Pezzaniti 2012). Using observations, previous studies have provided evidence of the hydrologic impacts of urbanization at the regional scale and decadal and longer timescales (Yang et al. 2013; Miller and Hutchins 2017). Through numerical experiments, Earth system models (ESMs) are important tools to further our understanding of the various pathways of how urbanization impacts hydrologic processes and its potential to exacerbate or mitigate flooding in an evolving environment with increasing extreme precipitation. However, current ESMs lack the spatial resolutions needed to resolve fine-scale heterogeneity in urban environments (e.g., completely missing BUSNs and their interactions with urban surface hydrologic processes). This oversimplification leads to inaccuracies in predicting urban floods and other urban hydrologic processes such as infiltration, evapotranspiration, and groundwater recharge.
Hydraulics-based models are invaluable for detailed urban hydraulic and hydrologic simulations. The integration of storm sewer networks into hydrologic-hydraulic models is essential for accurately representing urban flood dynamics, as these networks play a pivotal role in managing stormwater runoff and mitigating flood risks (Guo et al. 2021). In these models, BUSNs are represented at the level of individual stormwater pipes, with water and other fluxes simulated using hydraulic equations. Because of the intensive data requirements, expensive computational costs, and site-specific parameters not being easily transferable, hydraulics-based models are impractical for use in representing urban hydrology in ESMs.
Recently, a new modeling strategy was developed to represent BUSNs directly at the network level, capturing their main hydrologic functions instead of modeling individual pipes and their interactions. This strategy features a BUSN parameterization based on Graph Theory to delineate storm-sewersheds within a watershed, derive sub-BUSNs associated with each storm-sewershed, and determine the BUSN-relevant parameters a priori (Chegini and Li 2022). This modeling strategy has been implemented as a new urban module that is physically based, parameter-parsimonious, and computationally efficient within the Model for Scale Adaptive River Transport (MOSART-urban) (Chegini et al. in press), holding promises to be applied at large spatial and long temporal scales without losing sight of below-ground urban hydrologic processes.
In this report, we first demonstrate the MOSART-urban modeling framework in nine representative watersheds exhibiting various levels of urbanization, ranging from natural to highly developed watersheds in the Houston metropolitan area. As a first step to integrate MOSART-urban into the Energy Exascale Earth System Model (E3SM), we further demonstrate the performance of MOSART-urban driven by the runoff simulated by the calibrated E3SM Land Model (ELM) in simulations across river basins in the contiguous United States (CONUS). Validation against observed daily streamflow shows that MOSART-urban can capture small-to-large flood peaks and seasonal and annual water balance over the nine watersheds in the Houston area, with better performance compared to the National Water Model. MOSART-urban simulations driven by the runoff simulated by ELM show good performance in ~65% of the 299 watersheds across CONUS, with an urban fraction > 30%. Comparisons of MOSART-urban with and without BUSNs for both the simulations over the Houston metropolitan area, a region chosen for its diverse landscapes and a rich history of floods, and across CONUS demonstrate the flood-mitigation capacity of BUSN, which increases with the urban area coverage in the watersheds. MOSART-urban bridges the gap between detailed hydraulic and large-scale hydrologic models, providing a valuable tool for urban flood prediction and management across broader spatial and temporal scales and for Earth system modeling of the impacts of urbanization on the regional and global water cycles.
Product Documentation
This report documents the MOSART-urban modeling framework for representing network-level storm drainage pipes and urban hydrologic processes. MOSART-urban was first validated in the Houston metropolitan area. To test MOSART-urban within E3SM, ELM was calibrated using observed groundwater table depth and inundated surface water fraction data to constrain the simulation. The ELM-simulated runoff was then used to drive MOSART-urban simulations with and without BUSNs to isolate their impacts on streamflow and flood peaks in CONUS river basins with at least 30% urban coverage.
1st QUARTER METRIC COMPLETED:
Demonstrating Use of E3SM With an Urban Canopy Parameterization for Modeling Urban Impact on Local-to-Regional Heat Extremes
Product Definition
Urban areas exhibit distinct biophysical, morphological, and thermodynamic characteristics that influence local and regional weather and climate. These impacts include changes to the surface energy budget, near-surface meteorology, atmospheric composition, hydrological cycle, energy systems, and carbon cycle (Arnfield 2003, Qian et al. 2022). Urbanization disproportionately affects local populations by altering weather, particularly heat extremes (Tuholske et al. 2021). However, limited observations and model deficiencies hinder the understanding of urban systems (Grimmond 2011, Muller et al. 2013). Despite advancements, key aspects of urbanization, such as spatial variability and temporal evolution, remain poorly captured in models, especially Earth system models (ESMs), which are essential for studying climate processes (Chakraborty and Qian 2024). As we move from rural to suburban to urban regions, we expect sharp gradients in temperature, humidity, wind, and cloud cover that ESMs at coarse resolutions cannot resolve. The challenges of accurately representing within-city variability, relevant for examining community-scale heat hazards, highlight the need for improving ESM modeling capabilities to address uncertainties in urban environments (Sharma et al. 2021).
Most ESMs have limited or no representation of urban areas due to a legacy focus on large-scale climate impacts and coarse model grids that overlook urban processes (Zheng et al. 2021). The U.S. Department of Energy’s (DOE) Energy Exascale Earth System Model (E3SM; Golaz et al. 2019) is one of the few exceptions, incorporating an urban canopy model from the Community Land Model (CLM) using an "urban canyon" approach (Oleson et al. 2010) that represents roofs, walls, and canyon floors (Figure 1a). A global urban surface biophysical data set (Jackson et al. 2010), critical for constraining the surface energy budget, is also embedded in the E3SM Land Model (ELM). However, this model has two major limitations.
First, the urban surface data set in ELM is coarse-grained and outdated. Urban areas are categorized into three density classes across 33 global regions, with uniform radiative, morphological, and thermal properties that fail to capture real-world variability. For example, the prescribed albedo values differ significantly from satellite observations, which reveal greater variability within and across these 33 regions (Figure 1b). Poorly constrained urban parameters have a larger impact on model performance than model complexity, as evidenced by discrepancies in urban heat island (UHI) intensity simulations (Chakraborty et al. 2021, Grimmond 2011). Second, it neglects a key feature of many urban areas by not considering any vegetation within the urban canyon. As such, urban areas are treated as biologically inactive, with pervious surfaces modeled as bare soil. This neglects the critical role of urban vegetation in influencing temperature, air quality, and the water, energy, and carbon cycles within cities (Paschalis et al. 2021). In the context of extreme heat, it is well established that urban greenery mitigates daytime heat hazards (Ziter et al. 2019), a key requirement for effective climate adaptation in a warming world. While meso- and micro-scale models have advanced to include urban vegetation explicitly (Krayenhoff 2021), ESMs have lagged, often relying on outdated tiled approaches that cannot resolve spatial variability within urban areas (Krayenhoff 2020).
With ESMs now running at finer spatial resolutions, these deficiencies become more critical. For example, even when the Simple Cloud Resolving E3SM Atmosphere Model (SCREAM; Caldwell 2021) is run at km scale, the land cover constraints are still coarse, thus underestimating near-surface heterogeneity. High-resolution modeling, in theory, can reveal urban heat hotspots, UHI patterns, and urban-induced cloud formation (Theeuwes et al. 2019), aiding urban planning, isolating communities vulnerable to urban environmental stressors (Chakraborty et al. 2023), and guiding stormwater management strategies. However, SCREAM would still underestimate variability within and across cities due to its simplified representation of urban areas. To accurately capture urban climate signals, models must incorporate spatially continuous and biologically active urban representations, particularly for fine scale simulations. This would improve insights into urban impacts and enhance planning for climate adaptation and resilience.
This report documents recent improvements in ELM and the ELM urban model through the Integrated Coastal Modeling (ICoM) project and a DOE Early Career project that advance surface constraints for high-resolution urban-resolving E3SM simulations. We explore these improvements and their implications for urban heat and heat stress using both land-atmosphere coupled SCREAM simulations at 3.2 km and ELM land-only simulations over the United States for a heatwave period in July 2020. The coupled simulations run with and without spatially explicit urban surface data sets show improvements in capturing urban heat signals and their spatial and diurnal variability, while the land-only simulations allow us to estimate the sensitivity of urban heat to radiative and morphological urban parameters.
Product Documentation
This report documents the modeling of urban heat extremes in the contiguous United States using SCREAM with regional refinement capability. To accommodate km-scale modeling, we update the surface data of SCREAM based on several 1-km data sets to resolve spatial variability more realistically.
High-Resolution Land Cover Constraints
A global 1-km data set of land surface parameters was developed by Li et al. (2024) as part of the ICoM project using a combination of data sources. This data set, covering the years 2001 to 2020, provides higher-resolution information and more spatial variability of land use and land cover (LULC), vegetation properties, soil properties, and topography than conventional coarse-resolution data sets. Specifically, we update organic matter, clay, and sand percentages in soil, lake, and glacier percentages, the fractions of different plant function types in natural vegetation, leaf area index, stem area index, and canopy height information in the SCREAM surface data.
Global Spatially Explicit Urban Biophysical Properties for Urban Scale Modeling
The urban component of E3SM requires facet-level (roofs, walls, roads, etc.) properties critical for constraining the surface energy budget and anthropogenic signals. Recent work supported by a DOE Early Career project has developed U-Surf, a new global data set of urban surface parameters at a 1-km resolution (Figure 2b). U-Surf leverages high-resolution satellite remote-sensing data, machine learning techniques, and planetary-scale geospatial analyses on cloud computing platforms to generate a comprehensive and internally consistent set of biophysical parameters for urban areas worldwide. These parameters include radiative, structural, and thermal properties, aligning with the structural requirements of various urban canopy models, particularly E3SM (Figure 2). This compatibility enables realistic comparisons of urban climate signals both within and across cities. More details can be found in Cheng et al. 2024.
We update urban properties in the SCREAM surface data using U-Surf except for urban fractions, which are from another global 1-km data set of annual urban dynamics between 1870 and 2100 developed by Li et al. (2021), who trained an urban cellular automata model using satellite observations of urban extent between 1992 and 2013 and ran the model to simulate urban dynamics before 1990 and after 2020 under five Shared Socioeconomic Pathways (SSPs). The data set thus provides temporally continuous self-consistent high-resolution urban coverages, which is helpful in studies relevant to urban expansion and shrinkage. Urban fractions in 2010 from the data set are used to update our SCREAM surface data.
Regionally Refined Simulations for Km-Scale Urban-Resolving Continental Simulations
Although the update of LULC percentages, soil properties, and vegetation parameters in the surface data can also influence the simulation of heat extremes, its comparison with the default ELM surface data set has already been performed by Li et al. (2024). In this report, we use this data set in all the simulations and focus on understanding how updated urban properties from U-Surf can alter the heat extreme simulation since these properties are more relevant to the urban model. We conduct two atmosphere-land coupled SCREAM simulations, one with the default urban properties (hereafter named the default simulation) and the other with U-Surf urban properties (hereafter named the U-Surf simulation) to investigate the potential impact of updated urban properties on the simulation of a heat wave that occurred in late July 2020 in the eastern United States. The simulation period is from July 18 to July 30 with the first day as spin-up. To reduce the computational cost, we run SCREAM on a regional refinement mesh (RRM) with a horizontal resolution of ~3.2 km over the contiguous U.S. (CONUS) and ~100 km for other areas globally (Figure 3a). The atmospheric initial condition is based on a combination of the hourly 3-km High-Resolution Rapid Refresh (HRRR) and 25-km ERA5 reanalysis data sets. The land's initial condition is obtained from a 10-year land-only simulation on the same SCREAM RRM constrained by atmospheric forcing in 2020 from ERA5. The SCREAM simulation uses prescribed hourly sea surface temperature (SST) from ERA5. Three-dimensional zonal and meridional winds, temperature, and specific humidity above 850 hPa are nudged towards hourly ERA5 reanalysis with a relaxation timescale of six hours for the coarse-resolution grids with equivalent physical grid spacing larger than 5 km. Finally, we regrid the coupled simulation outputs to 4 km to facilitate comparison and analysis.
Besides the above two coupled simulations, we run five additional land-only simulations to resolve the distinct impacts of facet-level albedo, facet-level emissivity, and morphological (roof height, roof fraction, pervious road fraction, height of wind in canyon, and canyon height to width ratio) urban parameters from U-Surf on the heat extreme simulation (Table 1). The land-only simulation period is from July 19 to July 30, and the simulations are constrained by atmospheric forcing from the default coupled simulation. The restart file at 0:00Z on July 19 from the default coupled simulation is used as the initial condition for the land-only simulations except for the “All New Properties” simulation, which uses the restart file from the U-Surf coupled simulation due to different urban levels between the default and U Surf urban property data set.