Linked atmosphere-hydrology models at the macroscale

Macroscale Modelling of the Hydrosphere (Proceedings of the Yokohama Symposium, July 1993). IAHS Publ.no. 214,1993. 3 Linked atmosphere-hydrology models at the macroscale C. J. VÔRÔSMARTY Institute for the Study of Earth, Oceans, and Space, University of New Hampshire, Durham, Hampshire 03824, USA New W. J. GUTOWSKI Department of Geologic and Atmospheric Sciences, 3010 Agronomy, Iowa State University, Iowa 50011, USA Ames, M. PERSON Department of Geology and Geophysics, Minneapolis, Minnesota 55455, USA University of Minnesota, 310 Pillsbury Drive SE, T.-C. CHEN Department of Geologic and Atmospheric Sciences, 3010 Agronomy, Iowa State University, Iowa 50011, USA Ames, D. CASE ARC, Inc., 8201 Corporate Drive, Suite 1120, handover, Maryland 20785', USA Abstract In this paper we discuss numerous aspects of macroscale hydrologie models and their linkage to atmospheric simulations. We discuss spatial and temporal scaling and conclude that the region is the most critical scale for addressing global change issues. Macroscale hydrologie models might reasonably be considered as a hierarchical, nested system of models focusing on the mesoscale. The nested approach requires a GCM, a mesoscale atmospheric model, a soil-vegetation-atmosphere transfer scheme, and surface/subsurface hydrology models, the latter down to hillslope scale in small representative catchments. A hydrologie analysis system manages input/output functions, scientific visualization, model preparation, and execution. The entire system of data, models and computing tools is being developed to distill the facets of global change to the regional level and to explore feedbacks from the land surface to the atmosphere. INTRODUCTION The emerging field of Earth System Science has fostered the development of large-scale, synoptic views of the Earth's hydrology. This capability is particularly important given a growing consensus within the scientific community of imminent and unprecedented change in the Earth's climate. Although the greenhouse effect is considered global in extent, its critical impacts will occur at regional and smaller scales, such as river basins and agricultural zones. The need to plan for possible climatic change creates a profound challenge to our modelling and monitoring capabilities. The development of international programs such as the Global Energy and Water Cycle Experiment (GEWEX) and the International Geosphere-Biosphere Program (IGBP) is in part a recognition of the need to assess, understand and model the response of regions to potential climate change. In this paper we discuss the use of linked models of atmospheric dynamics, surface water balance and riverine transport to assess regional water resources in the context of global change. The paper has three goals. The first is to develop an operational 4 C. J. Vôrôsmarty et al. definition of a Macroscale Hydrologie Model (MHM), with particular emphasis on linkages between atmospheric and land processes. The second is to reveal the utility of such models at regional and continental scales. Our third goal is to offer a modelling strategy for quantifying regional water status, either in a natural state or as influenced by anthropogenic change. MACROSCALE HYDROLOGIC MODELS (MHMs) A recent series of publications (Eagleson, 1986; Becker & Nemec, 1987; Shuttleworth, 1988; Vôrôsmarty, 1991) has defined a broad set of issues regarding macroscale hydrology. The macroscale encompasses a spatial domain characterized by large units of landscape and, for the purposes of addressing hydrological issues, the organizing concept is the watershed. Comparative watershed studies at the local scale have yielded important insights into surface and subsurface hydrology, nutrient biogeochemistry and constituent transport. Researchers constructing MHMs hope to derive similar benefit by considering whole catchments, but over much broader scales. Macroscale hydrologie models simulate water flux in two or three spatial dimensions. These models are derived by combining expressions for the conservation of fluid mass with constitutive flow laws (e.g. Darcy's Law for groundwater flow; Manning's equation for channel discharge). Vertically, they exchange water across the land surface-atmosphere boundary through precipitation, évapotranspiration and recharge to subsurface storage. Mathematically, this results in a series of coupled differential equations for the vertical exchanges of water, energy and momentum. Vertical fluxes are also computed for soil infiltration and the delivery of surplus water to groundwater storage. In the horizontal plane, water is transported laterally using simulated network topologies which route surface and groundwater. The models discretize the spatial and temporal domains, with length scales varying from 50 km (Vôrôsmarty et al., 1989) to 10 km (Solomon et al., 1990) and monthly time steps. A number of studies have attempted to represent watershed transport processes at even finer spatial and temporal resolutions (e.g. Freeze, 1980; Beven & Wood 1983; Loague & Freeze, 1985; Bathurst, 1986), but few (Solomon et al., 1990) have done so in support of macroscale basin studies. Nonetheless, modelling at fine scales to support continental analysis is seen as an important area of research that can capitalize on continuing improvements in our monitoring capabilities (Chahine, 1992; World Climate Research Programme, 1991). With a growing interest in how climate and land use change affect hydrology must come steps to formally integrate models of the atmosphere and land-based water cycle. Land surface - atmosphere interactions are currently embedded within general circulation models (GCMs) and mesoscale atmospheric models through Soil -Vegetation-Atmosphere Transfer schemes (SVATs) of varying complexity (Henderson-Sellers, 1992). Although these models estimate time-varying runoff, they have seldom been cast in a hydrological context (Miller & Russell, 1992; Kuhl & Miller, 1992). A linked Atmosphere-MHM (A-MHM) would necessarily require a shared S VAT scheme, but would specifically generate and transport runoff laterally via overland flow paths, groundwater/stream interactions and open channel networks. Interactive models of this sort can be cast to capture the behaviour of drainage basins within the context of a dynamic climate system, possibly uncovering feedbacks not Linked atmosphere-hydrology models at the macroscale 5 otherwise detectable using hydrologie simulations driven by predetermined meteorological fields. The incorporation of carbon and nutrient cycling into such models (McGuire et ah, 1992; Running & Coughlan, 1988) provides even more promise for understanding how terrestrial ecosystems are coupled to the climate system. UTILITY OF MHMs AT REGIONAL AND CONTINENTAL SCALES Regional and continental-scale hydrologie models have successfully quantified the balance between precipitation, évapotranspiration and runoff, and hence are of use in a wide spectrum of water resource assessments. For example, they have been used to quantify water availability for agriculture, flood hazard, hydroelectric potential and sediment transport (Beran et ah, 1990; Solomon et al., 1990). Their predictions of water status have been used as inputs to terrestrial ecosystem models (Raich et al., 1991; McGuire et al., 1992). These models are also being formulated to improve land surface-atmosphere couplings to GCMs, thereby permitting a more direct assessment of the potential impacts of climate change. Such simulations can also be used to validate climate model predictions of the terrestrial water balance. Long-term predictability is a standing challenge in the hydrologie sciences (Alley, 1984). Retrospective analyses of palaeoflood hydrology (Kochel & Baker 1982; Hupp, 1982; Stedinger & Baker, 1987), palaeolake levels (Winter & Wright, 1977; Engstrom & Nelson, 1991; Fritz et al., 1991), and distributions of Holocene sediment facies (Kolterman & Gorelick, 1992; Keen & Shane, 1990) in tandem with MHMs can be used to explore long-term variability of the water cycle. An MHM was recently developed (Vôrôsmarty et al., 1989) as part of a larger effort to study the impacts of human activity on global water, nutrient and carbon cycling (Moore et al., 1989). Experiments using this model are instructive with respect to the promise and challenges of MHMs. The simulation is composed of a linked Water Balance and Water Transport Model (WBM/WTM). Both models operate at monthly time steps and discretize their domain using a 0.5 x 0.5 degree grid. The WBM converts information on land cover, soils and climate into predictions of soil moisture, évapotranspiration and runoff. The WTM routes the runoff using a simulated network topology derived from maps and calculates rates of water transport between adjacent grid cells. Corrections can be made for floodplain inundation and the impact of engineering works. The first test of the model was at the continental scale in South America, and its water balance component performed well in predicting the overall distribution and magnitude of continental runoff. In the Amazon Basin, the discharge algorithm predicted the timing and magnitude of river flow at a series of gauging stations. At a downriver site, representing the integrated regional water balance, predictions were within 1 % of observed values on an annual basis. Tests on another tropical river, the Zambezi in southern Africa (Vôrôsmarty & Moore, 1991; Vôrôsmarty et al., 1991a), were more problematic, and the model consistently overestimated runoff in a series of regional subcatchments. Konikow (1986) discusses a similar problem in modelling groundwater status in the western US, as did Alley (1984) attempting to predict the long-term (50-year) record of stream flows in ten watersheds in New Jersey (USA). 6 C. J. Vôrôsmarty et al. Difficulties in establishing an accurate water balance for the Zambezi catchment suggest that caution must be exercised in applying generic macroscale models. These experiments indicate that some level of site-specificity needs to be built into any macroscale analysis. Models must therefore be developed to explicitly treat basin-specific phenomena such as the shifting patterns of land use, evaporative losses by swamps, seasonal floodplain inundation, the role of riparian vegetation, impoundments and other engineering works. The experiments on the Zambezi also indicate the possibility that monthly time steps and coarse grids cannot adequately capture the dynamics of such heterogeneous environments, and more detailed physically-based analysis may be warranted. Furthermore, although much effort can be expended in developing observational data sets to provide model inputs as well as calibration targets, the error propagation associated with this information cannot always be established (Vôrôsmarty, 1991). Climatic fields used to force the model may have the additional problem of interpolation bias (Willmott et al. 1985), exacerbated whenever the original observational network is sparse or when the data inherently possess high spatial or temporal variability. Despite these obstacles, we are confident that the successful development of a validated A-MHM will make an important contribution to our understanding of regional water cycling and the effects of anthropogenic change. Linked atmosphere-hydrology models will enable us to better quantify the interplay between synoptic weather patterns and regional climate. So-called teleconnection patterns in the atmosphere attest to the importance of this interplay. Chen & Kpaeyeh (1992), for example, found that the intrusion of a low level water vapour jet from the Gulf of Mexico into the US Great Plains was in large measure controlled by far-field effects extending to the Rocky Mountains. On a broader scale, El Nino can exert substantial influence over weather and moisture cycling in regions far from its centre of activity in the tropical Pacific (Horel & Wallace, 1981; Ropelewski & Halpert, 1987). We must determine if such broad-scale climatic features either reinforce or interfere with more local recycling of precipitation and évapotranspiration, such as that described by Salati & Vose (1984). In addition, changes in the natural landscape may show marked changes in local recycling. Three independent modelling studies of Amazonia (Dickinson & Henderson-Sellers, 1988; Shukla et al, 1990; Lean & Warrilow, 1989) showed a weakening of the recycling processes with widespread deforestation. Although we have some general understanding about how synoptic and local conditions affect the regional atmosphere/land surface boundary, their impact on groundwater and river discharge remains an open question. ISSUES OF SCALE A consideration central to developing A-MHM models, is how to represent and then couple individual processes whose respective scales encompass several orders of magnitude in space and time. Figure 1 shows a matrix that conceptually divides the realm of hydrologie modelling into four quadrants. Successful analyses have been performed in two domains: a) empirical or semi-mechanistic models with relatively coarse-scale data and time steps, generating a 'synoptic' view of hydrological processes, and b) physically-based models with fine-scale data and short time steps. The question at hand is how to build upon this legacy to address water cycling issues Linked atmosphere-hydrology models at the macroscale MODEL PHYSICALLY-BASED D FINE <1 km subdally A T y ? x • • COARSE 10-50 km weekly monthly ? • SYNOPTIC V Fig. 1 Conceptual matrix for hydrologie modelling and data utilization. Successful analyses have been performed using physically-based models with fine resolution data and using synoptic hydrologie models with coarse-scale data. Application to the meso or regional scale may require all available combinations of model sophistication and data resolution. at intermediate, regional scales while simultaneously linking land and atmospheric systems. Spatial Scaling The coarsest A-MHM scales are associated with GCMs, which are important for simulating the atmosphere's global circulation and its associated cycles of heat, moisture and momentum. The global circulation patterns interact with surface and subsurface hydrology and link the planet's regions into a global climate system. However, the spatial resolution of current global climate models, typically several hundred kilometres, is too coarse to simulate the climatic forcings of consequence to individual watersheds. Efforts like the US Department of Energy's Computer Hardware, Advanced Mathematics, and Model Physics (CHAMMP) program are underway to increase model resolution, but useful global simulations using a resolution finer than 100 km will require significant advances in physical modelling and computing hardware (US Department of Energy, 1990); only a very limited number of such simulations, if any, are likely to occur before the end of the century. An alternative to increasing the resolution of global models is to use a nested resolution approach within a larger GCM domain (Giorgi & Mearns, 1991). Another is to use specified climatic boundary conditions from either observation or GCM output to drive regional models with sufficient resolution. Gutowski et al. (1993), for example, used GCM-projected changes in hurricane intensity to analyze the effects of global warming on flooding potential in small drainage basins (length scale of several km) in southern Florida. Branscome & Gutowski (1992) used such an approach to explore interactions between global warming and the synoptic-scale dynamics of water and energy cycles. 8 C. /, Vôrôsmarty et al. Atmospheric models with 60 km (mesoscale) resolution have already been used to project climate variability to scales smaller than those resolved by global models (Giorgi & Mearns, 1991). Mesoscale models, such as the MM4 developed by scientists at the National Centre for Atmospheric Research (NCAR) and the Pennsylvania State University, have successfully simulated general atmospheric dynamics down to scales as small as 25 km (Anthes et al, 1987; Anthes, 1990). The recent development of non-hydrostatic mesoscale models should extend this limit down to scales as small as a few kilometres. Such capability raises the question of whether or not a lower limit truly exists on the scales that must be resolved for regional climate simulation. The drainage basin is a useful organizing concept through which to view the coupling of the Earth's atmosphere and land surface hydrology. Numerous watershed studies have added significantly to our understanding of how landscapes function 15 S 25 S 35 S 85 W AVAILABLE 0 50 75 W 65 W WATER CAPACITY (mm) 100 150 200 250 350 Fig. 2 Variation in the frequency distribution of available water capacity in three 5° X 5° grids based on an original 'A0 data set derived from FAO/UNESCO (1977). Sensitivity in the distribution is due both to natural variations in soils and arbitrary positioning of the larger grids. Linked atmosphere-hydrology models at the macroscale 9 hydrologically and how constituents are processed and transported by the water cycle (Alley, 1984; Likens et al 1977; Correll, 1986; Swank & Crossley, 1988; Bartell & Brenkert, 1992; Elder, 1985; Richey & Ribeiro, 1987; Fédérer, 1992). The inherent advantages of such studies should be realized as well at regional and continental scale studies. However watershed and hillslope-scale studies generate characteristic rainfall-runoff relationships that have not yet been systematically incorporated into MHMs. The dynamics, at a minimum, must encompass a quantification of interception, throughfall, soil infiltration, groundwater recharge and discharge both from storm and base flow. Large drainage basins are a complex mosaic of land cover domains, each with characteristic water and energy dynamics. For this reason, any coupled A-MHM must include some measure of surface variability. This variability has been represented as a set of statistical functions (e.g. Entekabi & Eagleson, 1989; Freeze, 1980). However, calculations using observed land surface data (Fig. 2) show that the subgrid distribution of surface and subsurface characteristics can be quite complex. In addition, the structure of the distribution can vary greatly between grid domains. Further, simple experiments using representative 0.5 degree resolution data in conjunction with the WBM/WTM described earlier, show that surface hydrology can be quite sensitive to the loss of surface resolution (Fig. 3). Describing heterogeneity with a priori functions can therefore obscure diversity clearly apparent in the real world. To model surface heterogeneity adequately will require an empirical, data-rich approach similar to that used by Avissar & Pielke (1988), in which important surface classes in a grid are represented by their proportional area coverage. A further challenge is to monitor the changing mosaic of surface properties in large catchments due to the appearance and disappearance of various land cover classes. Clearly, an inventory of land surface cover is essential in determining the appropriate spatial resolution of an A-MHM. Temporal Scaling An inherent problem in simulating the various land surface and atmospheric phenomena associated with an A-MHM is the breadth of characteristic time scales required to move the coupled system of equations to dynamic steady state. Land surface-atmosphere interactions are elicited at a variety of temporal scales (Committee on Global Change, 1990). For the atmosphere, turbulent dynamics in the atmosphere's surface boundary layer occur on scales of seconds to hours, synoptic weather patterns span days to weeks, and the annual cycle of the atmosphere's general circulation is characterized over months, years and decades. For terrestrial ecosystems, metabolic activities associated with growth and maintenance constitute the most rapid interactions, of the order of seconds to days and determine latent heat, energy, water and C0 2 gas exchange through gross photosynthesis and respiration. Intermediate processes, from days to weeks, include the development of leaf area, root systems, soil water balances and runoff generation. Annual and decadal time steps encompass ecosystem production, long-term changes in carbon and nutrient stocks, the growth and succession of plant communities, structural changes in canopies, and the response of local and regional groundwater supplies. These large differences in time scales present operational problems in assigning an appropriate temporal discretization to each of the component subsystems. C. J. Vôrôsmarty et al. 10 30-1—• nui O 5 - 40H 4 16 64 256 NUMBER OF CELLS fH p m •• -.vv AVAILABLE WATER CAPACITY (mm water ) 0 50 100 -C. 125 150 200 350 Fig. 3 The influence of grid resolution on water balance components during a dry to wet season transition using a.simple water balance model (Vôrôsmarty et al., 1989). Each 8° X 8° block maintains the same average water capacity. Recent attempts at enhancing GCM's treatment of the land-surface boundary layer (e.g. the SiB and BATS models) have recognized the importance of terrestrial vegetation in exchanges of water and energy (Sellers et al, 1986; Dickinson et al, 1986). These models in some sense couple the land-biosphere and atmosphere, but they do so specifically to quantify the short-term exchange of water and energy from the atmosphere's perspective. Models which reflect a hydrologie or ecosystem perspective, and which are capable of extrapolation over large spatial scales, integrate some key state variables over much longer time scales, monthly to annually (Parton et al., 1987; Vôrôsmarty et al., 1989; Running, 1992; Raich et al., 1991). Regional models have explicitly linked models of C0 2 exchange and transpiration by modelling fluxes across the canopy boundary, but have not explicitly nested the calculations within an interactive atmospheric model. At the scale of a small watershed, models of water and Linked atmosphere-hydrology models at the macroscale 11 energy exchange between the atmosphere and land surface often operate at daily or subdaily time steps (Waring et al., 1981; Larson, 1982; Running & Coughlan, 1988). To understand the influence of climate change on hydrologie systems, we will need to simultaneously address the issues of natural climatic variability as well as greenhouse forcing, both of which span time scales extending to years and decades. The natural climatic variations of interest to hydrologists include events such as extended drought, low river flow and floods. Such events have been linked to remote forcings associated with sea surface temperature anomalies (e.g. Nicholls, 1989; Richey et al., 1989; Trenberth et al., 1988) and it is important to understand how both the water cycle and biosphere function are affected. This constitutes a distinctive signal-noise problem wherein separating one influence (e.g. increased greenhouse gases) from others requires an understanding of physical feedbacks across a broad range of scales. Analyzing these events will require detailed time series of climatic forcing fields and runoff derived principally from observational data and will probably restrict analysis to data-rich areas of the globe. The difficulty, of course, is that those areas with rich data resources are precisely those in which water resources are intensively manipulated, hence obscuring the natural water cycle. MODELLING STRATEGY FOR REGIONAL ASSESSMENTS The foregoing discussion suggests that (a) the region is the important scale to focus upon when considering the issue of hydrologie change, and (b) our understanding of atmosphere-hydrology interactions over regions will depend on our ability to combine both fine-scale and synoptic dynamics. Although an enormous body of hydrologie literature reflects work on physically-based models at more local scales, it would be both impractical and of questionable scientific value to apply intact versions of such models substantially beyond their intended time and space domains. Obviously, major problems would arise with data collection and storage, parameterization, validation and required computing power. The more tractable structure and relatively modest data requirements of a synoptic model therefore have appeal. However, if we wish to understand which elements of the water cycle are influenced by anthropogenic change, highly empirical, coarsely resolved models may not provide adequate insight. Macroscale Hydrologie Models may therefore be viewed more appropriately as hierarchical structures with finer scale, site-specific submodules interacting with simulations over broader domains (Table 1, Fig. 4). Such a configuration purposefully seeks to distill the impacts of global climate change down to regional and local levels, while simultaneously propagating physically valid dynamics up to regional and synoptic scales — a coupled 'top-down' and 'bottom-up' approach. The coarse-scale level serves as an integrator of the numerous subregional effects. It maintains models plus various climatic and biophysical data sets whose domains encompass regions or whole continents. The coarse-scale not only provides a mechanism by which to obtain a picture of regional hydrology, but it serves as a useful check of the overall modelling approach when compared to regional well log and river discharge data. The models are nested within a still coarser-scale GCM. The spatial and temporal scale for the atmospheric calculations are typical of current GCMs (a 100 km length scale and subhourly time steps) while the MHM at this resolution will 12 C. J. Vôrôsmarty et al. Table 1 Key characteristics of the multiple scale analysis. SCALE MODEL BOUNDARY RESOLUTION ATMOSPHERE HYDROLOGY Continental Linked GCM - MHM Interactive in GCM > 100 km (sub hourly) 10-50 km (weekly to monthly) Meso (Regional) Linked meso - MHM Prescribed or interactive in nested GCM 10-50 km (minutes) 1 - 10 km (daily) Local Hillslope small catchment Topographically determined Prescribed point forcings (subdaily) <km (subdaily) operate at length scales from 10 to 50 km with weekly or monthly time steps. Lateral boundaries are interactive within a GCM. At the other end of the spectrum, the local scale maintains finely-resolved data sets for detailed temporal and spatial land-surface exchanges and runoff generation. Hillslope-catchment models will operate on an initial set of representative basins defined by topographic divides from digital elevation models (DEMs) using grid MODELING SYSTEM FOR MULTI - LEVEL HYDROLOGICAL ANALYSIS Linkage with GCM's Broad Climatic Context Regional Linkage to Atmospheric Models Combines Continental Integration with Extrapolated Local Dynamics 9 0 Hillslope Processes Physically - Based Fig. 4 Variable resolution strategy to develop linked atmosphere - macrohydrology models for regional scale studies. Linked atmosphere-hydrology models at the macroscale 13 spacings well below 1 km. Due to the small spatial domain, météorologie forcings are prescribed from subdaily station data. Although there are inherent advantages to both coarse and fine-scale analysis, neither alone will provide a process-based view of hydrology at the regional scale. For this we will require an intermediate scale of resolution. A mesoscale atmosphere-hydrology model will simulate a series of representative subregional basins characteristic of the overall regional drainage system. The atmospheric component will operate with time steps of the order of minutes and a 10-50 km spatial resolution. The hydrologie module will operate at daily time steps and a resolution of from 1 to 10 km. Boundaries will be prescribed from either observation or GCMs. This model arises from the application of reciprocal scaling functions between the local and continental-scale endpoints. These take the form of spatially-correlated statistical distributions that can be based on observed local data (Eagleson, 1978; Zimmerman & Wilson, 1990). Once a classification of all subregional basins is applied to the entire domain at the GCM scale, a characteristic set of scaling functions is used to interpret GCM-derived climatic forcings for each subregional mesoscale drainage basin. Simultaneously, dynamics developed at local hillslope or small catchment resolutions (i.e. for energy, momentum, latent heat, runoff) are scaled back to the meso scale, integrated, and later passed to the synoptic scale for incorporation into the basin-wide GCM-hydrology model. In this way, an area-integrated flux of latent heat can be computed as a lower boundary condition within a GCM grid and local distributed runoff can be aggregated to produce large river discharges. Such a multi-tiered approach requires an integrated nesting of models and associated data sets within an interactive geographic information system as described later. Component Processes A coupled A-MHM is essentially a suite of interacting process models. Below we list a set of specific candidates that we believe can be applied to the multi-tiered strategy outlined above. The listing is in no way intended to be exclusive nor exhaustive. It simply reflects the authors' collective expertise in applying each particular model and our initial judgment as to a sensible choice within each category. GCM-scale Atmospheric Model Component A strong candidate for global climate simulation in this context is the NCAR Community Climate Model, Version 1 (CCM1) (Williamson et al., 1987). Although this model is typically run on a supercomputer, versions capable of running on workstations exist and have been used successfully for short-term climate simulation. A workstation version at Iowa State University can run at spectral resolutions up to 9 layers at triangular 42 horizontal truncation (roughly 2.8 x 2.8 degree spatial resolution). Even under the computational limitations of a workstation, numerous successful studies have been conducted with this version of the CCM1. Chen et al. (1992) have studied the response of North American circulation patterns to changes in Pacific Ocean surface temperatures, illustrating aspects of teleconnection patterns that may be relevant for studying changes in land-atmosphere interaction in the central United States. Further study by Chen (1992) has shown the relationship of remote, ocean temperature changes to the Indian Monsoon. 14 C. /. Vôrôsmarty et al. Typically, this has been done as a one-way coupling: the global model provides boundary conditions for a mesoscale model, but the physics of the mesoscale simulation do not feed back to the global model. While one-way coupling represents a reasonable first step in linking models of different scale, the approach eliminates local influences on large-scale circulation that may be resolved in the mesoscale model but not the global model. For example, more highly resolved topography in a mesoscale model can produce markedly different precipitation patterns compared to those produced by the global driving model (e.g. Anthes, 1990). The differing patterns of atmospheric heating could potentially yield differences in regional if not global circulation. Mesoscale Atmospheric Model Component An example of the mesoscale component is the Pennsylvania State/NCAR MM4 simulation (Anthes et al., 1987; Anthes, 1990) which has been used successfully in conjunction with CCM1 simulations as a link between global and regional simulation (Giorgi & Mearns, 1991). The MM4 was developed as a regional model with nominal resolution of 0.5° latitude x longitude, which should prove advantageous for coupling with the coarse- scale hydrologie model. A recent upgrade of this model, the MM5, includes multiple layers of grid nesting and a non-hydrostatic formulation that permits simulation down to scales of a few kilometres. For developing and testing atmosphere-surface coupling schemes, less computationally demanding models may be desirable. We are currently using one such model, the AER Local Forecast Adaptation (ALFA) model. The ALFA model solves essentially the same equations as a climate model, but in a single column only, with horizontal boundary conditions specified by large-scale, three-dimensional analyses. Being unidimensional, the ALFA model can include many vertical levels, a sophisticated description of atmospheric physics, and a detailed computation of the interactions between the atmosphere and the land surface, while remaining computationally inexpensive. Wetzel & Chang (1988) and Koster & Eagleson (1990) have also used a column model to emulate a GCM grid box. By simulating a single box, we remove feedbacks between the large-scale environment and local dynamics. However, the column model is more efficient than a three-dimensional model, and it allows us to control more easily in our experiments the large-scale atmospheric conditions. The ALFA model, in particular, is being used by its developer, Dr. J.-F. Louis, to assimilate observations produced by the US Department of Energy's (DOE) Atmospheric Radiation Measurement (ARM) program. A portion of the ARM effort aims at observing scale interactions in climatological land-atmosphere exchange, so that use of the ALFA model as a development tool should ease access to and enhance the utility of an important data base being developed by DOE. The Soil-Vegetation-Atmosphere Transfer (SVAT) Component Atmospheric models use a host of techniques to simulate the atmosphere's link with the surface. On the broadest scale, GCMs typically use relatively simple approaches, representing surface exchange fluxes with bulk formulas that simulate interaction between the land surface and the lowest layer of the atmospheric model. At higher resolutions, the techniques are generally much more complex. Models will diagnose the depth of the atmospheric boundary layer and flux processes operating within it. Vertical fluxes are typically modeled using diffusion coefficients that depend on the structure of the Linked atmosphere-hydrology models at the macroscale 15 larger-scale (resolved) environment. Depending on the degree of vertical resolution, up to several atmospheric layers may be involved in simulating boundary-layer dynamics and their interaction with the surface. At the land surface itself, a SVAT scheme must simulate a variety of processes of hydrological importance. Additionally, we are searching for a form that can eventually be embedded within a terrestrial ecosystem model for carbon, nutrients, and the water cycle. Transpiration will be simulated using a physically-based Penman-Monteith approach. Transpiration will thus be a function of radiation, wind speed, humidity, leaf area index, aerodynamic resistance and stomatal (canopy) resistance. The Penman-Monteith equation can also be used for interception, but with reduced canopy resistance. Throughfall and stemflow can be calculated by difference between precipitation and interception. Direct soil losses, although less important in forests, need to be taken into account in short vegetation, both natural and agricultural. A simple model would assume that this loss term is directly related to radiation and saturation vapour pressure and inversely to leaf area index. The division of remaining effective precipitation to storm flow (direct runoff) and eventual baseflow may follow either a simple empirical approach for the coarse scale simulations (e.g. use of the Holton, Philip, Kostiakov equations) or the solution of a two-dimensional Richards equation at the hillslope/catchment scale (Skaggs, 1982). The excess water so partitioned will then enter the groundwater/surface water model for subsequent horizontal routing. The Groundwater/Surface Water Model Component Ground and surface water processes will be represented at each scale of resolution, and in all cases, excess water from the SVAT model will be partitioned between surface and subsurface flow paths. At the hillslope-catchment scale, lateral groundwater flow rates will be determined using: (1) « * = ' K ^ ' qy= ~KTy where qx, qy are the specific discharges (or Darcy velocities) in thex - y plane, h is the head, and Kis the hydraulic conductivity. This constitutive law is incorporated into the continuity equation to yield a transmissivity-based equation that can be easily solved using finite difference methods (Wang & Anderson, 1982). This distributed parameter model is capable of computing changes in groundwater storage in a heterogeneous porous medium: Tdh dx dx + Tdh dy "dy~ Sy^r{x,y,t) (2) where t is time, Sy is the specific yield (volume released per unit area of aquifer per unit loss of head), r{x,y,t) is a time and space dependent recharge into the aquifer from either excess rainfall or snowmelt from the SVAT model, and T is the transmissivity (= Kb; where b is the saturated thickness). Source/sink terms can also be added to simulate the impact of water usage by humans. This construct borrows from other topographically-based models, like those of Beven & Wood (1983) (TOPMOD), O'Loughlin (1990) (TOPOG) and Gupta & Solomon (1977). Note that under this configuration, the dynamics of groundwater and C. J. Vôrôsmarty et al. 16 stream flow are linked at topographic low points. Digital elevation models will be used to formulate small drainage networks and help to determine the direction of subsurface flow by establishing spatial variations in the aquifer basement and precipitation/recharge (i.e. orographic effects). Maps of land cover, soils and aquifer characteristics will also be required. In conjunction with specified météorologie data, this model will serve as a runoff "generator" at the mesoscale based on results obtained in small representative catchments. The groundwater/surface water (GW/SW) flow models will be simplified for use at the meso- and GCM-scales and will incorporate a river discharge algorithm. Recent experiments (Espirito Santo et al., in preparation) comparing distributed and lumped parameter surface/subsurface hydrology models indicate that in typical field settings there is a characteristic length scale adjacent to open water channels in which the channel and ground water interact significantly. Beyond this length (i.e. upslope) there is a net uni-directional discharge downward. Coarse-scale GW/SW coupling will therefore be determined by combining a unidirectional, linear decay model with a lumped parameter groundwater-surface water model. Based on estimates from either ground-based or remote sensing surveys, each grid at the meso- or GCM-scale will be assigned an estimate of the proportion of landscape within and outside of the interactive zone. Within a discretized spatial element the upslope model will calculate, for any time step, coarse-scale groundwater discharge and head changes using the following: Qu = -CuhuAxuAyu àhu (QJAx.AyJ+R» & Syu (3a) (3b) where Qu is the total groundwater flow from upslope into the interactive zone per unit time, Cu is a linear coefficient, hu is the mean head above the base of the upslope aquifer, Ru is the total groundwater recharge from the SVAT module for the upslope area, Axu Ayu is the elemental grid area, and Syu is the upslope mean specific yield. In the interactive zone, water balance expressions for groundwater and surface water reservoirs are formally coupled using the following: Qt = -Ci{hrhr)AxiAyi Mh _ {[QrQuVàxAy^Rj ât " Sy{ (4a) (4b) where Q{ is the total net discharge from the interactive groundwater pool to the stream, C, is a linear coefficient, ht is the average aquifer head (relative to a datum, i.e.the aquifer base) in the interactive zone of the cell, hr is the areal average stream or river head (relative to datum) in the cell, Axt Ayt is the elemental grid area, Rt is the total groundwater recharge from the SVAT module into the interactive zone, and Syt is the interactive zone's mean specific yield. In this approach stream and aquifer water level gradients within a cell are replaced by spatially averaged quantities. The aquifer/stream exchange is then determined by considering the magnitude of the difference in average pool sizes (water levels) between the stream and aquifer reservoirs (Fig. 5). While the approach is simpler than traditional "distributed parameter" groundwater flow models Linked atmosphere-hydrology models at the macroscale Hillslope to Catchment Scale 17 Meso to GCM Scale High Relief v r NONINTERACTIVE INTERACTIVE Low Relief y .-owOC*-*- -*— -*» NONINTERACTIVE INTERACTIVE INTERACTIVE NONINTERACTIVE Fig. 5 The linkage between groundwater and streams at the hillslope-catchment and coarse scales. The interactive zone refers to that domain over which adjacent streams convey significant influence on groundwater dynamics and vice versa. Storm flow or direct runoff (not shown) is implicitly calculated by the SVAT module and delivered to the stream. (Loague & Freeze, 1985), it has produced results comparable to more sophisticated distributed parameter models. McLin (1981), for example, used a lumped parameter model to reproduce changes in stream flow, groundwater levels, and salinity along an 18 km reach of the Arkansas River in southeast Colorado (USA). His model compared favourably to average hydrologie results obtained from distributed models applied to the same system (Konikow & Bredehoeft, 1974). Surface flow will be simulated by a set of differential equations representing each discretized river reach. Within each river element: àSç àt = D -D. + Q+F (5) in which Sc is the storage of water in the channel reach, Du represents the sum of all upstream discharge inputs, Dd is downriver discharge from the reach, and Fs is storm flow quantified and implicitly routed off the land surface and through small streams by the SVAT. Although Dd defines the downstream flux of water from a particular cell, it also serves as an upstream input in the continuity equation for its adjacent downstream reach. A flow formula, such as Manning's, is employed to simulate flow within rivers: DA m n ARV*S'A (6) where Cm is a constant, n is Manning's roughness coefficient, A is the cross-sectional 18 C. J. Vôrôsmarty et al. area, R is the hydraulic radius, and S is the slope of the energy grade line (Streeter & Wylie, 1979). River geometry can be either specified from field survey or inferred from géomorphologie principles in different climatic and topographic zones. Equations (5) and (6) can be modified to account for flood-plain inundation and engineering works such as impoundments and stream channelization. Model Coupling Before the component models can be linked as an integrated A-MHM, suitable study areas must be defined for each spatial and temporal scale. In those areas, each of the modules must be tested independently using available land surface biophysical data, observed forcing functions, calibration and validation data sets. Spatial and temporal scaling functions linking each level of resolution must be formulated. The final linkage must progress in stepwise fashion, linking, for instance, the atmospheric and SVAT models, the SVAT simulation to the GW/SW model and the GW/SW model to the river routing algorithm. Each step will require a careful consideration of numerical stability and computational demands, and a mechanism to recognize emerging feedbacks. If, for example, a study of the Mississippi River drainage basin is sought (3.3 x 106 km2), the GCM-scale could be defined as a series of nested subregional drainage basins (i.e. mesoscale, 104 to 105 km2) residing within the larger domain. A series of representative subregional basins would be identified, and mesoscale models would be developed for each particular class of subregional drainage basin. Much of the Kansas or Platte River systems could, for instance, be used to define the dynamics of grassland basins in rolling topography. Parameters derived from the representative basin models would then be assigned to hydrologically similar mesoscale basins using a pre-determined classification system residing within a geographic information system (GIS). Although representative parameters (e.g. the fraction of runoff in storm or baseflow) will be assigned to each class of mesoscale basin, site specificity will be afforded through geographically-distributed differences in topography, soils, vegetation and time-varying climatic variables. Numerous representative hillslope-catchment sites would in turn be nested within each mesoscale basin and representative parameterizations of hillslope-catchment models would be geographically distributed using a procedure similar to that used at the mesoscale. With the landscape so defined, each of the component models would need to be both validated and checked for numerical stability. At this point, the atmospheric and hydrological models also could be coupled at the meso- and GCM-scales and be used to identify feedbacks. A simple first step would be to compare estimates generated by the linked atmospheric-MHM model to those made by the GW/SW model driven by prescribed climatic data. Differences in the two sets of estimates would provide an initial clue to understanding how inputs from the atmosphere alter the hydrologie response and how stable the computational linkage will be. The coupling would also require formulating the scaling functions (as discussed earlier) to carry dynamics from one level of resolution to the next. Account must also be made of the impact of engineering works including irrigation, evaporative losses from reservoirs, interbasin transfers and other such manipulations of the region's hydrology. In highly managed basins like the Mississippi, this will constitute a major undertaking. Linked atmosphere-hydrology models at the macroscale 19 Work with the coupled models could be inhibited by their potentially heavy demand on computing resources. Emerging high performance computing technologies, especially in the area of parallel processing, could circumvent this problem if the technology can be properly exploited. The coupled models should be amenable to parallel/vector processing on two levels: (1) parallelism among the spatially distributed surface elements and (2) parallelism in solving coupled sets of differential equations within each submodel. How we exploit this potential will, of course, depend on the processor architecture adopted for the global climate model as a whole. For example, a cluster of grid boxes could be assigned to a cluster of processors, each grid box having access to the entire cluster of processors in turn. Many efforts are underway to port models to parallel machines. For example, scientists at Iowa State University and the Department of Energy's Ames Lab (on the ISU campus) are adapting the MM4 to massively parallel computing environments. They are also exploring parallel computing in an environment that links multiple workstations. Thus, a wide variety of computing approaches can be developed but careful study is needed to determine optimal configurations for the task at hand. Finally, although the most immediate application of such advanced computing is with the coupled models themselves, managing and visualizing the simulated and observed data needed for model development and validation can also extract heavy computing demands. Resources must be developed to exploit advances in computing technology that will improve our ability to manipulate and comprehend the enormous data bases being generated by the research community. Data Support The Global Hydrologie Archive and Analysis System (GHAAS) is a data and model management tool which will integrate the analysis into a coherent framework (Fig. 6 a,b). It is similar in concept to other hydrologically-oriented analysis systems such as RESSys (Running, 1992) and GEnESIS (Solomon et al, 1990). The system is global in coverage, but can be used to analyze regional and local issues. The GHAAS is geographically-referenced and can handle both vector and raster-based data sets. It links together various input data sets with models and their output, in combinations defined by the user. It links to third-party software products (e.g. ARC/INFO, GRASS, PV WAVE) for visualization, spatial analysis and image processing. The system resides on a UNIX-based workstation. The package is used in assembling data sets from disparate sources, data checking and archiving in a geo-referenced format. The GHAAS facilitates calculations on both distributed water balance components and discharge dynamics in stream and river systems. For a particular analysis, the user identifies a target location and scale of analysis. The user can then browse available holdings and select from numerous data sets on soils, vegetation, remote sensing imagery, digital elevation, drainage basin topology and drainage basin structure. The data can be manipulated using functions that permit, for example, the spatial interpolation of point data. It also houses a growing inventory of potential anthropogenic changes associated with large water resource projects, land use change, and results from climate change scenarios. Links are available to tabular information that can be used for model calibration and testing. The user then selects from a suite of models and in the current context will be able to specify the way in which the component models are coupled. The user can customize output in the form of tables, C. J. Vôrôsmarty et al. 20 GLOBAL HYDROLOGIC ARCHIVE AND ANALYSIS SYSTEM USER FUNCTIONS ARCHIVE HOLDINGS AND TOOLS Browse j LOCAL META DATA NETWORK PREVISUALIZATIOh I REGION Data CLIMATE UNO SURFACE • •L3 • •II n^rnrz an 1 Scale FINE • MESO Models Output Variables NETWORK TOPOLOG Y HILLSLOPE •1111 i:ÊË!;:i!i SOIL WATER LATENT HEAT • 1 COARSE INTERPOLATED SVAT GROUND / SURFACE WATEF WIND FIELDS DISCHARGE POINT DATA GCM RUNOFF Visualization till!!! MAPS TABLES STATISTICS ANIMATIO N Fig. 6 (a) Elements of the Global Hydrologie Archive and Analysis System, showing data, models, and output management tools. The system is set in a geographically-referenced context and provides a framework for hydrologie study over multiple scales. The shaded boxes represent one possible combination of menu selections made by a user. statistics, maps and animated colour graphics. The GHAAS relies on NFS and TELNET links to client-server systems and can connect to various on-line distributed database systems (e.g. NASA Master Directory, Earth Data System [Ellery Systems, Inc.]) using TCP/IP protocols. Remotely sensed data can significantly strengthen the type of macroscale study we have outlined. The frequent temporal and large geographical coverage of select sensor data onboard satellite platforms such as the National Oceanic and Atmospheric Administration's (NOAA) Polar Orbiters may be particularly valuable to hydrologie investigations (Dozier, 1992). Table 2 provides examples of sensor data currently available and geophysical products that can be developed from processing these Linked atmosphere-hydrology models at the macroscale 21 KANSAS RIVER STUDY 1986 GAGING STATION 1987 1988 1989 1990 SITE: Fort Riley, Kansas Fig. 6 (b) An example using the GHAAS to preview and understand drainage system characteristics prior to simulation. The study area is in the State of Kansas (USA). The particular interface shown was generated using the ARC/INFO commercial software package. radiance data. These data require substantial processing to ensure that inherent problems such as cloud and atmospheric attenuation of the recorded radiances are removed (Hall et al., 1988). The additional step of interpretation is required to yield biophysical information of value in hydrological studies. The AVHRR, for example, provides a vegetation index that has been used in numerous land cover studies around the world (e.g. Justice et al., 1985; Heilkema et al., 1987; Goward et al., 1991). The great strength of these data sets is in the classification of vegetation into functional units that are of meaning in the hydrological sense (e.g. tree vs. crop vs. urban). In areas of shifting land use, such a capability will be of enormous importance. These data sets also can depict phenology across the terrestrial biosphere. In addition, Nemani & Running (1989) inferred canopy resistance from the thermal signal of AVHRR. Active microwave sensors and laser altimetry can give information on land surface structure (e.g. roughness [NASA, 1987; Harding, 1993]). At particular wavelengths, passive and active microwaves can delineate the behaviour of rivers and associated floodplains (Giddings & Choudhury, 1989; Imhoff et al., 1987) and moisture content in the soil surface (Schmugge, 1985). The advantage of long-term monitoring through remote sensing may soon be realized through ongoing efforts such as the joint NASA/NOAA Pathfinder Project to reprocess archival data from select NO A A satellites. Two efforts have relevance to hydrological modelling - the Advanced Very High Resolution (AVHRR) and TIROS Operational Vertical Sounder (TOVS) Pathfinder data sets. For the AVHRR, 12 years (1981 to 1993) of Level lb Global Area Coverage (GAC) data at 4 x 4 km spatial 22 C. /. Vôrôsmarty et al. Table 2 Examples of satellite sensor data currently available and associated geophyscial products that may be useful in hydrological modelling efforts. Satellite and Daily Sampling Frequency (for given area) Instrument(s) Onboard and Spectral and Spatial Res. Geophysical Products Retrieved Polar Orbiter Environmental Satellite (POES) - NOAA 4 times daily with 2 satellites in orbit 1. Advanced Very High Resolution Radiometer (AVHRR); 4/5 channels (2 reflective chls. 2/3 Thermal IR chls.); 1km and 4km data avail. Vegetation Indices (Vis). Land Surface Temperature; Broad-scale Classes of Surface Cover; Inferred Estimate of Net Primary Production. 2. TIROS Operational Vertical Sounder (TOVS) - 2 instruments a. NIRS - 1 reflective and 19 Thermal IR chls; 17.5km at nadir b. MSU 4 chl passive mircowave radio-meter; 125 km at nadir Land Surface Temperature; Atmospheric Temperature at various levels; Cloud Cover; Total precipitable Water; and OutGoing Longwave Flux 1. Optical Line Scanner (OLS)2chls. 1 reflect., 1 Thermal IR; 1.5km at nadir Snow/Ice Extent; Land/Sea Boundaries 2. Special Sensor Mircowave/Imager 7 chls. passive microwave 5 at 15km and 2 at approximately 28km at nadir Snow/Ice Extent; Soil Moisture at Surface; Total Precipitable Water; Cloud Cover; Sea Extent 1. Multispectral Scanner (MSS) 4 reflective channels in vis/NIR; 57 meters at nadir Vis; Species Classification (deciduous vs. conifers) Surface Cover Mapping; Monitoring Flood Stages; Anthropogenic Changes of Surface Cover (e.g., forest clearing) 2. Thematic Mapper (TM) 6 reflective chls. and 1 Thermal IR chl; 30m and 120m resolution respectively Same as for MSS Defense Meteorological Satellite Platform (DMSP-F8) U.S. Dept. of Defense 4-6 daily with 2 satellites in orbit NOAA's Landsat 4 and 5 Satellites 16 day repeat cycle Linked atmosphere-hydrology models at the macroscale 23 resolution are being obtained and processed. The processed data will yield various geophysical parameters such as the Normalized Difference Vegetation Index (NDVI). A comprehensive set of processing steps (for cloud contamination, intercalibration of sensors from different satellites, etc.) is being carried out using the best reference data and techniques currently available. The current plan is to supply products on an 8.5 km grid and to include the original processed channel data for researchers who wish to develop additional products (James, 1992). The TOVS Pathfinder Project is also developing a time series of thermal IR and passive microwave products from 1978 to the end of 1994. The products to be released will have a horizontal resolution of 25 km and be available for different levels within the atmosphere. Geophysical products to be generated include: atmospheric temperature, total precipitable water, cloud top heights and temperatures and specific humidity. Other products under consideration are surface temperature, outgoing long-wave radiation and precipitation estimates (Susskind, 1992). A twelve-year record of 37 GHz frequency passive microwave data sets collected on board the Nimbus-7 (SMMR) and DMSP-F8 (SMM/I) satellites, also provides an important long-term data set (Choudhury et al., 1992) for exploring the dynamics of drainage basins (Vôrôsmarty et al,. 1991 #b). Experience gained through intensive field campaigns such as FIFE, BOREAS, HAPEX, and the Boardman Regional Flux Experiment should help to provide land surface boundary data and interpretation of remotely sensed imagery. Recent releases of the associated data sets on optical media (e.g. for FIFE, Strebel et al., 1991) will help greatly in the calibration and validation of local and, potentially, regional S VAT models that are integral to the macroscale analysis. A growing interest in sites with disturbed landscapes (Shuttleworth, 1991) will make important contributions toward understanding the impact of widespread transformation of the Earth's surface by humans. CONCLUSIONS In this paper we have discussed numerous aspects of macroscale hydrologie models, with an emphasis on linkage to atmospheric simulations. We summarized key scaling issues for both space and time and concluded that the regional scale is the most critical to consider for global change studies. The region is also the level at which policy will be formulated in the wake of such change. Macroscale hydrologie models might reasonably be considered as a hierarchical, nested system of models, emphasizing the mesoscale, but with a "bottom up" linkage to relatively finer scales and a "top down" linkage to coarser scales. The nested approach requires the operation of a GCM, a mesoscale atmospheric model, a soil-vegetation-atmosphere transfer scheme and surface/subsurface hydrology models, the latter down to hillslope scale in small representative catchments. The multiple models and associated data sets need to be linked in a common analysis system which can manage sophisticated input/output functions, scientific visualization, model preparation and execution. The entire system of data, models and computing tools is being developed to distill the facets of global change to the regional level and to explore feedbacks from the land surface back to the atmosphere. Tests have recently begun for the Kansas and Platte River systems on a linked A-MHM that can be embedded within a synoptic-scale model of the larger Mississippi 24 C. J. Vôrôsmarty et al. basin. Results from the initial experiments will provide important guidance in studies of other drainage systems of the world, many in data-poor regions. Although our emphasis in this paper has been on water per se, we have implicitly set the stage for linking the water cycle to the biogeochemistry of carbon, nitrogen, phosphorus and other constituents. The coupling would occur in both the soil-vegetation-atmosphere system using a terrestrial ecosystem model, and in simulated river systems that would receive, process and transport waterborne constituents. Linked A-MHMs are therefore an important first step in the development of truly interactive Earth Systems Models of the terrestrial biosphere. Acknowledgements The authors wish to recognize support from the following US funding agencies: Department of Energy (Grant # DE-FG02-92ER61473), Environmental Protection Agency (Cooperative Agreement #816278), National Aeronautics and Space Administration (Grants NAGW 2669 & NAGW 1888). We also wish to acknowledge the assistance of S. Mazurkiewicz, D. Wright, S. Palmer, A. Schloss and R. White in preparing the final text and graphics. REFERENCES Alley, W. M. (1984) On the treatment of évapotranspiration, soil moisture accounting, and aquifer recharge in monthly water balance models. Wat. Resour. Res. 20(8), 1137-1149. Anthes, R. A., Hsie, E.-Y., & Kuo, Y.-H. (1987) Description of the Venn Stale/NCAR Mesoscale Model Version 4 (MM4). NCAR Technical Note [NCAR/TN-282+STR]. Anthes, R. A. (1990) Recent applications of the Penn State/NCAR mesoscale model to synoptic, mesoscale and climate studies. Bull. Am. Met. Soc. 71, 1610-1629 Avissar, R. & Pielke, R. A. (1989) A parameterization of heterogeneous land surfaces for atmospheric numerical models and its impact on regional meteorology. Mon. Weath. Rev. 117, 2113-2136. Bartell, S. M. & Brenkert, A. L. (1992) A spatial-temporal model of nitrogen dynamics in a deciduous forest watershed. Quantitative Methods in Landscape Ecology, (ed. by Turner & Gardner). Springer-Verlag. Bathurst,J. C. (1986) Physically-based distributed modelling of an upland catchment using the Système Hydrologique European. J. Hydrol. 87, 79-102. Becker, A. & Nemec, J. (1987) Macroscale hydrological models in support to climate research. The Influence of Climate Change and Climate Variability on the Hydrologie Regime and Water Resources. IAHS Publication #168. Vancouver, 431-445. Beran, M. A., Brilly, M., Becker, A. &Bonacci, O. (1990) Regionalization in Hydrology. IAHS Publication tf 191. IAHS, Wallingford, UK. Beven, K. & Wood, E. F. (1983) Catchment geomorphology and the dynamics of runoff contributing areas. J. Hydrol. 65, 139-158. Branscome,L. E., &Gutowski, W. J. (1992) The impact of doubled C0 2 on the energetics and hydrologie processes of mid-latitude transient eddies. Climate Dynamics 8, 29-37. Chahine, M. T. (1992) The hydrological cycle and its influence on climate. Nature 359, 373-380. Chen, T.-C. (1992) The interannual variation of Indian Monsoon simulated by the NCAR Community Climate Model: The effect of the western tropical Pacific SST. Submitted to J. Climate. Chen, T.-C. & Kpaeyeh, J. A. (1992) The synoptic-scale environment associated with the low-level jet of the Great Plains. Mon. Weath. Rev. 121, 416-420. Chen, T.-C, van Loon, M. H., Wu, K.-D. & Yen, M.-C. (1992) Changes in the atmospheric circulation over the Pacific-North American area since 1950. / . Meteor. 70, 1137-1146. Choudhury, B. J., Major, E. R., Smith, E. A. & Becker F. (1992) Atmospheric effects on SMMR and SMMI/I 37 GHz polarization difference over Sahel. Int. J. Remote Sens. 13, 3443-3463. Committee on Global Change (1990) Research Strategies for the US Global Change Research Program. National Academy Press, Washington, DC. Correll, D., ed. (1986) Watershed Research Perspectives. Smithsonian Institution Press, Washington, DC. Linked atmosphere-hydrology models at the macroscale 25 Dickinson, R. E., Henderson-Sellers, A., Kennedy, P. J. & Wilson, M. F. (1986) Biosphere-Atmosphere Transfer Scheme (BATS) for the NCAR Community Climate Model. NCAR Technical Note, NCAR/TN-275+STR, National Centre for Atmospheric Research, Boulder, Colorado. Dickinson, R. E. & Henderson-Sellers, A. (1988) Modelling tropical deforestation: a study of GCM land-surface parameterizations. Quart. J. Roy. Met. Soc. 114, 439-462. Dozier, J. (1992) opportunities to improve hydrologie data. Rev. Geophysics 30, 315-331. Eagleson, P. R. (1978) Climate, soil, and vegetation; parts 1-7. Wat. Resour. Res. 14, 705-790. Eagleson, P. S. (1986) The emergence of global-scale hydrology. Wat. Resour. Res. 22, 6S-14S. Elder, J. F. (1985) Nitrogen and phosphorus speciation and flux in a large Florida river wetland system. Wat. Resour. Res. 21, 724-732. Engstrom, D. R. & Nelson. S. R. (1991) Paleosalinity from trace metals in fossil ostracods compared with observational records at Devils Lake, North Dakota, USA. Paleogeography, Paleoclimatology, Paleoecology 83, 295-312. Entekhabi, D. & Eagleson, P. S. (1989) Land surface hydrology parameterization for atmospheric general circulation models including subgrid scale spatial variability. J. Climate 2, 816-831. FAO/UNESCO: 1977, Soil Map of the World. 1:5 000 000. Vol. I (Legend), Vol. VI (Africa). UNESCO, Paris. Fédérer, C. A. (1992) Documentation for BROOK 90: A Simulation Model for Evapotranspiration, Soil Water and Slreamflow. USDA Forest Service, Northeast Forest Experiment Station, Durham, New Hampshire. Freeze, R. A. (1980) A stochastic-conceptual analysis of rainfall-runoff processes on a hillslope. Wat. Resour. Res. 16(2), 391-408. Fritz, S. C , Juggins, S., Battarbee, R. W., & Engstrom, D. R. (1991) Reconstruction of past changes in salinity and climate using a diatom-based transfer function. Nature 352. Giddings, L. & Choudhury, B. J. (1989) Observation of hydrological features with Nimbus-7 37 Ghz data applied to South America. Int. J. Remote Sens. 10, 1673-1686. Giorgi, F., & Mearns, L. O. (1991) Approaches to the simulation of regional climate change: a review. Rev. Geophys. 29, 191-216. Goward, S. N., Markham, B., Dye, D. G., Delaney, W. & Yang, J. (1991) Normalized difference vegetation index measurements from the AVHRR. Remote Sens. Environ. 35, 257-277'. Gupta, S. K. & Solomon, S. I. (1977) distributed numerical model for estimating runoff and sediment discharge of ungauged rivers, 1: The information systems. Wat. Resour. Res. 13, 613-618. Gutowski, W. J., McMahon, G. F., Kirshen, P. &Schluchter, S. (1993) Projecting climate change to small basins: potential impact of global warming on flooding in South Florida. Submitted to /. Wat. Res. Plan. Man. Hall, F., Strebel, D. E. & Sellers, P. J. (1988) Linking knowledge among spatial and temporal scales: vegetation, atmosphere, climate and remote sensing. Landscape Ecology 2, 3-22. Harding, D. E. (1993) Vegetation analysis with airborne laser altimeter data. Goddard Space Flight Centre Seminar, 23 February. Greenbelt, MD. Henderson-Sellers, A. (1992) PILPS: Project for Intercomparison of Land-Surface Parameterization Schemes. Working Paper to form basis of PILPS Planning Meeting. 1992. Manuscript No. 300392. Macquarie University, North Ryde, Australia. Hielkema, J. V., Prince, S. D. & Astle, W. L. (1987) Monitoring of global vegetation dynamics for assessment of primary production using NOAA AVHRR. Adv. Space Research 7, 81-88. Horel, J. D., & Wallace, J. M. (1981) Planetary scale atmospheric phenomena associated with the Southern Oscillation. Mon. Weath. Rev. 109, 813-829. Hupp, C. (1982) Extension of flood record for Passage Creek, Virginia, utilizing tree-ring analysis. Bull, of Ecological Soc. of Am. 63, 78. Imhoff, M. L., Vermillion, C , Story, M. H., Choudhury, A. M., Gafoor, A. & Polcyn, F. (1987) Monsoon flood boundary delineation and damage assessment using space-borne imaging radar and Landsat data. Photogramm. Engng 53, 405-413. James, M. (1992) AVHRR Pathfinder Progress. Report at GSFC DAAC Version O Meeting. November, 1992. Justice, C. O., Townshend, J. R. G., Holben, B. N. & Tucker, C. J. (1985) Analysis of the phenology of global vegetation using meteorological satellite data, Int. J. Remote Sens. 6, 1271-1318. Keen, K. L., & Shane, L. C. K. (1990) A continuous record of holocene eolian activity and vegetation change at Lake Ann, east-central Minnesota. Geol. Soc. Am. Bull. 102, 1646-1657. Kochel, R. C , & Baker, V. R. (1982) Paleoflood hydrology. Science. 215, 353-361. Kolterman, C. E., &Gorelick, S. M. (1992) Paleoclimatic signatures in terrestrial deposits. Science. 256, 1775-1782. Konikow, L. F., & Bredehoeft, J. D. (1974) Modelling flow and chemical quality changes in an irrigated stream-aquifer system. Wat. Resour. Res. 10(3), 546-562. Konikow, L. F. (1986) Predictive accuracy of a ground-water model - Lessons from a postaudit. Groundwater 24, 677-690. Koster, R. D. & Eagleson, P. S. (1990) A one-dimensional interactive soil- atmosphere model for testing formulations of surface hydrology. / . Climate 6, 593-606. 26 C. J. Vôrôsmarty et al. Kuhl, S. C. & Miller, J. R. (1992) Seasonal river runoff calculated from a global atmospheric model. Wat. Resour. Res. 28, 2029-2039. Larson, C. L. (1982) Some particular watershed models. In: Hydrologie Modelling of Small Watersheds (ed. by C. T. Haan, H. P. Johnson & D. L. Brakensiek), 407-434. American Society of Agricultural Engineers, St. Joseph, Michigan. Lean, J. &Warrilow, D. A. (1989) Simulation of the regional climatic impact of Amazon deforestation. Nature. 342, 411-413. Likens, G. E., Bormann.F. H., Pierce, R. S., Eaton, J. S., & Johnson, N. M. (1977) Biogeochemistry of a Forested Ecosystem. Springer-Verlag, New York, 146. Loague, K. M., & Freeze, R. A. (1985) A comparison of rainfall-runoff modelling techniques on small upland catchments. Wat. Resour. Res. 21(8), 229-248. McGuire, A. D., Melillo, J. M., Joyce, L. A., Kicklighter, D. W., Grace, A. L., Moore, B & Vorosmarty, C. J. (1992) Interactions between carbon and nitrogen dynamics in estimating net primary productivity for potential vegetation in North America. Global Biogeochem. Cycles. 6, 101-24. McLin, S. (1981) Validity of generalized lumped parameter hydrosalinity model in predicting irrigation return flow. PhD. dissertation, New Mexico Institute of Mining and Technology, 351. Miller, J. R. & Russell, G. L. (1992) The impact of global warming on river runoff. J. Geophys. Res. 97(D3), 2757-2764. Moore, B., Gildea, M. P., Vorosmarty, C. J., Skole, D. L., Melillo, J. M., Peterson, B. J., Rastetter, E. B. & Steudler, P. A. (1989) Global Ecology: Towards a Science of the Biosphere, M. Ramble, L. Margulis and R. Fester (éd.). Biogeochemical Cycles. San Diego, California, 113-142. NASA (1987) SAR: Synthetic Aperture Radar, Earth Observing System Instrument Panel Report: Volume Ilf, National Aeronautics and Space Administration, Washington, DC. 233 pp. Nemani, R. & Running, S. W. (1989) Estimating regional surface resistance to évapotranspiration from NDVI and Thermal-IR AVHRR data. J. Climate Appl. Meteorol. 28, 276-294. Nicholls, N. (1989) Sea surface temperature and Australian winter rainfall. /. Climate. 2, 965-73. O'Loughlin, E. M. (1990) Modelling soil water status in complex terrain. Agric. Forest Meteorol. 50, 23-28. Partem, W. J., Schimel, D. S., Cole, C. V. & Ojima, D. S. (1987) Analysis of factors controlling soil organic matter levels in Great Plains Grasslands. Soil Science Society of America Journal. 51, 1173-79. Raich, J. W., Rastetter, E. B., Melillo, J. M., Kicklighter, D. W., Steudler, P.A., Peterson, B. J., Grace, A. L., Moore, B. & Vorosmarty, C. J. (1991) Potential net primary productivity in South America: application of a global model. Ecological Applications. 1, 399-429. Richey, J. E. & Ribeiro, M. N. G. (1987) Element cycling in the Amazon Basin: A riverine perspective. In: The Geophysiology of Amazonia, (ed. by R. Dickinson). 245-250, John Wiley, New York. Richey, J. E., Nobre, C. & Deser, C. (1989) Amazon River discharge and climate variability: 1903-85. Science. 246, 101-103. Ropelewski, C. F., & Halpert, M. S. (1987) Global and regional scale precipitation patterns associated with El Nino/Southern Oscillation. Mon. Wealh. Rev. 115, 1606-1626. Running, S. W. (1992) A bottom-up evolution of terrestrial ecosystem modelling theory, and ideas toward global vegetation modelling. In: Modelling the Earth System (ed. by D. Ojima), 263-279. UCAR/OIES, Boulder, Colorado. Running, S. W. & Coughlan, J. C. (1988) A general model of forest ecosystem processes for regional applications I. Hydrologie balance, canopy gas exchange, and primary production processes. Ecological Modelling. 42, 125-154. Salati, E. & Vose, P.B. (1984) Amazon basin: a system in equilibrium. Science. 225, 129. Schmugge, T. (1985) Remote sensing of soil moisture. In: Hydrological Forecasting (ed. by M. G. Anderson & T. P. Burt), 101-124. John Wiley, New York. Shukla, J., Nobre, C. & Sellers, P. (1990) Amazon deforestation and climate change. Science 247, 1322-1325. Sellers, P. J., Mintz, Y., Sud, Y. C. & Dalcher, A. (1986) A simple biosphere model (SiB) for use within general circulation models. J. Atmos. Sci. 43, 505-531. Shuttleworth, J. (1988) Macrohydrology - the new challenge for process hydrology. J. Hydrol. 100, 31-56. Shuttleworth, W. J. (1991) Insight from large-scale observational studies of land/atmosphere interactions. Surv. Geophys. 12, 3-30. Skaggs, R. W. (1982) Infiltration. In: Hydrologie Modelling of Small Watersheds (ed. by C. T. Haan, H. P. Johnson & D. L. Brakensiek), 121-166. ASAE Monograph 5, St Joseph, MI. Solomon, S. I., Capone, C., Moreau, A., Dengo, M. & Lee, M. (1990) The geographic, environmental, economic, and social information system (GEnESIS) for artificial intelligence applications. 16th General Assembly, EGS, Wiesbaden, Germany. Stedinger, J. R., & Baker, V. R. (1987) Surface water hydrology: historical and paleoflood information. Reviews in Geophysics. 25(2), 119-124. Strebel, D. E., Landis, D., Newcomer, J. A., DeCampo, J. & Meeson, B. W. (1991) Collected Data of the First ISLSCP Field Experiment, Prototype Volume. CDROM, NASA, Washington, DC. Linked atmosphere-hydrology models at the macroscale 27 Streeter, V. L. & Wylie, E. B. (1979) Fluid Mechanics. McGraw-Hill, New York. 562. Susskind, J. (1992) TOVS pathfinder progress. Report presented at GSFC-DAAC Version O Meeting. November 1992. Swank, W . T. & Crossley, D. A. (1988) Forest Hydrology and Ecology. Ecol. Studies #66. Springer-Verlag, New York. Trenberth, K. E., Branstator, G. W. & Arkin, P. A. (1988) Origins of the 1988 North American drought. Science. JA1, 1640-45. US Department of Energy (1990) Building an Advanced Climate Model: Program Plan for the CHAMMP Climate Modelling Program. DOE/ER- 0479T. US DOE, Washington, DC. Vôrôsmarty, C. J., Moore, B., Grace, A. L., Gildea, M. P., Melillo, J. L., Peterson, B. J., Rastetter, E. B. & Steudler, P. A. (1989) Continental scale models of water balance and fluvial transport: an application to South America. Global Biogeochem. Cycles. 3, 241-265. Vôrôsmarty, C. J., Moore, B., Grace, A., Peterson, B. J., Rastetter, E. B. & Melillo, J. (1991a) Distributed parameter models to analyze the impact of human disturbance of the surface hydrology of a large tropical drainage basin in southern Africa. In: Hydrology for the Water Management of Large River Basins (F. H. M. van d e Ven, D. Gutnecht, D. P. Loucks & K. A. Salewicz (eds)), IAHS Publ. 201. IAHS Press, Wallingford, UK. Vôrôsmarty, C. J., Grace, A., Moore, B., Choudhury, B. J. & Willmott, C. J. (1991b) A strategy to study regional hydrology and terrestrial ecosystem processes using satellite remote sensing, ground-based data and computer modelling. Acta Astronautica. 25, 785-792. Vôrôsmarty, C. J. & Moore, B. (1991) Modelling basin-scale hydrology in support of physical climate and global biogeochemical studies: an example using the Zambezi River. Surv. Geophys. 12, 27l-311. Vôrôsmarty, C. J. (1991) Models of macro-scale hydrology for use in global change research: tests on two tropical river systems. PhD. Dissertation, Engineering Systems Design Program: University of New Hampshire. Wang, H . F. & Anderson, M. P. (1982) Introduction to Groundwater Modelling: Finite Difference and Finite Element Methods. W. H. Freeman, San Francisco. Waring, R . H., Rogers, J. J. & Swank, W. T. (1981) Water relations and hydrologie cycles. In: Dynamic Properties of Forest Ecosystems (ed. by D. E. Reichle), Cambridge Univ. Press, New York., IIBP, No. 23, 205-265, Wetzel, P . J. & Chang, J. -T. (1988) Evapotranspiration from nonuniform surfaces: a first approach for short-term numerical weather prediction. Mon. Wealh. Rev. 116, 600-621. Williamson, D. L., Kiehl, J. T., Ramanathan, V., Dickinson, R. E. & Hack, J. J. (1987) Description of the NCAR Community Climate Model (CCM1). NCAR Technical Note NCAR/TN-285+STR. Willmott, C. J., Rowe, C. M. & Philpot, W. D. (1985) Small-scale climate maps: a sensitivity analysis of some common assumptions associated with grid point interpolation and contouring. Am. Cartog. 12, 5-16. Winter, T . C , & Wright, H. E. (1977) Paleohydrologic phenomena recorded by lake sediments, EOS. 58(3), 188-196. World Climate Research Programme, W.C.R. (1991) Draft scientific plan for the GEWEX continental-scale international project. WMO Bull. Geneva. Zimmerman, D. A. & Wilson, J. L. (1990) Description of and user's manual for TUBA: a computer code for generating two-dimensional random fields via the turning bands method. GRAM Inc. Albuquerque, NM 104. View publication stats