Susquehanna River Basin Experiment (SRBEX)

Introduction

SRBEX is the first of several planned regional experiments which are to be undertaken in our EOS research program. The Susquheanna River Basin (SRB) is a 62,419 km sq watershed covering portions of New York, Pennsylvania, and Maryland (Figure 1). The overarching goals of SRBEX are to understand the hydrologic cycle of the basin through modeling and the analysis of observed data, and as result of this understanding to develop capabilities for monitoring and projecting changes in basin hydrology. The reason for SRBEX is clear: accurate projections mean that managers will be able 1) to lessen the impacts of variations in water quality and supply and 2) to optimize water resources. Thus, significant social and economic costs can be avoided and benefits can be realized by the research efforts of SRBEX.

More specifically, SRBEX includes the following policy-relevant objectives:

  1. Identification of hydrologic parameters sensitive to climatic variation and change
  2. Model experimentation to provide quantitative comparison of the hydrologic system sensitivities to various natural and human forcing functions
  3. Observational studies to provide the context for comparisons of natural versus human-induced variations
Figures 2 and 3 summarize the SRBEX suite of databases, data sets, and linked models. These models include: the GENESIS GCM, the Penn State/National Center for Atmospheric Research (NCAR) Mesoscale Model (MM), the Soil Hydrology Model (SHM), the Pennsylvania State University Biosphere-Atmosphere Model Scheme (PSUBAMS), the Terrestrial Hydrologic Model (THM), and the Water Quality Model (WQM). The linkages between the different models are under development and/or validation.

Coupled Global-Regional Climate Model Simulations

A primary goal of SRBEX has been to link global-scale climate change to regional climate change over the SRB with a particular emphasis on the hydrologic component. However, the coarse resolution of current GCMs, which range from 3 to 5 degrees in latitude and 3 to 7.5 degrees in longitude, limits our understanding of regional climate change. Important land features such as vegetation, orography and complex coastlines are poorly resolved with current GCMs. In order to address these issues we have undertaken a number of coupled experiments aimed at reducing the uncertainties associated with the coarse resolution of GCMs. In these coupled experiments a limited area model is embedded within the GCM (Dickinson et al., 1989). This technique has subsequently been used to simulate present-day and elevated atmospheric CO2 simulations over the United States and Europe (Giorgi, 1991; Marinucci and Giorgi, 1992; Giorgi and Marinucci, 1992; Giorgi et al., 1993a,b,c; Giorgi et al., 1994).

Model Description and Experiments

The GCM used in this study is GENESIS (Thompson and Pollard, 1995) which has a horizontal resolution of 4.5 degrees of latitude by 7.5 degrees of longitude and 12 vertical levels. It uses a land-surface package (LSX) in which exchanges of heat, momentum and moisture fluxes between the biosphere and atmosphere are computed on a 2 x 2 degree grid. The regional climate model is known as the RegCM2 (Giorgi et al., 1993a,b,c) and is based on the Penn State/NCAR Mesoscale Model Version 4, MM4 (Anthes et al., 1987). Various components of RegCM2 have been altered (solar and infrared radiation calculation, land surface package, PBL scheme) from the original MM4 to make it compatible for undertaking long-term model integrations.

We have undertaken several sensitivity studies which take information from GENESIS, with a horizontal grid spacing of approximately 500 km, to drive RegCM2 with a horizontal grid spacing of 108 km. The time period of these integrations is 1979-1988. Time-varying observed sea surface temperatures (SSTs) during this period were used for these simulations.

GENESIS output was used to drive the nested RegCM2 at its lateral boundaries. The purpose of these experiments was to examine the nested RegCM2 sensitivity to domain size and to compare output from a long-term integration (1979-1988) to the stand-alone GENESIS output and observations. In the first simulation the domain of RegCM2 covers the Eastern two-thirds the United States, while in the second and third simulations the domain covers the entire United States, the boundaries extending from the Pacific to the Atlantic and from Mexico to Southern Canada. Completed and future simulations are shown in Table 1.

GENESIS 1979-1988 AMIP Simulation Results

The stand-alone GENESIS simulation overestimates rainfall considerably during the 1979-1988 winter and summer seasons when compared to observations. During the winter season the largest overestimated precipitation occurred over the Great Plains, Southern Atlantic coastal regions and the Rocky mountains. During the summer season, precipitation amounts were overestimated for most of the US (Figure 4). The GENESIS interannual variability of precipitation patterns and amounts during winter and summer seasons were smaller than the observations.



The GENESIS large-scale circulation features 200 mb zonal wind and 500 mb heights did not compare favorably with the observations. In particular, the GENESIS 200 mb winds were displaced poleward during the winter season and equatorward and westward of the observed position during the summer season. The wind speeds during the winter season were overestimated by 10 m/s. The 500 mb geopotential patterns were displaced westward of their observed positions during the summer and winter seasons. The GENESIS winter season Pacific ridge was persistently located 10-20 degrees west of its observed position, while the summertime continental ridge was located closer to its observed positions, but still westward by 5-10 degrees. Furthermore, a trough at 500 mb persisted over the central and eastern US in GENESIS during the 1979-1988 summer seasons. The GENESIS summertime displaced ridge led to a failure to capture the drought events of 1980 and 1988. The overall weaknesses of GENESIS in simulating the large-scale circulation had negative implications for the nested-grid simulation results which are discussed below.

Domain Sensitivity Simulation Results

The deficiencies of the stand-alone GENESIS GCM in simulating the climate over North America prompted us to evaluate how well the nested GENESIS/RegCM2 simulations would characterize the winter and spring periods of 1979 and 1980. An overriding concern was that errors in the stand-alone GENESIS simulation would dominate the coupled simulation as reported by Giorgi et al. (1994). Therefore, we undertook two simulations in which the domains included only the Eastern US and the entire continental US.

The nested simulation using the eastern US as the domain was strongly affected by the driving climate of GENESIS. The differences in the large-scale circulation between the nested simulation and the stand-alone GENESIS over the domain were small for the six month period (Jenkins and Barron, 1995). Improvement in the nested simulation sea level pressure, however, occurred during this period when compared to observations. Further, synoptic weather disturbances had a resolvable structure compared to GENESIS results. Indeed, in some cases, weather disturbances which tracked northeast along the Atlantic Coast were rain events in GENESIS while snow events in the nested RegCM2 simulation.

The nested RegCM2 simulation using an extended domain produced favorable distributions of precipitation compared to GENESIS and observations during both the winter and spring of 1979-1980 (Jenkins and Barron, 1995). Regional features, such as the Pacific Coast floods during the winter of 1979-80 and the springtime drought of the Northern Rocky Mountains and Great Plains were captured in the nested simulation with the extended domain. Even with this success, we found that the large-scale circulation had similar structure to the driving GENESIS. As with the stand-alone GCM, the nested simulated 500 mb geopotential heights were displaced westward of the observed position.

1979-1988 Nested GENESIS/RegCM2 Simulations

Recently, we extended the integration of the nested simulation with the expanded domain from the last section. The simulation was integrated out to 31 December 1988. Similar to the results above, we found an improvement in the nested model simulated precipitation during the 1979-1988 winter and summer season precipitation (Figure 4). However, the large-scale circulation continued to show similar characteristics to the stand-alone GENESIS. We have reason to believe that large-scale stationary waves that have wavelengths which are equal to or larger than the continental domain, if not properly simulated in the driving GCM, will introduce sizeable errors in nested simulations. A good example of this is the westwardly-displaced 500 mb simulated ridge which persists in the RegCM2 simulations. The displaced ridge in the nested RegCM2 simulation is caused by errors in the geopotential heights from the GENESIS GCM which are read in at the western lateral boundary of RegCM2. It is not just the western boundary which causes the problem, but the northern and eastern boundaries also supply any error-prone GENESIS geopotential heights to the nested simulation. These boundaries act to constrain the stationary wave pattern of the nested simulation to that which is specified at the lateral boundaries.

The nested RegCM2 simulations for 1979-1988 also produced a winter season cold bias at the surface which extends into the lower troposphere over all of the continental US. The cold bias appears to be related to very high cloud fractions in the nested RegCM2 simulations. The cloud fractions are generated by a simple cloud parameterization which is based on relative humidity. Currently we are running several sensitivity tests which use higher relative humidity thresholds to reduce winter season cloud fractions. The cold bias also occurs in a few regions (Rocky Mountains and Northern Atlantic Coast) during the summer season. Excluding the few cooler regions, RegCM2 produces a slight warm bias except along the Pacific Coast where the warm bias is much larger. This is due to a lack of simulated clouds along the Pacific Coast in both the RegCM2 and the GENESIS simulations. Summertime clouds along the Pacific Coast are diurnal in nature and the cloud parameterizations in both models are too simple to capture this feature.

The nested RegCM2 simulation does capture some of the extreme climate events during the 1979-1988 time period. In Figure 5, a comparison of observed 1980 and 1988 summer air temperatures are shown. During both of these periods, the Northern Rocky Mountain and Great Plains regions experienced drought and above normal temperatures. Both precipitation (not shown) and air temperatures are better captured in the nested experiments than in the stand-alone GENESIS simulation.

During the summer of 1988, the temperature maximum is located east of the observed maximum in both simulations, but more eastward in the stand-alone GENESIS simulation. The eastward displacement in high temperatures in the lower troposphere appears to be related to the advection of warm air into this region from the Gulf of Mexico and the southwestern United States. In contrast, the ECMWF analyses show that the primary source of warm air over the Northern Great Plains was advected from the Gulf of Mexico and Southern Great Plains regions. Further, the source of warm air over the Northern Rocky Mountain region during this period was from the Pacific and the mean wind direction in the lower troposphere had a southwesterly component. In contrast, the mean wind flow into the Northern Rocky Mountains had a northwesterly component out of Canada in the GENESIS simulation.

Current Work and Future Plans

The following tasks are currently being undertaken: (1) An analysis of the surface hydrologic cycle over the 9-year period in the nested RegCM2 simulation and the stand-alone GENESIS simulation. These results will be reported soon. (2) A RegCM2 simulation driven by European Center for Medium-Range Weather Forecasting (ECMWF) analyses over a 1-year period to evaluate the errors of the nested simulation caused by weakness in the GENESIS large-scale circulation. (3) Cloud parameterization sensitivity simulations of 1- month duration.

Future plans with GENESIS are to begin mixed-layer ocean simulations with the nested GENESIS /RegCM2. The new simulations may take place with the current R15 version or the new T31 version of GENESIS. First, the large-scale circulation of the T31 version must be evaluated against the R15 version. During the coming year we will evaluate the synoptic circulation of the 9-year nested RegCM2 simulation and the stand-alone GENESIS simulation using the techniques of Frakes and Yarnal (1995a). We will compare observed sea-level pressure patterns which lead to high discharge rates over the SRBEX region to model simulated sea-level pressure patterns which lead to high precipitation and runoff rates over the same region. Finally we will drive the Soil Hydrology Model (SHM), described later in this report, with GENESIS and/or RegCM2 meteorological data over several regions and compare the GENESIS and/or RegCM2 surface hydrology to the SHM-predicted surface hydrology.


Table of Contents Previous Section Next Section Download Report
Graphics and HTML text markup by Phil Kolb
Comments to: R. White, raw@essc.psu.edu