The Project:
• Why SEA-BASINS
• Implementation
• Partner Network

NAGA
• Strategy
• SE Asia Models
• Informatics

Active Basins
• Mekong Basin
• Sub-basins

• Restricted Papers
• Restricted Data

NAGA: Strategy

The overall challenge is to understand and capture the spatial and temporal scale of both anthropogenic demands and natural processes in moving water across the landscape to the ocean relative to the key cross-boundary environmental issues confronting resource managers. This must be done with sufficient detail to be relevant locally for specific issues, and to map to a broad, often cross-boundary geographic base. This must be done in a quantifiable and ultimately predictable manner. But regional dynamic analyses based on detailed data are generally difficult to realize; data are simply too sparse. Computer models coupled to field observation systems can provide regional interpolations, use what data do exist as validation points, and in the process estimate overall dynamics in areas where data are unavailable. Models consist of the scientific theory of physical processes affecting a region, embodied in a computer program. While imperfect and at risk for misuse and misinterpretation, such models do summarize the cumulative understanding of a problem in ways that can be tested. A final challenge is how to perceive and to present the reality of the extremely complex information and data that would be derived through such a process. The challenge is to convey the resulting information to policy makers (perhaps through the utilization of enhanced visualization techniques).

In the following paragraphs, we derive the overall analytical approach we use to represent the characteristic structure and dynamics of drainage basins in the overall NAGA environment.

The Characteristic Structure and Intrinsic Spatial and Temporal Scales of Drainage Basins

The drainage basin, as the landscape through which all waters flow from their highest source before draining naturally to the sea, can be considered to be a fundamental organizing unit of the land surface and its population. A drainage basin is defined by landforms (which do not necessarily correspond to political boundaries). Their edges are the ridges and hilltops that direct water into a stream or river, and their surfaces are the landscapes where communities grow. The overall balance of energy over a watershed affects how water is partitioned between the atmosphere, soil, and river channel. Rain or snow falls to the landsurface, some of which is returned to the atmosphere (through the process of evapotranspiration), some is stored in the soil (as soil moisture), and the balance drains into stream networks (where it provides river flow, or discharge).

As streams descend, tributaries and groundwaters add to their volume, creating ever-larger rivers. As rivers leave the highlands, they slow and start to meander and braid. In the lower stretches of unmanaged rivers, water moves between the river's mainstream and its floodplain, modifying the flow regime and creating critical ecological niches. River and floodplain ecosystems are closely adapted to a river's flooding cycle. The diversity of a river lies not only in the various types of land surfaces (or land uses) it flows through but also in the changing seasons and the differences between wet and dry years.

Disruption of the linkages between the landscape and rivers and between rivers and their floodplains through human intervention fundamentally alters the nature of riverine ecosystems. The impact on river systems occurs through erosion of the land surface, changes in the nature of the sediment and its associated organic matter, and nutrient content from agricultural and urban sources. Changes in hydrology are immediate consequences of dam construction and large-scale water diversion for irrigation. Longer-term changes in regional weather patterns and climate will result in altered flow regimes and thus impact downstream ecosystems including the coastal zone. Coastal ecosystem production relies very strongly on material inputs from the land. Deterioration of water quality, due to natural causes such as salt and acidity, and anthropogenic causes such as domestic, agriculture and industry, is problematic in most if not all countries.

Hence the hydrology of a small stream is determined by the larger watershed that stream is nested in, which probably has its headwaters in another county or even country. Impacts of land-use change on an entire river basin cannot be defined by simply "summing up" the impacts observed on individual streams; extrapolations based on "scaling" must be made. Overall, we must recognize the spatial and temporal relationships between dynamic ecosystems within river basins, where a landscape is composed of ever-changing elements, according to how the system is observed.

So what are the scales at which watersheds and their models can realistically be represented? Ultimately, this is a tradeoff between how finely individual processes can be described, the data that are actually available, and the ability of computers to make calculations. Theoretically, there is a continuous range of scales represented in a watershed. In practice, the nature of how observations can even be made imposes some serious constraints. For example, an observer flying over a section of the Amazon floodplain will see a series of discrete and recognizable landscape patches, made up of streams, main river, pastures and forest, and open and closed lake environments.

What the environment looks like as seen with observation systems appropriate at different scales: The Amazon floodplain from an observer in a light airplane (upper left), the Amazon floodplain from LANDSAT TM (30m pixels, lower left), a section of Rio Xingu from AVHRR (1km pixels, upper right), and Taiwan at 1km (lower right, left) and at 100km (lower right, right).

The processes occurring in each patch can be measured directly. A landscape of this scale can best be described by satellites of the type LANDSAT Thematic Mapper (TM), which can "see" up to 180 kilometers in extent. But the distinction of what can be observed is reduced to 30-meter homogeneous patches ("pixels") in fewer spectral bands than the eye itself will observe, resulting in a more blurred view of the landscape. In moving to larger scales, which is necessary to characterize regions, our ability to observe distinct features is degraded further. For example, the Advanced Very High Resolution Radiometer (AVHRR) satellite can "see" for hundreds of kilometers, but with no better than one kilometer-sized pixels, in even fewer bands. As a result, for example, the Rio Xingu in the Brazilian Amazon is barely distinguishable, never mind details of the floodplain. In moving from the AVHRR-scale view to the global scale pixels of one-degree latitude/longitude (about 100 km on a side in low to mid latitudes) large regions are rendered essentially uniform (e.g. Taiwan becomes 2 pixels). At these larger scales, specific objects in space are less important then the general "pulse" between longer temporal patterns and larger spatial extents. Time and magnitude of spatial variance are perhaps the most important variables at these scales. Issues of data aggregation and temporal phasing drive the modeling requirements.

The temporal scales of watersheds cover as much range as do the spatial scales. While the shape of the landscape itself (hills, valleys) will evolve (on geological time scales), these changes occur much more slowly than the seasonal and yearly evolution of the landcover and landuse. These seasonal changes in turn occur much more slowly than a rainstorm, whose characteristic time scale is on the order of minutes to hours. The processes with which we are concerned, for example the translation of rainfall into runoff across different types of landscapes, are themselves described differently at different space scales. The overall ensemble of these landscape features, which we can observe at multiple scales, can be thought of as the "physical template" upon which the more rapidly dynamic processes can operate.

Spatially-explicit, Dynamic Models of Large-scale Basins

The "Earth System Science" approach we use to capturing the spatial and temporal dynamics of drainage basins is to build a spatial data model of the slowly changing structure of the region (what we call the physical template), the more rapidly evolving landscape, and then bringing that template "to life" by deriving dynamic models of the climatology, water and material movement, and forcing by human populations.

The Template - The "slowly-changing" elements of the landscape (topography, river networks, and soil texture) can be considered as the (geo)physical template upon which the more dynamic systems evolve. Landcover and landuse (often determined through satellite imagery) can be considered as the biophysical template which changes more rapidly, and has attributes that may need to be updated on characteristic time scales from seasonal to annual (depending on the objective). It is in the biophysical template that human influences on the landscape can be expressed. For example, changes in the landcover and land use driven by changes in demography (and captured in demographic models), which affect the physical movement of water.

Schematized spatial model of the landscape and dynamics of the hydrological cycle.

To bring the multi-scaled physical template into a computational environment, the template can be thought of as being made up of a series of multiple elements, (called grid cells or drainage basin elements), each of which has uniform attributes which can be characterized. These attributes (themselves spatial models) must include those representations of the landscape which subsequent dynamic models require (e.g., the hydraulic properties of soil imparted by soil texture). A region is then made up of multiple grid cells, according to the resolution desired and what data is feasible to acquire. The ensemble of these attributes can be represented in a computer format with which calculations can be made. The emergence of Geographic Information System (GIS) computer software allows such an organization of information.

The template is more then a GIS database of thematic layers. It is the explicit geographically referenced statement of the relationship between these data layers over both space and time. The template captures the multiplicity of time and space scales over which the watershed environment changes and which humans must respond to. The template is restricted, of course, to the data available to describe it.

External Forcing - A significant challenge for biophysical models is providing the climatological data (or "drivers"). The two primary variables most widely available in long-term data archives are precipitation, and daily temperature minimum and maximum. Ideally, gridded fields of all the required model forcings would be derived directly from observations. However, in-situ data are far too sparse to support such an approach except for precipitation and temperature, which are observed at a comparatively dense network of climatological stations. (The typical average spacing over the continental U.S. is 30-40 km; with global spacing averaging 100-200 km). Therefore procedures to estimate all the forcing variables required by models based on a minimum set of variables have to be developed. Given the heterogeneous nature of precipitation and precipitation collectors, river discharge is perhaps the most robust integrator of the long-term hydrologic properties of a drainage basin.

Modeling the movement of Water (Energy) and Materials - Water moves between the atmosphere, the land surface, and the oceans. The problem is to describe these flows by sector. A tributary basin analysis is predicated on determining the interannual patterns in precipitation and runoff; runoff pathways and soil residence time can lead to differences in how chemical species are mobilized and particulates eroded. The hydrology of a regional-scale river system can be modeled as a geospatially-explicit water mass balance for each grid cell within the basin contributing to stream flow and downstream routing. A water balance model is applied at each individual grid cell over the defined region. It separates precipitation into evapotranspiration (a calculation which requires knowing solar radiation, temperature and ideally surface winds), soil moisture change (a calculation which requires information about soil texture and how much water reaches the soil surface, after being "intercepted" by the tree canopy), and runoff from the land surface to streams. The actual rates at which these processes occur fundamentally depend on such attributes as soil texture, rooting depth, slope, and other physical characteristics, which must be described for each cell.

Schematic of a pixel-based water balance model embedded in a spatial grid, where flow is subsequently accumulated and routed down a river network, eventually to the ocean.

The final step is to couple the spatial and hydrologic models to models of the origin and transport of selected chemicals. Water "quality," both in freshwaters and the immediate near-shore marine environments, is a crucial problem. Development of biogeochemical models is much more problematic than spatial and hydrological models. Large-scale models of chemical flux typically use a multiple regression approach of aggregate basin attributes and measured fluxes (e.g., Ludwig et al 1996). Such models are intrinsically not based on actual processes, and may be less sensitive to capturing regional differences or changes. More process-based models are required. Mayorga et al (2000), for example, proposed an integrated modeling framework aimed at quantitatively describing the dynamics of organic matter and nutrient cycling in mesoscale to large rivers.