Contents - Previous - Next

This is the old United Nations University website. Visit the new site at http://unu.edu

Effect of surface cover on land surface processes

Major Types of Surface Cover

The effect of variations in land surface cover or the effect of changes in surface cover may influence general circulation models directly through the parameters of the atmospheric model or indirectly through the hydrologic coupling. In the context of general circulation models and the study of climate, we are concerned with broad classifications of surface cover and with long-term effects. From the point of view of hydrologic and atmospheric processes, regional land surfaces can be classified into the following zonobiomes (Perrier 1982):

1. Evergreen equatorial forest, which occurs in countries with wet climates (2,0007,000 mm per year) and high, fairly even air temperature (around 27°C). It is characterized by a very dense canopy. Precipitation exceeds evaporation throughout the year and there is a continuous growing period.
2. Dense ombrophile forest, which occurs in some warm, humid areas in the Southern Hemisphere. This evergreen forest thrives at mean temperatures between 10° and 18° C and rainfall above 2,000 mm. The water balance becomes positive every year.
3. Boreal coniferous forest, which covers a large portion of the land surface area of the Northern Hemisphere, is characterized by a short growing period. The density of the forest cover increases with the number of days with a mean temperature above 10°C. The root zone is quite shallow (200 - 400 mm). The water balance is positive over the year. Interception of precipitation by the canopy reaches values of 50%, compared with 10% for deciduous temperate forest, (7) below.
4. Mediterranean forest, characterized by rainy winters and dry summers with an annual rainfall between 800 and 2,000 mm. The mean annual water balance is positive, but may be negative for any one year.
5. Monsoon forest, which sheds leaves during the dry period in spite of a total precipitation between 1,000 and 2,000 mm. This land cover may be considered as intermediate between the evergreen forest and the savanna.
6. Deciduous boreal forest, which replaces the boreal evergreen forest as the climate becomes more continental. When the climate becomes more temperate, the boreal forest is replaced by mixed forest and ultimately by temperate deciduous forest.
7. Deciduous temperate forest, which includes a tree stratum, a shrub stratum, a bush layer, and sometimes a herb layer and completely covers the soil for over six months. This type of forest intercepts only 10 to 15% of precipitation. Rainfall is well distributed throughout the year and there is positive water balance in winter and spring.
8. Savanna, which develops over large areas where the duration of the dry season is increased and the precipitation does not exceed 1,500 mm. The biomass of the roots is relatively high and is important for survival during the dry season.
9. Temperate grassland, which occurs with cold winters and hot summers and also has a high root biomass. Water interception is high. The annual water balance is positive, but during the summer there are long periods of negative water balance.
10. Steppes, characterized by intermittent vegetation that has a strong root system and is adapted to limit water loss. Evaporation is relatively low and for rainfall of about 300 mm the runoff may be 100 mm or more.
11. Tundras, which develop when only a few months have mean temperatures above 0° C and which develop sparse vegetation during the summer. In summer the runoff creates swamps and the evaporation is close to the potential rate. During the rest of the year only sublimation will take place.
12. Deserts, which have practically no vegetation and may be either hot desert with annual rainfall less than 50 mm or cold deserts of permanent snow and ice and low rates of precipitation and sublimation.

The division between these zonobiomes depends largely on precipitation and temperature. The water balance and the individual hydrological processes will follow a distinctive pattern in each case. Types (1) to (7) have a closed canopy and well-covered soil surface. Types (10) to (12) approximate to a bare or sparsely covered soil surface.

An estimate of the distribution of major agricultural land surface types for each of the continents due to Perrier (1982) is shown in table 2 below.

TABLE 2. Distribution of land surface types (areas in 106 km, based on FAO data)

Direct Effects of Surface Cover

Surface cover affects atmospheric processes directly through albedo, emissivity, and surface roughness. Values of albedo are available for the ocean at different latitudes, cloud cover, and sea state; for snow of various structures and for ice of different types; for soils of various colours, both dry and moist; for crops of various types at various stages; for forests of different types at different times of the year.

For general circulation models what is required is a set of reliable values of regional albedo over a wide area for each month of the year. These regional values may either be parametrized from point measurements at observing stations or measured on a scale of about 150 mm by 150 mm by scanning radiometer from a satellite. The variation of the short-wave albedo for various surfaces has been recently reviewed by Kondratyev in the context of general circulation models, and table 3 shows the longterm albedo values for each month of the year for a number of landscape-climatic zones in the Northern Hemisphere (Kondratyev et al. 1982).

TABLE 3. Typical regional values of albedo

The actual assumptions used for albedo in general circulation models have been reviewed by Carson (1982). The most notable study of the effect of albedo change relates to the hypothesis by Charney (1975) that increased albedo in deserts produces a positive feedback by increasing the tendency to subsidence in the atmosphere and thus further inhibiting precipitation.

The range of long-wave emissivity (E) is much narrower than that for albedo. Reported values vary from 0.85 for yellow loam and 0.87 for black chernozem soil to 0.99 for fresh snow (Kondratyev et al. 1982). The value given for forested areas is usually 0.96 or 0.97. In view of the small range of variation in emissivity and of uncertainty in computing surface temperature, the use of an emissivity value of 1.0 for all surfaces is unlikely to be a significant source of error. However, the question of the transfer of long-wave radiation within a forest stand is important for the transfer processes of heat and water (Monteith 1976) and may have to be taken into account in specialized regional models or in a general circulation model designed to explore the effect of vegetation changes.

The surface roughness affects the momentum transfer in the atmospheric boundary layer. The approximate order of magnitude of the momentum roughness parameter for different surfaces is given in table 4 below (Deacon 1953; Baumgartner and Kirchner 1980; Brutsaert 1982; McNaughton and Jarvis 1983).

Thus, deforestation for cropping involves a change of one order of magnitude in surface roughness and deforestation for grazing involves a change of two orders of magnitude. Some general circulation models use roughness lengths of 0.1 mm for the ocean and 100 mm for the land. Numerical experiments indicate that an increase in roughness length would decrease the wind speed and the surface temperature and tend to produce more rapid infilling of cyclones (Delsol, Miyakoda, and Clarke 1971). The Study Conference on Land Surface Processes in Atmospheric General Circulation Models suggested a numerical experiment to see whether the cutting of the forests in the Amazon Basin would produce an effect on climate through changes in the roughness alone (Eagleson 1981).

TABLE 4. Order of magnitude of roughness parameter

Baumgartner and Kirchner (1980) have reported on a simulation of a deforested world. They estimated an increase in global average surface albedo from 0.167 to 0.174, with the largest increase in the tropics and in the Northern Hemisphere, a decrease in the global average value of the roughness parameter from 150 mm to 30 mm (with the largest decrease in the Northern Hemisphere), and a decrease of roughly one degree in the average deviation angle between the surface geostropic wind at the isobars. They were of the opinion that the greatest impact would be on the carbon dioxide cycle.

Hydrologic Effects of Surface Cover

Surface cover has a marked effect on hydrological processes such as infiltration and evaporation and hence can have indirect effect on climate through the interaction between the hydrosphere and the atmosphere. Physical hydrology is concerned with variation at a local scale and with short-term changes. Parametrization of the effects of surface cover changes to GCM scale is a major undertaking (Dooge 1982).

Forests in good condition promote high infiltration rates and facilitate the recharge of ground water. Table 5, due to Lull (1964), shows the infiltration capacity (i.e. rate of potential infiltration) for successive stages over 100 years of the reversion of abandoned pasture to forest in the south-eastern United States.

The type of forest management practices will also have an influence on rates of potential infiltration as shown in table 6, which is also due to Lull (1964).

The high rates of potential infiltration under forests are due to a combination of soil structure and adequate soil moisture deficit due to high evaporation.

In the case of vegetated surfaces, the total evaporation comprises evaporation from the ground surface and evaporation of intercepted and transpired water. The distribution of the total evaporation between these three components depends on the nature of the vegetation stand and its effect on the distribution of energy and on the movement of water vapour within the canopy. Table 7 shows an approximate distribution of the three components of evaporation as given by Baumgartner (1967). This data can be supplemented by figures available for interception and stomata! resistance for different types of vegetation (Rutter 1975).

TABLE 5. Potential infiltration rate for successive vegetation

TABLE 6. Effect on infiltration of forest management

TABLE 7. Distribution of components of evaporation (% of total evaporation)

The influence of vegetation on runoff and on the prevention of soil erosion is also important for the hydrological cycle and hence indirectly for general circulation models. Information is available on these aspects in the contributions to the International Symposium on Forest Hydrology, edited by Sopper and Lull (1967). In particular, reference can be made to the paper by Pereira on effects of land use on the water and energy budgets of tropical watersheds and to the paper by Hibbert on forest treatment effects on water yield. The general problem of the effect of land use on water resources has been reviewed by Pereira (1973). Balek (1977) discusses the role of the tropical forest and of the savanna in the hydrological cycle. Tropical agricultural hydrology is discussed in the conference proceedings edited by Lal and Rusell (1981), and the question of the hydrological effects of deforestation of the tropical rainforest is discussed in a paper by Lal (1981). Bosch and Hewlett (1982) reviewed the effect of changes of vegetation on 94 experimental catchments, including the 39 catchments previously reviewed by Hibbert (1967).

The hydrology of humid tropical regions was the subject of a symposium at the recent IAHS General Assembly at Hamburg (IAHS 1983). Particular reference may be made to the paper by WMO (1983) on operational hydrology in humid tropical regions and the paper by Brinkmann (1983) on the nutrient balance of a central Amazonian rain forest.

Seeking new concepts of macro-hydrology

Soil Moisture-Vegetation-Climate Equilibrium

For the past few years a number of research hydrologists have been seeking to avoid parametrization from the microscale of hydrological physics by seeking to establish new laws of hydrological behaviour at the scale of the catchment and the region. If hydrological laws on a catchment scale are to be discovered then they must depend on some type of uniformity or regularity other than that which provides the foundation of the laws of fluid mechanics used at the scale of hydrological physics. It will be recalled from previous mention in this paper that the assumptions underlying the commonly accepted laws of fluid mechanics are at variance with the assumptions of physical chemistry that are necessary to explain the physical properties of water. This suggests that the hydrologic laws at catchment scale need not resemble the laws of hydrologic physics at a local scale. If such marcoscale laws can be developed and verified then they would provide a hydrologic input at a scale compatible with general circulation models. Sufficient progress has been made to indicate that the further pursuit of this approach would be worthwhile. Accordingly, a brief review will be given here of some of this work.

The first successful attempt to derive laws at the macroscale was in the study of the composition of drainage networks and in the detection of regularities arising from geomorphological equilibrium. Smart (1972) has written a review of the first 20 years of this work on quantitative geomorphology. More recently the regularity in the pat tern of the drainage network has been linked to the regularity of the shape of the direct catchment response to storm rainfall (Rodriguez-lturbe and Valdez 1979).

More immediately applicable to general circulation models is the work by Eagleson (1978, 1982a) on the equilibrium relationships between climate, soil moisture, and vegetation. Eagleson (1978) used a statistical dynamic approach in order to incorporate the short-term dynamics of the soil moisture accounting into the long-term water balance. He modelled the rate of potential infiltration at a wet surface by a formula of the type given by equation (22) above, with A and B dependent on the amount of moisture (assumed uniformly distributed) in the root zone. The rate of potential evaporation at a dry surface (sometimes called exfiltration) between storms was modelled by a formula of the same type, but with the addition of a term representing the uptake of the soil moisture by the plant roots.

Using representative probability density functions for the independent climatic variables (interstorm interval, storm duration, rainfall intensity, and rate of potential evapotranspiration) and assuming an initial moisture content and the corresponding rate of deep percolation, Eagleson (1978) derived the probability density functions of the actual infiltration during storms and the actual evaporation between storms. The average number of storms in the rainy season was then used to find the average volumes of infiltration and evaporation in the average long-term water balance. If water-table variations can be ignored, the water balance can be defined in terms of three soil parameters (effective porosity, saturated intrinsic permeability, and an index of pore disconnectedness) and one vegetation parameter.

Figure 5 shows the relationship between the ratio of the actual to the potential evapotranspiration as a function of a compound climate-soil parameter for the case of bare soil. The relationship is very similar for vegetated surfaces (Eagleson 1978). For homogenous conditions the water balance at a point can be extended to the macroscale through the use of areally-averaged atmospheric inputs and of lumped physical parameters.

In a further development Eagleson (1982b) suggested that we should look to the state of the natural vegetation canopy, which is readily observable, for information in the time and space average of the actual evapotranspiration. He studied catchment systems whose biology is limited by either available water or available energy. Noting the maximization of soil moisture at an intermediate canopy density (L'vovich 1974; Eagleson 1978), Eagleson (1982a) hypothesized that a water-limited natural vegetal system would produce a canopy density that created minimum water demand stress under the local climate and soil conditions. This is the equivalent to assuming that the system would seek an equilibrium involving maximum soil moisture. For moist climate-soil systems Eagleson hypothesized that the system would tend to maximize biomass productivity for the given amount of available energy. This is equivalent to maximization of canopy density. Applying these hypotheses to his 1978 model, Eagleson (1982a) derived the equilibrium conditions shown in figure 6 for the relation between the species-dependent plant coefficient ki (i.e. ratio of unstressed rate of transpiration to potential rate of bare soil evaporation) and the vegetation canopy cover M (i.e. shadowed area). Preliminary results tend to confirm the predictions for humid-arid conditions (Eagleson and Tellers 1982).

FIG. 5. Dynamic function for evaporation efficiency

FIG. 6. Vegetation characteristics

Problems of Spatial Variability

General circulation climate models operate on a resolution of about 250 km and provide outputs averaged on a monthly basis. Precipitation (and consequent soil moisture changes) are not adequately defined at such scales and are better represented at resolutions of about 10 km and one day. This poses an important research problem of parametrization. In catchment hydrology the problem of rainfall representation has been tackled by simulation based on the estimated spatial correlation structure (Mejia and Rodriguez-lturbe 1974).

Some studies have also been made of the effect of spatial variation in soil properties on the water balance of small catchments. Peck, Luxmoore, and Stolzy (1977) simulated heterogeneity of the micro-geometrical soil scaling factor (Miller and Miller 1956) for a 9.7 hectare catchment in Tennessee. The mean values of the measured soil properties were used and the spatial variability in the soil scaling factor taken as normally distributed with a coefficient of variation of 0.25. The authors concluded that this particular spatial variability did not give rise to significant changes in the water balance of the 9.7 hectare catchment and that the lumping of the values of the soil properties at their mean values was justified.

Sharma and Luxmoore (1979) made similar studies based on a Monte Carlo simulation, but came to a different conclusion. They assumed the soil scaling factor to be log-normally distributed with a coefficient of variation of 0.6 and used fourteen different values of the scaling factor. They applied the simulation to a 9.6 hectare field in Oklahoma. Using the Penman-Monteith equation for evaporation, they calculated the water balance changes at 15-minute intervals during precipitation and at hourly intervals otherwise. Monthly values of the components of water balance were computed for all 14 values of the scaling factor. They found, as might be expected, that with increasing values of the scale factor the deep drainage increased and the evaporation and surface runoff decreased. Once surface runoff was generated, the amount increased very rapidly for lower values of the scaling factor. The effects of spatial variability were amplified in months with high rainfall and low evapotranspiration. The effects were also amplified when the normal distribution was used instead of the log-normal. In contrast to the results found by Peck, Luxmoore, and Stolzy (1977), Sharma and Luxmoore found that the water balance, simulated by using the mean scale factor alone, was significantly different from the integrated response predicted on the basis of either the log-normal or the normal distribution.

Freeze (1980) also applied Monte Carlo simulation to spatial variability of saturated hydraulic conductivity. He applied stochastic rainfall patterns to both a Horton infiltration model and a variable source area model, in each of which there was an assumed correlation structure in the hydraulic conductivity distribution. It was found that the statistical properties of the predicted runoff were quite different from those predicted on the basis of a lumped value of hydraulic conductivity equal to the mean value.

Milly and Eagleson (1982a) adapted the statistical dynamic approach used previously by Eagleson (1978) for time-averaging to the problem of studying the effects of the spatial variability of the storm inputs and of the soil on the infiltration function. They found that increased spatial variability almost invariably leads to decreased infiltration and increased surface runoff. They also applied this approach to the effect on average areal evapotranspiration, of spatial variations in surface roughness, canopy resistance, and available energy (Milly and Eagleson 1982b). This problem is complicated by atmospheric advection, which results in the downstream evaporation being affected by the evaporation further upstream. The effect of local advection was shown to be most significant when there was a large variation in canopy resistance over the area.

Proposed Large-Scale Experiment

If hydrologic inputs to general circulation models are to be improved, then reliable field measurements at a scale comparable to the GCM grid scale must be available to enable a rational choice to be made between alternative methods of simulating the coupling of the hydrosphere to the atmosphere. The Study Conference on Land Surface Processes in Atmospheric General Circulation Models held in January 1981 recommended that consideration be given to the organization of a joint hydrologic-atmospheric experiment over a watershed of grid square size to validate current and future hydrologic parametrization in detail (Eagleson 1981).

The conference working group on the physics of land surface processes and their parametrization recommended the use of a catchment hydrology model to generate "synthetic areal observations" and the evaluation of the hydrologic parametrization of these synthetic observations. The working group stated that the model should (a) account for soil moisture explicitly, (b) should provide as output areal values of soil moisture, actual evapotranspiration and runoff, and (c) should be capable of producing realistic runoff (daily, 10-day, monthly, yearly) for the main river basins. It was suggested that the first set of experiments be limited to regions of abundant rainfall, evaporation, and runoff, like the basins of the Amazon, Zaire, Ganges, and Brahmaputra, Mekong, and so forth.

The suggestion by the Conference for a large-scale joint hydrology-atmosphere experiment was endorsed in principle at the second meeting of the Joint Scientific Committe of the World Climate Research Programme in March 1981. The question was again considered by the JSC at its third meeting in Dublin in March 1982. At the latter meeting, the following additional criteria were proposed: (i) the experiment should be carried out in both a wet and a dry season to provide data against which parametrization could be tested under the extremes of soil moisture; (ii) the climate should be continental; (iii) the host nation (or nations) should have good infrastructure and technical support capability; and (iv) the results should have immediate economic utility to the host country. It was also recommended that when confidence was felt in the results likely to be obtained from the climate models, the climate sensitivity to the following land surface perturbations should be tested: deforestation (e.g. the Amazon Basin); drainage (e.g. the Sudan region of the White Nile); irrigation (e.g. the central United States).

The question of land-atmospheric processes was a special topic at the third meeting of the JSC in Venice in March 1983. Seven scientific presentations on hydrology and land-atmosphere processes were made to the Committee and the first two days of the meeting were devoted to this topic. The Committee agreed with the conclusion of the invited experts that further progress in the parametrization of hydrology and land surface processes on a scale suitable for climate models would require a conceptual development. It was also agreed that such conceptual advances would be greatly helped by the acquisition of a consistent and comprehensive data set over a representatively large and heterogeneous area. It was agreed that embarking on a full grid size would not be warranted and that a less ambitious pilot project limited to 50 km x 50 km (or almost 100 km x 100 km) should be considered first to establish the feasibility and accuracy of the proposed measurements. The purpose of such a pilot experiment would be to provide, in addition to the best possible conventional hydrological measurements for the catchment, independent records of total actual evaporation as well as simultaneous values of the relevant radiative and aerodynamic quantities near the surface.

The proposal of the Venice meeting of the JSC for a pilot experiment was developed by a consultant expert and his report was considerd by a multidisciplinary meeting of experts at the end of November 1983. This meeting of experts supported the proposal to organize a pilot atmospheric-hydrological experiment and formulated a number of actions that would need to be pursued with regard to its planning and implementation. They expressed as the main objective of the experiment the provision of data that would:

1. improve our understanding of the hydrological processes at the air-surface interface ranging from a few kilometres up to 100 kilometres and, in particular, aid in estimating evaporation by as many different independent methods as possible;
2. test current parametrization schemes for the prediction of evaporation, soil moisture, and runoff in climate models and help develop and test new conceptual models with a meaningful physical basis;
3. aid in the interpretation of remote sensing data through the provision of concurrent remote and ground measurements.

The expert group recommended that, if the pilot experiment were successful, further experiments be organized so as to cover as soon as possible (i) a temperate climate region with forest, agricultural land, and pasture, (ii) a semi-arid region, (iii) a northern region with evergreen forest giving a long snowy period and high summer evapotranspiration, and (iv) a region of evergreen equatorial forest.

The expert group endorsed the proposed size of area between 50 km x 50 km and 100 km x 100 km. They felt that to increase the scale beyond the upper limit would not increase significantly the variability of the surface conditions and of the hydrological processes but would bring in an undesirable variability of atmospheric conditions. The group recommended that the hydrological measurements should con tinue over three years with background flux measurements over one year and shorter intensive periods of measurements in the atmospheric boundary layer by radiosonde and aircraft.

The report of the multidisciplinary group of experts was approved by the JSC at its meeting in March 1984 and authority given to establish a steering committee to plan for the implementation of the project in 1986-1989.

Contents - Previous - Next