Contents - Previous - Next

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

Methods of detecting the effects of forests on water yield

In order to detect the effects of forest on streamflow characteristics three operations are usually necessary: calibration of the watershed against the control, treatment of the watershed, and analysis of the resulting data to detect any changes the treatment may have on the streamflow or yield. Much of our knowledge of forest watershed hydrology is derived from experiments.

Calibration Methods

Calibration is the determination of the normal (pre-treatment) relationship between a streamflow characteristic of interest and some other variable or variables. Several calibration methods have been used to characterize hydrological relationships. A limited number of studies have been conducted using only single watersheds. In this case a watershed is calibrated on climatic data. During the calibration period the flow characteristic under study is related to climatic variables. According to Golding (1980) this method is more informative than the paired (control) basin method because it relates streamflow to factors that influence it. It also costs less than instrumenting two basins and avoids the problem of searching for two or more similar basins. Clearly, no advantage can be taken of possible contrasts between different types of cover on otherwise similar watersheds, which may facilitate analysis of the results (Ward 971, 49).

There is also the paired watershed approach, in which two comparable watersheds are selected and the hydrological variables of interest measured and uncontrolled in fluences indexed during the calibration period, which is usually a minimum of five years. One of the watersheds is chosen as a control while the other is treated. The paired watershed approach is most common and was at first assumed to be the most suitable for solving the watershed problems investigated in South Africa (Wicht 1967). Some of the best known paired watershed experiments in humid tropical regions are those of East Africa, of which two are in Kenya and one in Tanzania. Many years may be necessary to attain the level of precision required, and the length of this calibration period is difficult to determine (Wilm 1949; Kovner and Evans 1954; Singh 1974) in spite of the several prescriptions made for its determination. The period should not only be long enough to yield sufficient values for computing the significant statistics, it should also be representative of the regional climate. To permit accurate isolation of treatment effects the climatic conditions during the treatment period should be similar to those prevailing during the calibration period. It has been found in some cases that such stability cannot be guaranteed. This has led to a search for rapid calibration methods. Bethlahmy (1963) advocates the calibration of watersheds by comparing their responses to individual storm events. It is suggested that in humid regions the annual number of adequate storms will exceed 100, so that a maximum of three years will be required for calibration.

The use of multiple watershed experiments partially eliminates the deficiencies of the paired watershed approach. In this approach streamflow, for instance, from a number of treated watersheds may be compared to that from a number of control watersheds. This may eliminate the need for a pre-treatment calibration period. Here again there is need to determine the number of watersheds required to detect a specified increase in water yield at a given probability level (Swanson and Hillman 1977). There are examples of this approach in the humid tropics; some of the earlier ones include those reported by Rodier (1959) and Wicht (1961, 1967) for the tropical and subtropical regions in Africa. The multiple watershed approach requires a greater input of financial and human resources over the short period, but the problem of unstable control treatment due to natural or man-made changes in the physical environment is much reduced.

Types of Treatment

Several types of treatment have been, and are being, applied to experimental watersheds. Clear-cutting of a large part, or even the entire experimental watershed, is one of the common treatment methods. There are also cases of strip cutting in which clear-cut strips alternate with uncut strips of similar dimension (Ward 1971). Examples of clearcutting experiments in tropical rain forest include Lien-Hua-Chi Watershed in Central Taiwan; Mbeya Watershed in Tanzania; IITA experiments in Ibadan, Nigeria; the Doon Valley experiments in India; and the multiple watershed experiments in French Guiana.

Selective logging or thinning, based on either age, species, or quality, and sometimes selective patch or block cutting are also adopted as the method of treatment of an experimental watershed. The Man and the Biosphere experiment in East Kalimantan

Province, Indonesia, is a good example of treatment by selective logging. In some more recent experiments the role of chemical treatments and burning in forest clearance has also been investigated.

Substitution or changing forest cover is a rather common type of treatment in humid tropical regions, where the choice is often between tea, pine, and bamboo species. Experimental watersheds in which the natural forest cover is changed to tea or pines are found in Kericho and Kimakia, both in Kenya. There is also the Dehra Dun experiment in the Doon Valley, India, where the forest was clear-cut and reforested with Eucalyptus grandis and E. camaldulensis (Mathur et al. 1976; Bahadur, Chadra, and Gupta 1980).

The removal of forest litter is another method of treatment that may help to separate the effects of tree canopy from those of the forest floor, particularly pertinent to tropical forests. Unfortunately no example of this in the humid tropics is known to the writer. However, Tsukamoto (1975) has demonstrated the effect of forest litter in the 1.42 ha North Watershed in Aichi Forest of the University of Tokyo.

Finally, afforestation has been applied as the main treatment in a number of experimental watersheds in the tropical regions. Effects of afforestation on watersheds require longer periods of observation since tree growth is relatively slow. In South Africa watersheds under sclerophyllous scrub were afforested with pines. The treatment is similar to that of substituting one forest cover for another in the East African watersheds, except that forest development may be faster in the latter.

Examples of properly designed watershed experiments, which have as their main objective the evaluation of the effect of forests on streamflow characteristics, are few in the humid tropical regions, even though they abound in the temperate zones. There are, nevertheless, a number of watershed investigations that use forested watersheds that are being progressively deforested to examine the effects of such forest removal. These studies include those by Raghunath, Das, and Thomas (1970), who analysed data from 17 watersheds whose forest cover ranged from 2.47% to 69.33% in the Nilgiris, India. Others include the Kali Mondo Basin, Java, in Indonesia (Bruijnzeel 1983), 10 watershed investigations in French Guiana of the Amazon region (Roche 1981), and the montane cloud forests area of Costa Rica (Zadroga 1981).

According to Reinhart (1967) an ideal treatment would take full effect immediately upon application and remain in effect, at the same level, indefinitely. This situation is approximated by a clear-cutting of forests followed by frequent re-cuts. If the treatment departs from this ideal situation, analysis becomes more difficult and failure of the experiment is possible. Afforestation and reforestation treatments are exceptions to Reinhart's maxim, but time-trend analyses can be used to evaluate the changes, if any, in the streamflow characteristics.

Effects of tropical forests on water yield

Hibbert (1967) summarized 39 studies of clearing forest (all the results available by that time). He shows that the upper limit of yield increase is 4.5 mm y^-1 for each percentage of reduction in forest cover. As may be expected most treatments produced less than 2.5 mm increase per year. First-year responses to complete deforestation varied from 34 mm to 457 mm of increased runoff and most produced less than 300 mm. The plot of the first-year increases for thirty treatments does not, according to Hibbert (1967), reveal any strong relation between increased water yield and percentage of reduction of forest. It is, however, interesting that the highest first-year increase in water yield reported was for a 100% clear-cutting of a watershed at Kimakia (Kenya) under tropical forest cover. Hibbert drew three generalizations:

1. reduction of forest cover increases water yield;
2. establishment of forest cover on sparsely vegetated land decreases water yield; and
3. response to treatment is highly variable and, for the most part, unpredictable.

A total of 94 forest clearing or planting experiments, including those already reviewed by Hibbert (1967), were studied by Bosch and Hewlett (1982). They also made generalizations and criticisms that may lead to more careful design of catchment experiments. Incidentally, two treatments reported for Maimai, New Zealand, produced the highest first-year yield increases, of 650 and 540 mm. Bosch and Hewlett (1982) dispute Hibbert's conclusion that water yield responses to afforestation and deforestation is unpredictable and conclude that coniferous forest, deciduous hardwood, brush and grass cover have, in that order, a decreasing influence on water yield of the source areas compared to bare ground.

The following specific inferences are also made, though the authors cannot set error limits on the coefficients whose sign and magnitude they consider to be correct.

1. Conifers and Eucalyptus cover types cause about 40 mm change in annual water yield per 10% change in the area of forest cover.
2. Deciduous hardwoods produce some 25 mm change in yield per 10% change in cover.
3. Brush and grass lands produce about 10 mm change in yield per 10°/o change in cover.

Results from Forested Tropical Catchment Experiments

Table 5 presents brief descriptions of a number of paired catchment experiments under tropical and subtropical forest conditions. The table gives details of treatments and effects on water yield during the years following treatment. Experiments 21 and 22 are not strictly controlled, while number 12 is a promising multiple catchment experiment yet to yield results. The following brief description of the selected experiments provides additional information on the treatment effects.

East Africa. Two of the experiments are in Kenya. One deals with the replacement of indigenous rain forest by plantation tea at Kericho and the other by exotic conifers, patula pine, at Kimakia. A third experiment is at Mbeya in southern Tanzania, where a pair of otherwise similar catchments, one under indigenous forest cover and the other cultivated without soil conservation, are being compared for streamflow and sediment yield. In all three experiments the water balance equation was used to evaluate, over a series of water years, the unknown components, particularly the actual water use of the paired catchments. Other conceptual models were then used to simulate streamflow responses.

TABLE 5a. Tropical and subtropical watershed experiments: Basin characteristics

^aSee Appendix: ^bhighland;^cciting Van Wyk 1977;^d citing Bailly et al. 1974

TABLE 5b. Tropical and subtropical watershed experiments: Treatment and water balance

At Kericho (Sambret Watershed) there was overall reduction in water use (about 11%) and some increases in water yield during the initial clearing and clean-weeding. The replacement of tropical forest by tea estates has, however, not resulted in long-term reduction in water yield. Differences revealed in interception (lower from tea) and transpiration (higher from tea) might alter the seasonal distribution of streamflow, which for some water uses may be as important as changing the annual yield.

At Kimakia replacement of bamboo forest by patula pine and vegetables initially decreased the water use by 19% and resulted in a large increase in streamflow. Once the pine canopy closed (1967-1973) no significant difference in water yield could be detected as water use by both bamboo and pine was 76% of the Penman evaporation from water surface (E`,).

At Mbeya, when evergreen forest gave way to shamba or small holder cultivation, a large increase in water yield resulted. Very little of this increase resulted from overland flow because of the remarkably high infiltration capacity of the very stable and porous ash-derived soils. On the other hand, the dry-season baseflow was doubled. The difference in dry-season transpiration accounted for much of the change in water yield, for water use by plants decreased from 0.92 E₀, under forest to 0.65 E₀ for cultivated land. Also, under forest only 28% of annual precipitation resulted in streamflow, while the percentage was 40% for the cultivated watershed during 1958-1968.

Taiwan. A small low-altitude watershed, Lien-Hua-Chi, near Sun Moon Lake in central Taiwan, originally under subtropical montane hardwood forest, was clear-cut during 1978-1979. The surface disturbance was kept to a minimum as skyline logging was used; roads were constructed around the basin periphery, away from streams, and all yarding was uphill. The yield increases for the first and second years after clear-cutting were 58% and 51% respectively of the annual flow. The effect was greater on the low flows, which increased by 108% and 293% respectively during those years, while the wet-season flow increased by 55% and 47% respectively.

Ibadan, Nigeria. The effect of clearing a 44 ha watershed at Ibadan, western Nigeria, in early 1979 was a significant increase in total water yield. Prior to the clearing streamflow was negligible. Following clearing runoff increased by some 340 mm, or 23% of the annual rainfall. Both direct runoff and baseflow components increased after deforestation (fig. 2). The baseflow increased steadily from "an unmeasurable trace" in the forested catchment in 1978 to about 0.1, 0.12, and 3.22 mm per month in 1979, 1980, and 1981 respectively. Gradual and persistent deterioration of the surface soil structure is said to account for the increase in overland flow. The reduced water use was an important factor of the increase in total water yield.

Experiments 6 to 11 compare methods of land clearing and tillage systems. These six experiments are calibrated against a 15 ha forested catchment that experienced only an insignificant amount of runoff during 1979-1981. Traditional farming is based on incomplete clearing. Among basin treatments involving complete clearing (i.e. 2-2 to 2-6), manually cleared plots are associated with an average of 16 mm per year increase of yield in 1979-1981. The corresponding figure for mechanically cleared plots is 67 mm per year.

FIG. 2. Water yield from a 44-ha cleared catchment at Ibadan, Nigeria (table 5, item 5) for three consecutive years after deforestation. Note increase in the baseflow from November to March for 1980 and 1981 compared with 1979. (Source: Lal 1983)

The results of these experiments indicate the need to investigate further the effect of clearing and tillage methods, as much of the differences in experimental results might be explained by the techniques of deforestation employed.

Dehra Dun, India. A pair of small forested (brush) watersheds were calibrated for 8 years (1961-1968). W₁F was then clear-cut and re-afforested with Eucalyptus species. The post-calibration relationship indicates that reforestation resulted in 28% reduction in streamflow. The resulting plantation was fully stocked, with dense undergrowth. Information was not available on the effect of clearing on water yield during the first year after clearing (1969) due to failure of the flow-recording gauge.

North Creek, Babinda, Queensland, Australia. Two catchments were selected in the tropical rain forest in Babinda, North Queensland, in 1969 to study the effects of logging and land clearing. The experimental catchment, North Creek, was calibrated against the South Creek control watershed from 1969 to 1971. It was then logged in June 1971, and in 1973 67% of the experimental catchment was completely cleared, stick-raked, and partially ploughed. The treated areas of the catchment remained completely bare for two years.

Logging produced little detectable hydrological change, but the two years following clearing produced an average increase in annual water yield of 10.2%, or 293 mm. Furthermore, weekly minimum discharge in the treated catchment increased by 40 to 60% for flows less than 5 litres s^-1 after clearing. Soil moisture levels in the experimental basin remained higher because of reduced transpirational demand, and soil moisture deficits were drastically reduced. For instance, at the end of the first dry season after clearing in 1973, the treated North Creek catchment required only 94 mm of rain to bring the soil moisture in the surface 3 m to field capacity, whereas the control catchment required 291 mm. Equations relating soil moisture in the two catchments before and after clearing indicate that the moisture level rose by 33% in North Creek after clearing.

South Africa. The replication in time over a set of six watersheds shown in table 5 for South Africa has brought the streamflow comparison method to its highest degree of development (Pereira 1973). The experiments were begun at Jonkershoek Research Station in 1940 and at the Cathedral Peak Research Station in 1951 to study management policies for the steep grass or scrub covered watersheds. Plantings were done at 8-year intervals on a 40-year forest rotation cycle. This made possible comparisons of the effects of pine and Eucalyptus at different stages of tree growth.

A simple model derived from the Cathedral Peak results by Nänni (1970) is used to predict the effect of afforestation on average annual water yield. The model consists of a set of curves relating mean annual streamflow reductions, due to afforestation, to mean annual runoff before afforestation (fig. 3). The model has limited use as it neither provides information on seasonal patterns of water yield nor on changes in the components of streamflow. Maximum flow reductions in the South African experiments following afforestation conform to expected first-year increases after clear felling (table 5).

First-Year or Maximum Yield Changes

Hibbert (1967) appeared narrowly to resist the temptation to pronounce 450 mm as the upper limit of the first-year increase following clear-cutting. He noted, however, that "exceptional climatic conditions must prevail if larger increases are to be obtained." The maximum value of 457 mm was produced in one of the two tropical forest experiments included in his review. However, the results in table 5 and the plot of 12 of the results (fig. 4) indicate that the first-year maximum yield changes following clearcutting or afforestation may be up to 6 mm for each percentage of change in forest cover. There is obviously some relationship between streamflow response to afforestation or deforestation and the regional mean annual rainfall in tropical forest environments. Unfortunately, the availability of results presently for the region is inadequate to depict any such relationship. Nevertheless, if any region is to provide maximal streamflow responses, that region appears to be the forested areas of the humid tropics, where high rainfalls coupled with large radiation energy surpluses react virtually continuously throughout the year with deep-rooted evergreen trees on deeply weathered soils.

FIG. 3. The Nänni curves used to predict the effect of afforestation on average annual water yield in Natal. (Source: Bosch 1982)

It would appear, however, that the effect of clear-cutting in high rainfall areas is shorter lived than in low rainfall areas due to rapid regrowth of vegetation in the former. The maximum yield reductions following afforestation are comparable to the first-year increases after clear-cutting. There is no evidence that such reductions are lower; they may indeed be higher than the observed increases following clear-cutting in the tropical forest (fig. 4).

FIG. 4. Maximum yield change after/during treatment versus changes in vegetation cover. (Numbers refer to items in table 5.)

Average Annual Changes in Total Yield

Figure 5 shows the plot of average annual changes in yield against the change in forest cover. A "best-fit" line (not shown) to the plot indicates that there is about 5 mm change in annual water yield for each percentage of change in forest cover. Average values obtained for afforestation over longer periods are more reliable in this respect, provided the effect of climatic trends has been removed. The plot also suggests that any change in forest cover of less than 15% may result in no change in streamflow.

Seasonal Distribution of Yield Changes

It is important to consider the behaviour of the stream during different seasons of the year. It is possible for the annual total yield to remain unchanged while the stream regime itself is so altered that the use of the streamflow is greatly impaired or enhanced. The role of forests in this case is similar to that of engineering structures on streamflow.

FIG. 5. Average yield changes following changes in vegetation cover. (Numbers refer to items in table 5.)

At Mbeya, East Africa, and North Creek (Babinda), Australia, there was no detectible change in overland flow, whereas the dry-season baseflow increased significantly. Where there is evidence of change in overland flow, as in Ibadan (Nigeria) and Taiwan, the increase in baseflow is much higher, even though the actual volume flow is relatively small.

The drastically reduced transpirational demand, especially during the dry season, from clear-felled areas is primarily responsible for the large increase in low flow of streams. Maintenance of a good soil structure and high infiltration after treatment further enhances the contribution of baseflow to the total yield.

Conclusion

Wide variations in experimental methods have often made pooling of results difficult.

In the present review of tropical forest experiments the methods adopted by investigators include the control watershed approach, in which the yield (R_A) of the treated watershed is predicted by the yield (R_C) of the control watershed by the relation

where a and b are intercept and slope constants respectively.

In the East African experiments with the paired catchments under the same land use, no initial calibration period was possible. But through the use of better methods of estimating evaporation and improved methods of measuring the major components of the hydrological cycle, it was possible to produce comparative estimates of the water use of different types of vegetation. This, however, means that any differences in seasonal streamflow distribution between the paired catchments could not be attributed, by pre- and post-treatment comparison, to a particular cause, such as land use change. Intensive and long-term methods (with calibration and treatment) should be regarded as complementary rather than as alternative methods in the tropical forest environments.

One of the major weaknesses of many traditional studies that should be avoided is the measurement of only one or two of the components of the water balance. All the components of the hydrological cycle ought to be measured or accurately estimated, if necessary, by employing watershed model techniques. Only as empirical coefficients are made more deterministic and bulked processes of the hydrological cycle further separated can hydrological models be improved. This is perhaps why Pereira (1973) announced "the era of calibrate, cut and publish, is over and a far more detailed study of the components of the hydrological cycle is necessary."

In evaluating the effects of land use on water yield, the use of the "best" hydrological year is desirable. The best hydrological year is associated with the highest coefficient of determination when runoff is related to rainfall by a simple correlation model (Sharp et al. 1960). They also found that the variable hydrological year did not prove to be as good as the "best" hydrological year of uniform length and dates.

There is certainly an urgent need to undertake properly designed catchment experiments in tropical forests in order to understand and control the impact of forest harvesting on water quantity in these active hydrological environments. The development of catchment studies should enable the developing nations that inhabit most of the region to apply the knowledge gained to land management, particularly in water conservation and erosion control. It is encouraging that a number of such studies recently established in parts of Brazil, Guyana, Argentina, India, and Malaysia are expected to produce results that will considerably increase our present knowledge of the hydrology of the enigmatic tropical forest.

Contents - Previous - Next