THEME 3. LOSS OF WATERSHED FUNCTIONS

Papers on Watershed Functions

Process-based Distributed Hydrological Modelling

by Rob Vertessy¹

Introduction

All over the world, catchments are being manipulated through land use, leading to major changes in catchment water and material balances. Water quantity and quality are changing, sometimes for the better, but usually for the worst. Catchment managers requires tools to help them assess how particular forms of catchment stewardship will affect water values. The 'empirical age' of catchment treatment experimentation has served us very well but is now waning because of cost and time constraints. From the experiments we have learnt much about basic hydrologic processes and how different land uses affect catchment balances. Managers have also learnt that many of their pressing questions have a spatial dimension to them. For instance, how does topographic position effect the response of a hillslope to disturbance by forest logging.

Hydrologic models provide a way of forecasting how catchment might respond to different forms of management. Amongst the many different styles of hydrologic modelling sits the so-called 'process-based distributed' hydrologic models. The term 'process-based' implies that these models focus on the physical laws governing water and material (particulate and solute) movement in landscapes. The term ' distributed' climate, topography) is represented. This class of models is regarded as data-hungry and difficult to use, but also the only class of models suited to capturing the complex feedbacks that occur in hydrological systems when they are perturbed.

One of our main arguments for the use of such models is that they capture important spatial dependencies in hydrologic systems. Hydrologic processes and rates vary enormously in space according to system properties such as soil type, land cover type and rainfall rate. Similarly, the response of any catchment or part of a catchment to disturbance will depend on the particular distribution of system properties applying in that area. Furthermore, land use systems usually have a complicated mosaic pattern. We have observed that the particular configuration of that mosaic (eg. The proximity of a logging road to a stream) is a major determinant of catchment response to land use. This is another compelling reason to use process-based distributed hydrologic models.

Some example models

The Cooperative Research Centre for Catchment Hydrology has developed two such models, referred to as Topog and Macaque. These are suited to different scales and problems but reflect where the state-of-the-art with respect to process-based distributed hydrologic modelling. In this presentation we describe both models through their application to example catchment management problems.

Topog is suited to simulating the dynamic function of small catchment (usually less than 10 km2) over long time periods. It runs using timesteps ranging between minutes and days, depending on the particular processes of focus. An innovative feature of the model is that it uses a novel flow net to describe how water and entrained materials move laterally through the landscape. This is based on a contour and flow trajectory system which is demonstrably superior to conventional grid-based methods used in most other hydrologic models. Topog integrates the water, carbon, solute and sediment balances and is thus useful for exploring complex feedbacks between system properties. It includes a physiologically based plant growth module which allocates carbon to above and below ground compartments of various plant forms (including grasses, crops and trees), dependent on water, nutrient and light availability, ambient temperature, and soil salinity. Thus plant growth and water use and soil-water solute dynamics are closely linked. In the past, the model has been used to assess how global warming and elevations in atmospheric C02 concentrations might affect forest growth, evapotranspiration and water yield from catchments. Developed since 1989, Topog is now well documented and published in the international literature. It is also available free of charge over the internet at www.clw.csiro.au/topog.

Macaque has some of the above-ground modelling features of Topog, though does not simulate carbon assimilation and allocation. Further, its below-ground modelling components are much simpler, permitting application of the model to large catchment scales (up to 1,000 km²). Is has many features common with GIS-based models, principal amongst these is a dependency on grid-based digital elevation models, vegetation cover maps, soil maps and climate maps. Like Topog, it runs on a daily timestep, though is only able to predict runoff at this stage. Macaque is undergoing very rapid development at the moment and has recently been modified to run under WINDOWS NT in a Borland C⁺⁺ Builder environment. Because of its rapid development cycle it is currently not available for public distribution.

Model applications

Topog is described by way of an application to an agroforestry application in south eastern Australia. The context of the application is that extensive tree planting is being advocated across grazing and cropping lands which have become degraded from excessive groundwater recharge and salinity. Why? Essentially, after European settlement the indigenous deep-rooted evergreen forests were replaced by annual, shallow rooted vegetation which evapotranspired much less than the original vegetation. This permitted more rainfall to percolate to groundwater, thus raising watertables which brought ancient salt stores with them into the plant root zone and onto the ground surface in certain places. The environmental consequences of this have been drastic and people are now scrambling to reverse the problem via tree planting. The problem is that most landholders cannot afford to give up their entire land resources base to trees.

The impetus for our Topog modelling of this problem is that methods are needed to say where the maximum impact from tree planting can be gained. Furthermore, because of high climate variability we experience in Australia, there is also a sustainability dimension to this problem. Landholders are expecting to generate commercial quality and volumes of timber in landscapes which previously supported scattered woodland or open forest. They are planting fast-growing species such as Tasmanian Bluegum (Eucalyptus globulus) using traditional commercial plantation management techniques. In intermediate annual rainfall areas (c.800 mm) initial growth has been very impressive, but once the soil water profile has been depleted the trees become susceptible to drought stress. Even mild droughts are sufficient to result in the mortality of over half the trees planted.

We have used Topog to design alternative plantation design systems, and to predict their performance over full rotations, usually lasting 30 years or so. Our focus has been on sloping terrain where there is some prospect for lateral water movement downslope via subsurface and surface pathways. These lateral flows can thus complement the natural rainfall and enhance tree growth. Specifically, we have compared the water balance and growth performance of two systems, namely block planting (as widely practised now) and strip planting. For the same total area planted our model results reveal very different outcomes in tree growth and thus water valance benefit which is positively correlated to growth. On the basis of our model results we are advocating the establishment of widely spaced and thin belts of trees on sloping land, which can benefit from lateral flows of water yielded from the interbelt grassland which is grazed by animals.

We describe Macaque through an application to water yield prediction on a large (161 km²) forested basin near Melbourne, Australia which is relied upon for urban water supply. The problem here is that managers have observed a strong relationship between forest age and water yield, with old growth forests aged 200⁺ years yielding about twice as much runoff that regrowth forest aged 30 years. Hence the demography of the forest is of critical importance to water supply. The catchment managers (Melbourne Water) need a means of forecasting the water yield consequences of wildfire and different logging regimes. Macaque has been used to estimate how basin water yield might change for different forest burning and logging patterns. To test Macaque we applied it to the Maroondah basin for an 85 year period spanning 1910-1995. During this period (primarily in 1939, but also on other occasions) much of the basin was burnt by wildfire, resulting in significant water yield changes. Macaque was able to simulate these changes effectively, giving us confidence that it can now be used in forecasting. It is currently being applied to the nearby Thomson basin (500+ km2) to help resolve a heated debate between water and timber harvesters regarding the relative value of wood and water and the effects of forest logging on the economic output from the catchment.

The problem of data

As stated earlier, Topog and Macaque belong to a class of models regarded as data hungry. Indeed, the need for data (for input, calibration and testing) has been a major constraint in their application, the models have really only been applied to intensively monitored research sites where sufficient banks of data exist. However, we are witnessing very rapid and exciting developments in spatial data acquisition which are set to transform the utility of these models and enable much wider application.

One of the most exciting developments is the advent if rainfall radar. Our previous large scale hydrologic modelling work has told us that errors in rainfall are by far and away the biggest source of error in hydrologic models. When modelling large systems it is imperative that the space-time distribution of rainfall is accurately represented. Doppler radar systems being operated by weather forecasting agencies such as the Bureau of Metheorology in Australia now enable the space-time distribution of rainfall to captured for input to models.

For many years, conventional remote sensing (e.g. AVHRR, LANDSAT, SPOT) has been of great use to hydrologists in terms of land cover mapping. In fact, our own Macaque modelling has depended on spatial maps of Leaf Area Index (LAI) derived through the calculation of greenness indices (eg. NDVI) obtained via LANDSAT-TM scenes. In more recent times however, aircraft-mounted hyperspectral (ie. Broad and intensively sampled band width) scanner data are revolutionising our ability to 'type' land cover. Such data can now be gathered at sub-1 m pixel resolution over large tracts of land and yield information on canopy structure and photosynthetic activity. Soon, we will be able to derive long transects of tree growth and water use to test our distributed models.

Until now, one of the main constraints to the application of distributed hydrologic models has been the mystery surrounding everything below the ground. Soil hydraulic properties have a major bearing on how water and materials move through the landscape. Yet, traditional field survey methods are so logistically demanding that soil properties cannot be sampled on anything but the smallest of study sites. The increasing using of aircraft mounted gamma radiometrics sensors are about to change all that. Used by geophysicists for some years in mineral prospecting, environmental scientists are now discovering that gamma radiometrics can reveal all kinds of geomorphic and pedologic features. For instance, CSIRO scientists have been able to derive maps of soil hydraulic properties over large tracts of forested land deposition of sediments have also been obtained, providing a new way of testing distributed models of erosion and deposition.

Last, but not least, new generation laser altimeters mounted on aircraft are yielding high resolution digital elevation models (DEM's) which will have a huge impact on our ability to simulate distrubuted hydrologic process. Horizontal resolution of 1 m and vertical resolution of 25 cm can be obtained at a cost of only $4000 per 10 km². In a couple of years time, such sensors will be based on satellites (probably with slightly coarser resolution), providing world wide coverage of high resolution topographic data. This will be define to define hydrologic flow paths.

The four new sensing technologies are still in their infancy and require much more testing. However, in the next decade, almost all of this technology will be operationalised, permitting routine process-based distributed hydrologic modelling over large tracts of land in almost any part of the world. In the meantime, catchment modellers must start to adjust their models to exploit these new forms of data.

Footnote_______________

¹ Cooperative Research Centre for Catchment Hydrology, CSIRO Land and Water, GPO Box 1666, Canberra, ACT, 2601, Australia