Contents - Previous - Next

This is the old United Nations University website. Visit the new site at

Hydrological processes in tropical forests

A background discussion of the influence of the tropical forest upon the hydrological processes that precipitation undergoes on delivery to the land surface is pertinent here. The hydrological processes in question include precipitation, interception, evaporation, and runoff. Interception splits precipitation into that delivered to the land and water surfaces and that caught on the forest canopy and returned to the atmosphere by evaporation. Water delivered to the land surface may run off directly, as overland flow into streams to drain by way of rivers and lakes back into the sea, or infiltrate the soil. This latter pathway has been considered the most important pathway for the sustenance of man (Pereira 1973, 1). From the soil, vegetation is supplied; the surplus draining further down to springs maintains the steady flow of rivers. Plants return much of the soil water through transpiration to the atmosphere Some water also evaporates directly from the soil and from the surfaces of lakes and rivers. It is known that part of the water that infiltrates the soil moves laterally through the upper horizons until it reaches a stream channel and does not become part of the ground water reservoir. This portion of subsurface flow is known as interflow or through-flow and, together with overland flow, constitutes what is generally referred to as surface runoff or, more properly, as direct runoff or quickflow. In practice interflow and overland flow cannot be accurately separated, but several methods of hydrograph separation are available for isolating their sum from baseflow.

The forest is characterized by three primary elements: (i) the foliage above the ground forming a number of layers that compose the total thickness of the protective canopy, (ii) the accumulation of dead and decaying plant remains on the ground surface constituting the forest floor, and (iii) the forest soils that are formed below together with the living and dead roots and subsurface stems that permeate the soil. These three elements account for the observed distinctive movement and action of water in the forest.

The tropical forest, especially the rain forest, owing to its opulence, forms an exceptionally effective screen or filter of climate between the free atmosphere above and the ground below. The unique hydrological effects of this forest reflect its special microclimate or bioclimate and moisture regime. According to Tricart (1965) the tropical forest owes its properties to the density of the vegetation, which is three or four times more than that of the forest of the temperate zone. It is often claimed that the vertical structure of the tropical forest consists of five layers, three of which are tree layers. In South-East Asia the trees in the upper stratum, or the emergents, are 40-60 m with umbrella-shaped crowns some tens of metres apart. The second tree layer, 20-30 m in height, is also discontinuous, with the gaps usually occurring below the emergent trees so that the upper two layers together form a more or less continuous canopy. The lower stratum is between 9 and 15 m high in West Africa and consists of many trees with narrow crowns reaching for light. The shrub layer then follows before the ground stratum, which is usually patchy and consists mainly of tree seedlings and a mixture of herbs.


Lockwood (1976, 91) cited the work by Bergemann and Libby, who, in 1957, used isotopes of water to obtain the ratio of maritime water to land water in the Upper Mississippi Valley, and concluded that one-third of the average precipitation is formed of re-evaporated (i.e. continental) water and two-thirds of ocean water. One may be tempted to conclude from this study that it is unlikely that changes in the nature of land surface, such as the removal of the forest, will have any significant influence on local rainfall. In the subtropics, where continuously clear skies and large amounts of solar radiation are available to evaporate water, the main sources of water vapour are the oceans. The same assumption has been extended to South-East Asia, which receives much of its rainfall in the form of water that has evaporated from the subtropical Indian Ocean.

The reported results from the Amazon Basin, where the mean annual precipitation is 2,000-2,400 mm, are, however, different. Here, the mean recycling time for water vapour in 1979 was found to be 5.5 days (Salati et al. 1979 and Salati and Matsui 1981). Measurements of oxygen isotopes in the rain and river waters confirmed the importance of recycled water in the hydrological balance throughout the basin. The studies concluded that 52% of the precipitation in the Amazon region between Belém and Manaus was accounted for by inflowing moisture from the Atlantic Ocean; the remainder by recycled vapour within the area. An alteration of this water balance by man's activities (such as by deforestation) might significantly affect the precipitation input. It is also noted that the Andes form an effective western water barrier to the horseshoe shaped Amazon Basin, so that the main influx of advected water is due to the trade winds from the east and the apparent main outflow is via the river to the east. There is, however, no information on the dynamics of the rain-producing processes and on the effect that changes in the evaporation will have on cloud formation and the precipitation pattern. These aspects need to be studied separately during different seasons. Other important works on hydrological processes in the South American rain forest are Lettau, Lettau, and Molion (1979), Marques et al. (1977), Molion (1975), and Potter et al. (1975).

Henderson-Sellers (1981) used the three dimensional general circulation model to predict the potential climatic effects of Amazon deforestation. This suggested a likely decrease in rainfall of 600 mm per year and a small temperature increase. Newell (1971) suggested that Amazon deforestation might affect the atmospheric general circulation, but Dickinson (1980) concluded that even complete deforestation of the tropical region is not likely to cause global climatic changes in excess of the natural climatic fluctuations. If the predicted reduction of 600 mm is correct, it may induce irreversible ecological changes in some areas since the effect is not likely to be evenly distributed either in time or space"

Interception and Evaporation

The obstacles presented by the tropical forest to the free fall of rain will cause changes in the quantity, rate, and time of water delivery to the ground. Interception in the forest occurs at two levels within the forest cover: in the canopy and in the ground litter. Three main components of canopy interception can be identified. These are the interception loss (i), water retained by the crown surfaces and later evaporated; throughfall (T), water falling through and from the leaves to the ground surface; and stemflow (S), water that trickles along twigs and branches and finally down to the ground surface via the main tree trunks. Interception loss is a primary water loss as it represents water that never enters the soil. The amount depends on the ability of the forest to collect and retain rainfall (interception capacity), storm size and intensity, and evaporation rate. The density, type, and height of the canopy will affect the interception capacity.

In the tropical rain forest interception loss values for heavy rainfalls are less than for light falls. Observation in Peninsular Malaysia indicates that with a total annual rainfall of about 2,500 mm, the interception loss amounts to 450 to 500 mm, or 18-20% Most of the rainfall seems to reach the ground surface by the process of throughfall (Lockwood 1976, 98-99). In Nigeria Lawson, Lal, and Oduro-Afiriyie (1981) obtained the results in table 2 from 30 storms, each with rainfall in excess of 5 mm, sampled in a tropical forest in Ibadan during the May to September wet season in 1979. The value of interception loss ranges from 13 to 19% for monthly rainfall of at least 100 mm and averages 17%. The average throughfall was 73%, plus 10% stemflow. Table 3 shows the values of T. S. and I for other tropical forest locations, including India. The effect of storm size on interception loss in a tropical forest in Tanzania is illustrated in table 4. For gross storm amounts equal to or greater than 10 mm the interception loss ranges from 13.5 to 18%. Table 4b presents comparative data for temperate forest environments.

Finally, in an open deciduous seasonal forest in Uganda, Hopkins (1960) observed over a seven-week period a rainfall of 1,130 mm on the top of a tower overlooking the forest, but only 66.4% of it reached the ground. He also observed that nearly the whole loss occurred below 9.2 m, in the lower stories of saplings, shrubs, and herbs. It has also often been claimed that secondary tropical forests characterized by dense undergrowth are probably a more effective screen against rain erosion than a virgin forest, which is more open at the ground level.

Ground litter, like the crown canopy, detains and retains precipitation that reaches it. Evaporation takes place both from the canopy surface and the litter. The quantity of rain-water that evaporates (interception loss) represents a loss of water so far as water yield is concerned and this holds true also for the water evaporated from the forest floor litter, however, this loss is partly compensated for by the reduction in evaporation from the protected soils.

TABLE 2. Rainfall and interception components over forested watershed at IITA Catchment, Ibadan, Nigeria, 1979

fall (T)
tion (1)
May 5 89.4 68.4 9.5 11.5 77 11 13
June 6 107.8 85.0 8.9 13.9 79 8 13
July 7 238.8 170.0 29.8 39.0 71 12 16
Aug. 5 76.3 50.8 6.9 18.6 67 9 24
Sept. 7 132.8 98.8 9.3 24.7 74 7 19
Season total 30 645.1 473. 0 64.4 107.7 73 10 17

Source: Lawson, Lal, and Oduro-Afiriyie 1981, p. 143, table 3.11
aonly some rainstorm events were samples

Evaporation from bare ground may be intense but is a short-term phenomenon, as it only affects the surface soil and the vapour flux declines rapidly without further rain. In contrast, dense tropical forest continues to transpire water from deeper soil horizons. Thus the tropical forest serves as an important source of water vapour because it transpires water in larger quantities than most other forms of vegetation. The soil-drying action of the tropical forest is considerable. Lomee (1961) found it reached 2,0002,300 mm for a 25 m rooting depth and an available soil moisture of 4,200 mm in Java and in the Congo Basin between 1,230 and 1,510 mm under forest, while it was only 950 -1,100 mm under savanna.

Actual evaporation and transpiration losses from tropical forest continue throughout the year at near potential rates in many cases. For instance, near Manaus in the Amazon Basin transpiration plus interception losses accounted for 80.7% and 74.1% of the rainfall in two representative basins with an average rainfall of 2,000 mm; transpiration alone accounted for 62% and 48% respectively. Similarly, results from the Loweo Catchment in the Yangambi Forest Reserve in Zaire showed that evapotranspiration accounted for 63% of the basin rainfall of 1,500 mm.

Infiltration and Overland Flow

Water that reaches the soil surface either enters the soil or runs off as overland flow. Infiltration is a complex phenomenon because both infiltration rates and capacity vary with time. Water entering the soil can bring about physical changes in the soil that can also reduce the rate of infiltration. Such changes may include swelling of colloidal material and in-wash of fine particles into pore spaces and destruction of soil structure by the impact of rain drops.

TABLE 3. Throughfall, stemflow, and interception from rainfall in selected tropical forests

Forest type Location Gross rain-
fall (mm)
Percentage of total rainfall Investigator
fall (T)
flow (S)
tion (1)
Eucalyptus hybrid (1,658 trees/ha) India 80.7 7.7 11.6 Tejwani et al. 1975a
Shorea robusta (1,678 trees/ha) India 66.4 8.3 25.3 Dabral et al. 1963a
Shorea robusta (668 trees/ha) India 54.6 7.2 38.2 Dabral et al. 1963a
Pinus roxburghii (1,156 trees/ha) India 74.3 3.6 22.1 Dabral et al. 1969a
Alstonia scholaris (1,675 trees/ha) India 57.0 17.0 26.0 Dabral et al. 1968a
China fir plantation Taiwan 1,165 91.7 0.8 7.5 Sopper and Lull 1967, pp. 89-94
Zelkova plantation Taiwan 868 90.7 1.4 7.9 Sopper and Lull 1967, pp. 89-94
Natural hardwoods Taiwan 971 85.9 1.7 12.4 Sopper and Lull 1967, pp. 89-94
Natural hardwoods Banco Forest, lvory Coast 86.9 Bernhard Reversat et al. 1972
Natural hardwoods Yapo Forest, lvory Coast 77 Bernhard Reversat et al. 1972
Natural hardwoods Garamba Forest, Zaire 74 Noirfalise (1956)b
Natural hardwoods Puerto Rico 18 12.2 Kline et al. (1968)b
Bamboo forest Kenya 18.2 Pereira (1952)b

aCited by Gupta 1980, p. 84, table 2
bCited by Lawson et al. 1981, pp. 143-144

TABLE 4. Storm size and interception ratios

a) Stemflow and interception in a tropical forest in Tanzania

Average stemflow for storm classes
Storm size (mm) 0-5.0 5.1-10 10.1-15 15.1-20 20.1-30 30.1-40 40.1-50
Stemflow (mm) 0 0 0.1 0.2 0.4 0.5 0.9
Gross rainfall and interception
Gross rainfall (mm) 1 2.5 5.0 7.5 10.0 15.0 20.0 30.0 40.0
Interception loss (mm) 0.7 0.9 1.2 1.5 1.8 2.4 3.0 4.2 5.4
Interception (%) 70 36 24 20 18 16 15 14 13.5

Source: Jackson 1971

b) Stemflow and interception in a temperate forest

Stemflow for storm classes (as a percent of rainfall)
Species Age (years) Rainfall size (mm)
    1-2 5-10 10-15 15-20 20-25 25-30 30-60
Oak 25 min. 0.24 1.18 2.08 5.6 10.9
max. 0.32 4.9 6.5 8.5 15.3
Aspen 50   0.8 3.2 4.8 6.4 8.1 9.4
Birch 65   0.7 2.8 3.9 5.6 7.3 8.5


Interception of rainfall by the canopy of temperate forests 220 - 225 years old (percentage of precipitation on forest glades)
Forest type Interception (%)
  Storm size (mm)
  0-1 2-5 5-10 10-15 15-20 25-30
Oak with sedge-goutweed 37 23 20 18 15 11
Oak with sedge-goutweed-linden 47 31 26 23 20 12
Oak with spindle tree 25 13 11 10 7 4
Oak in solonets 19 11 9 7 4.8 2
Glade precipitation (mm) 0.67 3.6 7.5 12.7 17.3 26.1

Source: Molchanov 1963

The tropical forest, and indeed any forest cover, produces litter that protects the soil beneath from rainfall impact and filters out the fine particles that may clog the larger pores. In addition the forest furnishes food and protection to insects and animals that burrow in the soil and increase soil permeability. Infiltration rates are therefore usually high under forest cover where the forest floor layer is well developed. Where it is disturbed by logging or removed by fire, protection may be decreased sufficiently to lead to overland flow. It thus appears that the forest floor and not the canopy is the part of the forest that directly regulates infiltration. It can also be argued, however, that the microclimate (high humidity, light wind, maintaining of low moisture fluxes) prevents the forest soils from drying (especially during the West African harmattan). The soil does not even harden, so that its permeability is maintained, preserving its infiltration capacity. Finally, the presence of a certain amount of humus in the top soil assures a soil structure favourable to infiltration. In Yangambi in Zaire, under an annual rainfall of 1,850 mm d'Hoore (1961) reports that yearly decomposition of organic matter in the litter produced by a secondary forest reaches 50%. The coefficient of decomposition of humus (the proportion of the total mass of litter that is decomposed annually) reaches 68-76% as against 6-12% under beech forest in California.

In the Indian black soil region at Bellary (Tejwani, Gupta, and Mathur 1975) the infiltration rate was lowest under agriculture, at 10 mm h-1; under woodland it was 112 mm h-1. At the Ootacamund Shola forest it was 112 mm h-1 under broom (Cytisus scoparius), 125-168 mm h-1 under Shola forest, and 207 mm h-1 in a blue gum (Eucalyptus globulus) plantation. In Bihar, Mistry and Chatterji (1965) recorded infiltration rates under forest of 260 mm h-1. The corresponding values for grass and crop lands are 120 and 90 mm h-1. The infiltration rate reported for an undisturbed primary tropical forest on yellowish red sandy clay loams to clays and on slopes of 5 to 15% in East Kalimantan Province of Indonesia is 2,772 mm h-1. (Kartawinata et al. 1981).

Infiltration and overland flow are two sides of the same coin, since what does not infiltrate the soil runs off the surface as overland flow. There is no doubt that infiltration rates are influenced by diverse factors that include pedological and slope conditions and the effects of forest cover. The resulting spatial variations together with the temporal variations of infiltration make comparison of results obtained in different forested tropical basins difficult. Klinge et al. (1981) found that high vertical permeability of soils in one of the three watersheds studied in the Amazon Basin, in French Guiana, resulted in high infiltration rates and hence low overland flow. In contrast, the other two catchments in the same area had low subsoil permeabilities, which resulted in low infiltration rates and transformed 60 to 70% of the incident rainfall into runoff. Cailleux (1959) also concluded that in primary tropical forest of French Guiana overland flow is insignificant. Measurements on experimental plots at Adiopodioume, near Abidjan in Ivory Coast (Dabin 1957), also reveal very little overland flow: from I to 3% under forest, with a maximum of 7.8% during a downpour of 193 mm. The terrain in this case is composed of Tertiary sands, but the slope is of only a few degrees.

Overland flow under forest is essentially discontinuous and the water infiltrates the soil every few metres. This makes its measurement more complex and interpretation more difficult. For instance, it has been observed that under the tropical forest overland flow often originates in streamlets dropping from trees; it frequently moves the litter, accumulating it against obstacles downhill; and the water spreads in different directions. This discontinuous overland flow in the forest of southern Ivory Coast was more marked than that of the forest of the state of Bahia in Brazil. He noted that in Ivory Coast litter is never abundant and the water runs underneath it. In contrast, in Bahia the litter is thick and the top soil is composed of a spongy humic horizon, both of which retard overland flow. The explanation seems to lie in the manner of decomposition of the organic matter, but the roles of the type of litter, microclimate, and organisms are not fully understood. It is known, however, that the interference by man has been more important in West Africa, and shifting cultivation could have been responsible for a certain amount of soil degradation that diminishes the soil permeability and facilitates overland flow.

Bonell, Gilmour, and Cassells (1983) found that widespread overland flow is common in the undisturbed rain forest during the summer monsoon in north-east Queensland, in Australia. They explain this by reference to the relationship between rainfall intensity and soil hydraulic properties. They find the surface soils (0-0.1 m) to be so highly permeable (the mean coefficient of soil permeability, K= 20.13 m d-1) that they can absorb even peak monsoonal intensities. However, the subsoil has low permeability (mean K of the 0.20-1.0 m layer is 0.02 m d-1), so that the moisture capacity of the 0.15-0.20 m layer is frequently exceeded during prolonged rainfall events. This results in early saturation of the top layer, which in turn leads to "saturated" overland flow as defined by Kirby and Chorley (1967) and rapid subsurface flow. Not surprisingly, therefore, some 46% of the total annual streamflow appears as quickflow (overland flow plus interflow), while frequently more than 45% of the rainfall from individual storms appears as quickflow (Gilmour, Cassells, and Bonell 1982).

Hitherto forest lands have generally been regarded by hydrologists as areas that have optimum infiltration and negligible overland flow. Available evidence from Australian, West African, and Amazonian tropical forests should make hydrologists pause and think again. There seems to be no doubt, however, that fire, trampling, and compacting (by humans, cattle, or vehicles) from logging disturb the natural forest floor and usually cause marked changes in normal infiltration. And often such disturbances initiate overland flow. On the other hand it is sometimes argued that if the forest floor is maintained in an undisturbed state when the forest cover is removed, infiltration remains unchanged and overland flow does not occur (Gilmour, Cassells, and Bonell 1982).

Water yield characteristics

Water yield is defined as the discharge of a stream at a particular cross section as calculated over a specified period of time: a day, a month, a season, or a year; or over a number of such intervals. The average yield is a measure of the volume of runoff, the quantity of water available for utilization and development (Benham 1974). The yield is often used to describe the regime of flow of the river empirically and stochastically. It is usually measured in a well-defined watershed to ensure that no significant leakage occurs.

Colman (1953) argues, however, that yield should not be considered in terms of the total amount of runoff alone but should include rate of flow and its temporal variations as well as water quality (incorporating sediment load). He submits that all are important aspects of water yield, and control over all of them is required if proper management of water yield is to be achieved. In this report, however, only streamflow aspects of water yield are included. The total water yield has several components that should be considered separately in order to detect the effect of forests. These include:

  1. total discharge for hours, days, months, seasons, years;
  2. overland flow and interflow (= direct runoff) for similar periods;
  3. baseflow discharge for similar periods;
  4. stormflow discharge for individual storms;
  5. discharge duration and retardation.

Contents - Previous - Next