TERRY E. EVANS
Former Director, Mott MacDonald Co., UK
Methods of evaluation
General circulation models
Hydrological models
Recent changes in global precipitation
Recent changes in global runoff
Conclusions
References
The development of large water resource systems can take 10 to 20 years from the initiation of preliminary studies to realization of projects, and therefore an assessment of the impact of global warming on water resources is urgently required. The aim of this chapter is to present evidence of hydrological changes which have already occurred, and to evaluate whether, at the present time, our knowledge is sufficient to make predictions of changes in the hydrologic cycle with enough accuracy to be of value in water resource planning.
Methods for evaluating future available water are discussed and examples given of impact studies using outputs from General Circulation Models (GCMs). In this respect, Beran (1986) has emphasized a distinction between hydrology and available water resources. The former deals with evaluating the total resource (naturalized river flow) whilst the latter requires the quantification of the exploitable amount.
The world's agricultural output is heavily dependent on irrigation. Out of a global cropped area of 1 500 million ha, 16% is irrigated. Global warming is likely to have a major impact on the hydrological cycle and consequently on irrigated agriculture. This impact may be more significant than the direct effect of higher temperatures. Even though this may be the case, it is essential to view the impact of climate change in context with other major global changes. Demands on water are multi-sectorial and irrigation is already in competition for potable water, industrial, power generation, and recreational and environmental uses: most of which command a higher price. Within many Near East and African countries current water resources are almost fully exploited and supplies will have to double over the next 20 to 30 years to maintain the current, even unsatisfactory status quo. When the problem of increasing levels of water pollution is added into the equation, a situation arises of 'Water in Crises' (Gleick, 1993).
Unfortunately, in most cases hydrology is treated on a national, rather than a regional, continental or global scale. It is heavily dependent on accurate data collection. With national water resource agencies generally small units within major ministries, and having low priority, funding has fallen to such an extent that in the majority of African countries reliable data are no longer collected (World Bank/UNDP/ADB/EC/French Government, 1993). The seriousness of the problem has yet to be addressed; although the World Bank is promoting a new 'Integrated Approach to Water Resource Management for Sub-Saharan Africa' and other regions of the world, whilst WMO is attempting to set-up a 'World Hydrological Cycle Observing System (WHYCOS)' (Rodda et al., 1993) and FAO is establishing a programme to develop a worldwide water database with hydrological modelling capability using Geographic Information System (GIS) and remote sensing technology.
In this chapter some of the problems associated with climate change and the hydrological cycle are discussed.
Different approaches can be used to estimate the effect on water resources of global warming of which the main ones are:
· use of instrumental records;
· palaeoclimatic analogues;
· GCM outputs.
In the first approach the natural variability of climate is used to predict rainfall patterns during sequences of extremes of dry or wet years. Difficulties arise due to the short period of records, although stochastic data generation techniques can be used to extend the database. Over such limited time spans, however, longer-term changes in vegetation, sea surface temperatures, ocean currents, etc., which would be expected to take place with a doubling of atmospheric forcing functions such as CO2 are not adequately accounted for. Irrespective of this limitation, significant long-term changes in surface runoff have occurred over the past 30 years in a number of major river basins, especially in Africa. Recently, strong scientific evidence has also been produced to suggest that such changes are directly related to global warming triggered by the increase in anthropogenic greenhouse gases (Thomson, 1995; Hadley Centre, 1995). As some of the recorded trends in river flows are of the same order of magnitude as would be expected for a 2 x CO2 scenario, a review of recent changes in the hydrological cycle adds to an understanding of the problems associated with long-term changes in gross water resources (section Recent changes in global runoff).
Palaeoclimatic analogues fall into the same category as instrumental records though with obvious drawbacks: estimating past climates is extremely unreliable and fragmentary. Their main benefit lies in being able to cover the range of temperatures likely to be experienced over the next century or two (Table 2.1).
Recent studies suggest that global temperatures during previous ice ages fluctuated widely over time periods as short as decades or centuries (Heinrich cycles) and temperatures were not relatively constant as generally assumed (Maslin, 1993).
Table 2.1. Past global temperatures
|
Mean global temperature above present-day values |
Date |
|
+ 1.5°C |
6 000 BP |
|
+2.5°C |
125 000 BP |
|
+ 4.0°C |
4 000 000 BP |
There is also no consensus amongst scientists over the prime cause for past temperature changes in the earth's atmosphere, or for the triggering of the ice ages; it certainly was not anthropogenic. The general accepted viewpoint is that changes in solar radiation received at the earth's surface are the prime instigators (Milankovitch, 1930). However, with current research spanning many disciplines, there has been a prolifigation of new theories related to the controlling factors responsible for producing ice ages and postglacial optimums. Current theories range from rates at which the earth's crust is formed, extent of coverage of peat bogs, release of methane from shallow seas due to formation of sea ice, rate of mountain building and resulting erosion and sediments deposited in the oceans, and cyclic changes in iceberg production affecting the north-south ocean 'conveyor belt', possibly due to instability from trapped thermal heat below glaciers, etc. Until the casual factors involved in past climate change are known, it would be unwise to assume that they would have no influence in determining the resulting climate; an inherent assumption in the climate analogue method. Present land use, vegetation, ice cover, atmospheric composition, etc. are all likely to be so different from those experienced during previous postglacial optimums as to make a direct comparison between past and present climates of speculative value only.
GCMs are the only credible tool for predicting climate change and for providing inputs to hydrological models. Unfortunately, at the present time it is not possible to use precipitation outputs from GCMs directly as inputs to hydrological models. The outputs are not sufficiently accurate to simulate daily and monthly sequences (see Table 2.2).
However, the use of GCMs is the only way forward and with improved understanding and better modelling of the systems (especially ocean circulation and the land phase of the hydrological cycle) and larger computer memories, it is only a matter of time before these deficiencies will be remedied.
Table 2.2. Comparisons between observed rainfall over England and estimates from three GCMs (Arnell et al., 1990)
|
Source |
Dec-Feb |
|
Observed |
2.5 mm/day |
|
UK Meteorological Office GCM (Bracknell) |
1.2 mm/day |
|
National Center for Atmospheric Research GCM (Boulder, Colorado) |
3.4 - 4.5 mm/day |
|
Goddard Institute for Space Studies GCM (New York) |
3.5 - 4.5 mm/day |
A GCM model typically has four components: atmospheric, land, ocean and sea ice. As the four models are interactive, production runs must include all four model components. Initially, the atmospheric model was developed to a relatively high level of sophistication, whilst the land phase and the ocean components were very simplistic.
Runoff enters into current GCM model simulations at two points (Rowntree, 1989). Firstly, at the boundary between the atmosphere and the land surface, where flux transfers are converted into surface runoff, and secondly at the boundary between the ocean and land, where inflow hydrographs are required as input to the ocean model. At present neither are simulated with sufficient accuracy. Kite et al. (1994) have demonstrated some of the problems associated with GCM modelling of the hydro-logical cycle, particularly the lack of adequate lateral transfer of water. The problem of inadequate routing of surface runoff is highlighted by studies of Semtner (1984, 1987) where a reduction in freshwater inflow to the Arctic Ocean would have a significant effect on salinity gradient and it was calculated that the outflow from the Kara and Barent seas could halve should there be a total diversion of rivers flowing to these seas. The oceans are therefore quite sensitive to freshwater inputs.
It has been demonstrated that a high degree of 'recycling' of rainfall occurs over land and if, for instance, land evaporation is much reduced or eliminated, then a significant reduction in precipitation results (Shukla and Minz, 1982; Hall and Sarenije, 1993). In order to determine actual evapotranspiration an accurate assessment of surface runoff is required. Current GCMs do not achieve this. It is likely, therefore, that shortcomings in the modelling of the hydrological cycle are reducing the accuracy of outputs from current GCMs, as has been highlighted by Rind et al. (1990), The GISS GCM, and others, have been shown to greatly underestimate actual evapotranspiration (ET) through overestimating the reduction in transpiration with increased soil moisture deficits. For example over the USA the ratio of Actual ET/Potential ET for current climatic conditions was estimated by GCM to be 0.23, whereas the actual ratios should lie between 0.6 and 0.9. Such a large discrepancy arises in part due to vegetation cover being inadequately represented in the land phase in GCMs. It was concluded by Rind (1995) that ET loss has been seriously underestimated by GCMs and that global precipitation increases of 9 to 15% (Grotch, 1989) are not compatible with a 30% increase in atmospheric water holding capacity corresponding to a 4°C warming. Either the rainfall estimates are underestimated or much more serious droughts will occur over continents.
Initially, climate change was seen, understandably, as a meteorological problem. The importance of the atmospheric-ocean interface was always known, but only now are results from coupled models becoming available. Similar improvements are required in the hydrological cycle components. The Global Energy and Water Cycle Experiment (GEWEX) of the World Climate Research Programme of ICSU/WMO/IOC should provide the impetus for such improvements. However, until the land phase is adequately modelled by GCMs then there is little hope that its outputs can be used with confidence in water resource planning. Furthermore, river basin impact studies undertaken to date have been based on results obtained from equilibrium GCMs in which the oceans are modelled simply as a fixed layer(s). Because of this major limitation, they are only used to simulate equilibrium states such as 1 x CO2 or 2 x CO2 neither states of which are, or will be, in equilibrium. All the GCMs are structurally similar and contain the same basic algorithms and will inherently include the same errors. It is rather disturbing, therefore, that such large differences in output values are produced (see Table 2.2 and Figure 2.1). Equilibrium models are relatively stable and do not suffer to the same extent from cold start and drift problems that afflict transient models. Because of the shortcomings of the equilibrium models the reliability of the outputs, particularly precipitation, is seriously compromised. The range in outputs from the different equilibrium models is generally so large that it is often not possible to determine the sign of the change in runoff over large regions or even over some continents. Because of this problem some researchers have used outputs from a number of GCMs and developed levels of probability for decreases (or increases) in temperature and precipitation (Howell and Allan, 1994). Although the results from such a risk analysis approach present a useful guide for water resource planning, when all the other uncertainties are included, such studies are of little practical value in the detailed planning of future water resource development, except to make the planner aware of the increased uncertainty due to climate change.
A significant improvement in recent years has been the development of transient, (ocean-coupled) GCMs. However, the computer time utilized and the costs in running these models are at present exorbitant. As an interim measure techniques have been developed which allow for the rescaling of equilibrium model results and integrating them with those of individual transient model runs to provide an indication of temporal changes (Viner and Hulme, 1993). Figure 2.1 shows the range of estimated precipitation changes using the outputs from seven GCMs scenarios (Table 2.3) over the period 1990 to 2050. The low and high rainfall changes are based on the lower and upper 90% confidence limits. Although this approach enables the results from equilibrium models to be modified to provide outputs for different time horizons, the serious drawbacks which apply to the basic equilibrium model results remain.
The first transient model experiment (UKTR) undertaken by the Hadley Centre (Hadley Centre, 1992) determined the climate's response for an increase in CO2 concentration of 1% per annum over a 75 year period. The northern hemisphere was found to warm twice as rapidly as the southern hemisphere. The rise in temperature of the Antarctic was of a similar magnitude to that for the southern hemisphere as a whole, i.e., very different from previous equilibrium GCM results. In addition, the temperature rise in the transient model experiment was only 60% of that for the equilibrium model, due to the large inertia of the oceans. The change in global rainfall was, however, similar to that produced by early UK Meteorological Office GCMs. Further experiments (Hadley Centre, 1995) have attempted to model historical changes in climate from 1860 to 1990 incorporating for the first time the attenuating effects of sulphate aerosol emissions. The results are encouraging, and the anomalous reversals in global temperature rise in the early 1940s, and the rapid warming from the mid-1970s, are reproduced by the model. Evidence is mounting that the recorded changes in climate over this century are consistent with those expected from increasing emissions of greenhouse gases.
Unfortunately, transient coupled GCMs are not yet capable of reproducing dynamic regional sea surface temperature patterns, and the associated precipitation changes, to the accuracy required to simulate the trends in regional precipitation which have occurred this century. To do this it would be necessary to understand the mechanisms which trigger the Southern Oscillation and to be able to simulate both them and those of the north Atlantic ocean conveyor system, and other significant ocean circulations. It is not possible, therefore, to confirm at the present time, using outputs from transient GCMs, that recorded global precipitation changes are the direct results of global warming. Until such calibrations are effected, then precipitation outputs from GCMs, transient and equilibrium, will remain speculative and should not be used directly in detailed planning of new water development projects.
TYPES
The models under discussion here are concerned with the land phase of the hydrological cycle, and generate river flow and groundwater recharge. The most commonly used is the deterministic model. It is a 'conceptual' type model used in catchment hydrology to simulate at various levels of detail the physical processes involved in moisture flux transfer: from interception, evapotranspiration, soil moisture changes, surface runoff, infiltration, percolation, interflow, groundwater storage and channel routing. They tend to be lumped processes and are the workhorse of applied catchment hydrology. An essential prerequisite in applying these models is a high level of calibration based on recorded data.
Table 2.3. Some characteristics of the seven GCM experiments used (Hulme, 1994)
|
Model acronym |
Resolution |
Ocean heat transport |
D Teq |
D Peq (%) |
W (%) |
Reference |
|
|
Longitude |
Latitude |
||||||
|
1 GISS |
10.0° |
7.8° |
Prescribed |
4.2 |
11 |
5 |
Hansen et al. (1984) |
|
2 GFDL |
7.5° |
4.5° |
None |
4.0 |
9 |
10 |
Wetherald and Manabe (1986) |
|
3 LLNL |
5.0° |
4.0° |
2 layer ocean |
3.8 |
11 |
5 |
W.L. Gates (pers. comm., LLNL) |
|
4 ECHAMI-LSG |
5.6° |
5.6° |
11 layer ocean |
1.6 |
3 |
10 |
Cubasch et al. (1992a) |
|
5 OSU |
5.0° |
4.0° |
None |
2.8 |
8 |
20 |
Schlesinger and Zhao (1989) |
|
6 UKLO |
7.5° |
5.0° |
Prescribed |
5.2 |
15 |
15 |
Wilson and Mitchell (1987) |
|
7 UKHI |
3.75° |
2.5° |
Prescribed |
3.5 |
11 |
35 |
Mitchell et al. (1990) |
Note: D Teq and D Peq are sensitivity of models for a 2 x CO2 scenario (except 4). W = weight to derive precipitation model-average, based on global pattern of correlation coefficients.GISS: Goddard Institute for Space Studies
GFDL: Geophysical Fluid Dynamics Laboratory
LLNL: Lawrence Livermore National Laboratory
ECHAM-LSG: European Centre Hamburg Model - Large Scale Geostrophic
OSU: Oregon State University
UKLO: United Kingdom Meteorological Office - Low
UKHI: United Kingdom Meteorological Office - High
Much research in pure hydrology is aimed at improving the individual processes and obtaining better spatial representation of input data. This, eventually, should reduce the importance of calibration. However, there are pitfalls in using over-sophisticated representation of the hydrological cycle in one or two processes at the expense of over-simplification in others, especially without evaluation and verification on actual catchments (Lockwood, 1985; Schnell, 1984). Similarly, statistical and black box models, often regression based, have little use in impact studies, where many parameters affecting runoff are non-stationary, e.g., evapotranspiration, vegetation, land use, groundwater abstractions, etc.
Simple water balance models (Wigley and Jones, 1985), or models incorporating evapotranspiration based solely on temperature, are also of little value either in water resources impact studies or in generalized estimates of regional changes. There is no short cut to evaluating regional change in water resources without involving all the components of the hydrological cycle at a basin or subbasin scale. Time scales in hydrology are extremely important. For example, in arid and semi-arid regions the annual runoff is dependent on rainfall intensities at an hourly time scale falling over relatively small areas and on percolation within the river bed. Times to peak for catchments of areas of thousands of square kilometres may be only one or two hours. In such catchments the intensity is as crucial to surface runoff as the rainfall amount. Similarly, if annual potential evapotranspiration were to increase by 25% over a catchment with few rainfall days in the year, it would have little impact on the volume of runoff generated. Simple annual and monthly water balance models will work satisfactorily in such conditions. In catchments with significant soil storage, the runoff is governed by available storage and rainfall intensity. In the simple balance equation:
Q = P-E
where Q = surface runoff, P = precipitation, E = catchment losses.
Q is not independent of E, and if treated as such, will produce significant distortion when calculating changes in Q relative to changing of P and E (Wigley and Jones, 1985).
Also, in most catchments both geology and vegetation cover can be extremely variable. These parameters alone produce large differences in seasonal and annual runoff. For example in the River Canje catchment in Guyana, half the catchment is covered by tropical forest with a closed canopy and underlain by pervious 'White Sands'; whilst the remainder is largely covered by savanna and relatively impermeable sediments. As a result, runoff ratios (Q/P) in the wet season are about 0.15 in the forested area, and 0.85 in the savanna; in the dry season the values are 0.50 and 0.10, respectively (MacDonald, 1965).
INTERFACING GCMS AND HYDROLOGICAL MODELS
There are many ways of using outputs from GCMs to provide inputs into hydrological models. The most obvious is to take the GCM simulations of daily rainfall, temperature and other meteorological variables and feed them directly into a calibrated hydrological model. Unfortunately precipitation outputs from different equilibrium models are extremely variable (see Figure 2.1 and Table 2.2). Also, because of the way clouds and precipitation are modelled, GCMs produce unrealistic simulations of day-to-day rainfall. An alternative approach is to use the 'change' in rainfall, temperature and other climate variables either to produce directly a perturbed time series or to produce stochastically generated sequences of wet or dry states with amounts based on conditional probability distributions. This latter method is a compromise to cover for the inadequacies of the GCM outputs and to produce more realistic looking data sequences. Similarly, regional GCM-generated pressure patterns can be used with a suitable regional multivariate model to generate sequences of rainfall and evapotranspiration (Mott MacDonald, 1992, 1993). However, the pressure patterns cannot be used directly to produce climatic sequences. None of the methods used gives confidence that results will be other than approximate and of limited value in water resource planning.
IMPACT STUDIES OF CLIMATE CHANGE
To date, most of the impact studies undertaken have not used hydrological models which adequately simulate the transient nature of the parameters of the hydrological cycle. Even fewer have attempted to include changes in water demand.
Many impact studies have found that changes in evapotranspiration rates due to global warming have only a marginal effect on surface runoff when compared with changes in precipitation. These conclusions have been questioned. Rind et al. (1990) have suggested that most GCMs fail to simulate the land phase satisfactorily and underestimate actual evapotranspiration rates. The effect of increased potential evapotranspiration may significantly reduce surface runoff in the tropics (Rind, 1995). In addition, often changes in temperature alone have been used to estimate changes in evapotranspiration. This again will underestimate actual evapotranspiration (Arnell and Reynard, 1993). Finally, many studies have used simplified hydrological models, often of the water balance type, to analyse changes in surface runoff (Wigley and Jones, 1985). Such methods can easily be demonstrated to be inappropriate.
The Institute of Hydrology (IH) has undertaken an impact study for 21 catchments in England and Wales (Arnell and Reynard, 1993). The wettest scenario produces a general increase in average runoff across England and Wales, whilst the driest produces a reduction. However, the most interesting findings relate to changes in evapotranspiration. Three different potential evapotranspiration (PET) scenarios were investigated: PET changes based on; Case 1 - increased temperatures only; Case 2 - changes in temperature, net radiation, relative humidity and windspeed; Case 3 - Case 2 plus changes in plant stomatal conductance and leaf area. The results are given in Table 2.4.
Table 2.4. Equilibrium climate change scenarios for 2050 (Arnell and Reynard, 1993)
|
Climate parameter |
Annual changes in England and Wales |
|
Temperature |
+2.2% |
|
Rainfall |
+4% |
|
PET 1 |
+10% |
|
PET 2 |
+35% |
|
PET 3 |
+ 24% |
An important feature of the results is the quite dramatic increase in PET predicted in England and Wales resulting from the inclusion of climatic elements other than temperature. A change in humidity produced the most significant increase based on PET calculations using the Penman-Monteith formula.
Kite (1993) has recognized the importance of land cover changes in water resource impact studies. The Kootenay catchment in Canada was divided into 50 Grouped Response Units (GRUs) and 10 land cover divisions, giving a potential for subdivision to 500 units. The land cover data were obtained from satellite data. Good calibration was obtained using a distributed hydrological model (SLURP). In the 2 x CO2 scenario with changed land cover, the frequency of high flows was found to more than double and evapotranspiration was found, surprisingly, to decrease by 10%.
Over the last century the hydrological cycle over most river basins has been subjected to extensive human changes with the construction of reservoirs, land-use changes, river abstractions, groundwater abstractions, inter-basin diversions, etc. Such changes significantly alter river flows and the hydrology of the basin. For example, the Nile loses 10% of its flow through reservoir evaporation above the Aswan Dam whilst the Zambezi loses the equivalent of almost 20% of its flow at Victoria Falls from evaporation losses from the Kariba reservoir. On a global scale it has been estimated that sea levels would have risen this century by 2 mm per year, rather than the recorded 1 mm per year, had it not been for increases in reservoir storage and river abstractions, mainly for irrigation purposes (Newman and Fairbridge, 1986). It is important, therefore, that existing modifications to natural river flows are evaluated in impact studies as well as likely future changes in demand and supply.
It is evident that forecasts of future water availability produced by climate impact studies to date have limited value in present-day detailed water development designs. Their main value lies in improving impact modelling techniques. Certainly, too much effort has been placed on routine impact studies at the expense of targeted research. Estimates of the effect of climate change on precipitation and evapotranspiration are not sufficiently accurate to provide credible hydrological predictions.
RECORDED RAINFALL, 1931-1960 AND 1961-1990
Hulme et al. (1992) have prepared maps (see Figures 2.2 and 2.3) showing global precipitation as well as annual and seasonal changes in precipitation between the periods 1931 to 1960 and 1961 to 1990. The two most prominent features are over northern Russia with increased precipitation of up to 20% and the African Sahel, between latitudes 10°N and 30°N, with reduced precipitation between 20 and 50%. Significant changes have occurred over all continents. These recorded changes over the last 60 years are of a similar magnitude to those predicted by GCMs for a 2 x CO2 scenario (see Figure 2.4). How far these recorded changes are related to radiative forcing due to increased greenhouse gases is controversial. The standard response is to state that the natural variability in climate will mask any forcing for two or more decades. This viewpoint can now be questioned following the results of several recent research studies, most notably Thomson (1995). In a major study of annual solar variations at the earth's surface, it has been demonstrated that, due to approximations in using the tropical year (equinox to equinox) in the Gregorian calendar rather than the anomalistic year (perihelion to perihelion), the starting dates of the seasons should be occurring later each year this century. A reversal occurred in 1923 which cannot be explained by changes in incoming solar radiation or by other phenomena such as the occurrence of volcanic eruptions or the Southern Oscillation, but is completely consistent with recorded increases in greenhouse gas emissions. Moreover, were the recorded increases in global temperatures due to greater solar radiation activity, then the amplitudes in seasonal temperature would be widening. The reverse is the case, which is consistent with the greenhouse effect. In addition, UK Meteorological Office GCM (Hadley Centre, February 1995) modelling of historical changes of climate (1960-1990) has reproduced both the anomalous reversals in global temperature and the rapid increases recorded later in the century. The evidence is mounting, therefore, that recorded changes in climate this century are the direct result of increased greenhouse gas emissions. If this is the case, then changes in river flow regime over this century probably present the best evidence available for water resource planning of the transient effect of global warming.
Although there are no transient GCM results to make comparisons between the observed precipitation change recorded between 1931 to 1960 and 1961 to 1990 (Figure 2.3), an interesting comparison can be made between present-day conditions and those predicted in 50 years' time (Figure 2.4). On a global scale the same pattern of change is modelled. The significant exception is that of Africa, north of the equator, which is shown to experience increased rainfall. The Sahelian drought may therefore be ameliorated to some degree. This is consistent with a predicted reversal in hemispheric warming with the north warming much more rapidly than the south; the opposite situation to that seen over the past 30 years. Aridity is predicted to increase in Australia, a trend which has been apparent in recent years, although clearly associated with the Southern Oscillation, a feature not represented at a decadal scale in model outputs. The other continents are shown to continue the trend of the past 30 years. However, considerable reservation must be placed on such GCM predictions until the changing rainfall/runoff patterns experienced over the past 100 years have been simulated with a measure of accuracy. This will require a much improved understanding of ocean circulation.
Figure 2.2. Approximate global precipitation
FACTORS AFFECTING AFRICAN HYDROLOGY
The Hurst effect
Hurst (1965) showed that annual Nile floods did not follow the classical theory of a stationary stochastic process. The estimated storage to maximize yield of the Nile was much higher than classical theory indicated, e.g., for a 500-year period Hurst calculated an accumulated departure of flow from the River Nile of 48 standard deviations from the mean compared with the theoretical 28. This problem was a dominant feature of hydrological research in the 1970s. Klemês (1974), in a comprehensive review of research into the Hurst phenomenon, warned of the dangers of deluding ourselves that statistical models such as fractional Brownian noise with its implicit acceptance of infinite memory could be used to explain the phenomenon. A highly successful operational model may turn out to be totally unacceptable from a physical point of view - the Ptolemaic planetary model for example. Klemês also demonstrated that the Hurst phenomenon could be produced using a zero memory process and non-stationary means. Klemês even suggested that operational models of the future may revert to extrapolations based on sound short-term knowledge rather than long-term synthetic data based on 'ignorance and guesses'.
Since the 1970s, interest has waned in trying to explain mathematically why so many long-term natural series fit distributions which are not Gaussian. As the modelling of physical processes has developed, the mechanisms which can lead to persistence of high and low rainfall have become better understood.
Physical feedback processes
The climate of Africa is rather unique. In comparison with the other continents, Africa south of the Atlas Mountains has no major continuous mountain barriers, either longitudinally or latitudinally orientated, to disrupt the circulation. Consequently, its climates and its seasonal variations are less complex and governed to a large extent by the oscillating inter-tropical front which follows the sun on its annual passage between the two hemispheres.
Nicholson (cited in Tyson, 1987) has clearly demonstrated, in a reconstruction of sequences of historical droughts and wet periods of Africa, that the African climate bridges the two hemispheres. The extreme drought years of 1972 and 1984 which devastated the Sahel region from the Atlantic coast to Ethiopia are clearly reflected in the flows of the Rivers Nile, Chari, Zaire, Niger, Zambezi and Orange in these years. Although a high degree of homogeneity does exist between climatic fluctuations in the north and south of the continent, the influences of such large-scale systems as the Southern Oscillation and the Atlantic Oscillation, both in part driven by sea temperature variations, and the monsoons of Asia, do produce spatial anomalies in climate extremes in different areas of Africa.
There is a tendency for an equatorial band across African to react in an opposite direction in terms of precipitation to the rest of Africa. This has been demonstrated clearly by the recent high levels recorded in Lake Victoria, Lake Tanganyika and Lake Malawi. Perhaps the clearest example of this anomaly is that the higher White Nile flows, originating in equatorial Africa, have persisted from 1961, whilst in contrast the Blue Nile has suffered the most serious reductions in discharge recorded this century, or over the past 1 300 years if the Roda Nilometer records are accepted as a reliable source (Howell and Allan, 1994). It seems that for extreme wet and dry years the whole continent can be affected with only minor deviations. Nicholson (1986) isolated four different preferential climatic patterns over the African continent. These modes are:
· northern and southern subtropical areas dry; equatorial regions wet;
· northern and southern subtropical areas wet; equatorial regions dry;
· whole continent dry;
· whole continent wet.
Dry continent and dry subtropical area modes have dominated the continent since 1970.
From a water resource planning viewpoint the major concern is whether or not the drought, which started in the late 1960s in the Sahel, and became dominant in southern Africa throughout the 1980s, will continue. One viewpoint, which has been strongly supported in the past in southern Africa, is the existence of cyclic behaviour (Tyson, 1987). Such fluctuations, and particularly the persistence in climate whereby one dry year seems to be followed by another and one wet year by another wet year, can be explained by biogeophysical short-term feedback processes, the so called 'Joseph' and 'Noah' effects: reduced vegetation leads to increased albedos and increased radiation losses, surface cooling and greater atmospheric stability which reduces rainfall and encourages persistence (Charney, 1975). Similar persistences are produced by lowering soil moisture levels in GCM simulations, indicating the importance of rainfall itself in initiating a significant feedback process, and by the increase in atmospheric particles from sandstorms. Such feedback processes have always been around and explain the Hurst effect and almost certainly explain much of the natural persistence of sequences of wet years and dry years found in the Sahel region, although fluctuations in sea surface temperature also play a role.
Water resource engineers in the past have relied on statistical data, whether stochastically generated or historic, for their estimation of reliability of yield. Long-term trends are generally ignored unless physical explanations can be presented to justify their inclusion. In this respect the Sahelian drought has been with us for almost 30 years and suggests that it has been maintained by causal factors other than normal feedback process. Global warming is an obvious explanation.
Global causes
The search for a causal explanation of the African drought received an impetus by research undertaken in the UK Meteorological Office (Folland et al., 1986), which demonstrated a strong correlation between sea surface temperature (sst) anomalies on a global scale with wet and dry periods in the Sahel. After the mid-1960s, a marked cooling of the oceans of the northern hemisphere and simultaneous warming in the southern hemisphere was observed. A reversal occurred around 1970, since when temperatures in both hemispheres have increased. A time series plot of sst differences between oceans of the southern hemisphere and those of the northern hemisphere, and rainfall anomalies for the Sahel, shows a strong negative correlation. The correlation between the July-September sst and Sahel rainfall for the period 1901 to 1984 was -0.62, which is significant at the 99.9% probability level (see Figure 2.5). Numerical equilibrium GCM experiments with prescribed sea temperatures were also undertaken by the UK Meteorological Office which were able to replicate rainfall reductions in the Sahel for recent drought years: 30% reductions over western Sahel, 20% over eastern Sahel and up to 50% over the mountains of southern Sudan and northeast Ethiopia.
The observed and unexpected warming of the southern oceans at a faster rate was thought to be due to a reduction in the heat transfer from southern to northern hemispheres, although the detailed mechanisms of the transfer are still the subject of much research. Alternative scenarios include increased deep water circulation from 1960 to 1970 in the Atlantic and the effect of sulphate aerosols which are dominant the northern hemisphere. If, however, the reduction in heat transfer is related to the north-south conveyor system combined with a slowdown in formation of north Atlantic deep water at high latitudes, owing to a reduction in the extent of sea ice (Street-Perrot and Perrot, 1990), then the Sahelian drought may persist until the greater land mass in the northern hemisphere starts to dominate the effects of a slowdown in ocean transfer and the attenuation effect of sulphate aerosols. However, confirmation will depend on further research developments into detailed coupled transient GCM models, which can be calibrated against recent climates and sea temperatures. It is likely that the results from the equilibrium GCMs will be found wanting once more reliable coupled models have been developed.
A plausible scenario of future global warming links a weakening in carbon sinks and radiation sinks in the polar regions with reduced deep water formation due to reduced heat transfers from the southern hemisphere to the north. It is postulated that reductions to the radiational and CO2 sinks could give rise to significant positive feedbacks leading to an increase in global warming (Lewis, 1989).
Figure 2.5. Sahel rainfall departures and sea surface temperature differences between hemispheres (after Folland et at., 1986).
(a) rainfall departures Sahel;
(b) sea surface temperature (sst) differences (southern hemisphere, including Indian Ocean, minus northern hemisphere)
LAKE CHAD
Lake Chad is situated on the border of the Sahara between latitudes 12°N and 14°30'N and longitudes 13°E and 15°30'E. The lake has no outlet to the sea and is the main focus of an internal drainage basin of 2 500 000 km2 that collects water from Algeria, Chad, Niger, Nigeria, Cameroon, Sudan and the Central African Republic. Over 90% of the total lake inflow comes from the Chari-Logone river system which rises on the southern margin of the basin and has a combined catchment of about 570 000 km2. Lake Chad is located centrally within the Sahel, which is often defined as the 250 to 500 mm rainfall band across Africa below the Sahara desert. Lake Chad is shallow; average depths vary between 1.5 and 5 m, with the result that its surface area is very sensitive to changes in inflows. The lake has a number of unusual features. In spite of being an enclosed basin, it remains a freshwater lake. Its size ranges from 100 km2 for the 'Gran Chad' to 10 to 30 km2 for the 'Petit Chad'. The water tends to be replaced every year or every other year, with an average annual inflow of about 30 to 40 km3. The saline water in normal times is pushed to the northwest of the lake where its salinity is concentrated and natural seepage to the north maintains the salt balance in equilibrium.
Feasibility studies began in 1972 to investigate the irrigation of lands bordering Lake Chad in Nigeria. As a result, the South Chad Irrigation Project (SCIP) was designed for three-stage implementation. Two stages have been built: Stage I in 1979 and Stage II in 1983. A pump station with a total capacity of 103 m3/s (current capacity 75 m3/s) was constructed to pump water from the lake via a canal. The full scheme has a gross area of 66 000 ha and would abstract approximately 3% of the annual inflow to Lake Chad. The layout of the scheme was based on 4 ha units to provide 16 000 households, or a 100 000 population, with a more secure livelihood. It was planned to produce two crops per year based on rice (50%), cotton (20%) and wheat (60%), with a cropping intensity of 130%.
In the mid-1960s Lake Chad had reached record levels for this century giving rise to speculation that the 'Gran Chad' as observed by explorers of the last century would be re-established. However, this failed to happen and lake levels fell during the 1970s and at the time that the feasibility studies were completed were similar to those experienced during the first half of the century. No one anticipated the drastic fall in rainfall which has occurred since (Figure 2.6).
The sudden onset of the Sahelian drought in the 1970s resulted in a severe contraction in the size of the lake, changing its character to a swampy delta at the mouth of the River Chari, with dire consequences for lake-side communities and the SCIP irrigation scheme.
Lake Chad divided into northern and southern pools in 1973 and by 1976 levels had fallen below the bed of the intake canal. Although water levels recovered during October and November, providing irrigation water for wheat crops planted in November, shortage of water in June to September prevented the cultivation of the main rice crop. The two pools have remained separated and since 1983 levels have continued to fall and have prevented even winter-season irrigation. The scheme is maintained in good condition awaiting the recovery of the lake.
Figure 2.6. Measured and reconstructed Lake Chad levels (1870-1992)
Table 2.5. River Chari annual runoff (km3) at N'Djamena (1971-1992) (mean flow before 1971 - 40 km3/annum)
|
Year |
Flow |
Year |
Flow |
Year |
Flow |
|
1971 |
31 |
1978 |
29 |
1986 |
15 |
|
1972 |
18 |
1979 |
21 |
1987 |
10 |
|
1973 |
18 |
1980 |
25 |
1988 |
27 |
|
1974 |
29 |
1981 |
19 |
1989 |
16 |
|
1975 |
35 |
1982 |
21 |
1990 |
12 |
|
1976 |
29 |
1983 |
17 |
1991 |
19 |
|
1977 |
25 |
1984 |
7.5 |
1992 |
21 |
|
|
|
1985 |
17 |
Mean annual flow |
22 |
Figure 2.7. Departures from mean annual flow, River Chari at N'Djamena
The reason for the fall in lake levels is simply that flows in the River Chari-Lagone basin failed. The total runoff in 1984 was only around 20% of the long-term mean. Since 1971 flows have been reduced to 50% of the previous long-term mean annual flows of over 40 km3, although rainfall is estimated to have fallen by only 25%. Of all the major basins in the world, probably Lake Chad has been affected most by climate change. The most ambitious irrigation scheme built in Africa since the mid-1940s has yet to be operated in earnest (Table 2.5 and Figure 2.7)
Falling Lake Chad levels demonstrate the effect of a rainfall reduction of 25% on river flow in a semi-arid region where rainfall gradients are steep and small shifts in the atmospheric circulation can cause large rainfall changes. As a result, proposals have been put forward for a major inter-basin transfer from the River Zaire to the River Chari basin. With an annual runoff of 1 250 km3, potential supplies are available.
RIVER NILE
To a large extent Egypt has remained unscathed by the drought owing to the large over-year storage in Lake Nasser. However, other factors have contributed to reducing the drought's impact. One of these is the exceptionally high levels in Lake Victoria which have helped maintain higher White Nile flows. The higher levels have resulted from very heavy rainfall in Kenya and Uganda between 1961 and 1963 and above average rainfall since. The higher White Nile flows have helped to compensate for the lower discharges in the Blue Nile, which have been more adversely affected by the Sahelian drought.
The construction of the High Aswan Dam in 1963 provided the complete regulation of the annual Nile flood and also gave sufficient over-year storage to supply Egypt's share under the Nile Waters Agreement (1959) of 55.5 km3, with a very high level of reliability (96%).
Figure 2.8. Lake Nasser live storage (1968-1992)
Following the failure of Nile flows in 1984 (a naturalized inflow to Lake Nasser of 59 km3 was recorded, compared with a mean flow of 84 km3), there was considerable concern that Lake Nasser's resources would fall to a level where irrigation supplies would have to be reduced with both agricultural and political implications (Figure 2.8).
Effect of Sahel drought on Nile flows
A main feature of the Nile flow series, apart from the high flow period at the end of the last century, is the steep fall in discharge from the mid-1960s. It is more pronounced and persistent than any previous low flow periods. An analysis of Roda Nilometer data also indicates that during the last two centuries the variability of the annual flood far exceeds that observed since records began in AD 622 (Howell and Allan, 1994).
Most previous yield studies of the River Nile have analysed River Nile flows at Aswan. However, the flow records over the past two decades have demonstrated the importance of the different rainfall regimes over the catchments of the Blue and White Nile. Their influence on White and Blue Nile flows can clearly be seen by comparing Figures 2.9 and 2.10. Flows in the White Nile between 1962 and 1985 have increased by 32% or 8 km3 above the 1912 to 1961 mean. This has occurred at a time when Blue Nile flows have decreased by 9 km3 over the period 1965 to 1986, or 16% below their 1912 to 1964 mean. It could be inferred that the two rainfall regimes are negatively correlated. However, past records show that the relationship between annual inflow to Lake Victoria and recorded Blue Nile flows at Khartoum is random (MacDonald, 1988). This, however, contradicts results of GCM outputs which suggest that rainfall over the White and Blue Nile catchments may be negatively correlated (Howell and Allan, 1994).
Lake Nasser simulations
River Nile flow sequences were used to investigate the ability of the Nile systems to withstand drought (MacDonald, 1988) should the Sahelian drought prove to be a persistent feature of climate of the region in the coming decades. Two sequences were prepared for severe and moderate droughts (based on flow statistics from 1968 to 1988) and the results are shown in Table 2.6. They indicate that Lake Nasser releases would have to be reduced below 55.5 km3 (Egypt's allocation under the 1959 Nile Waters Agreement) to 52 km3 and 49 km3, respectively, and that these yields would have reliabilities of 87% and 88%, i.e., water shortages would occur on average in 13 and 12 years respectively every 100 years. The effect of the Sahelian drought on River Nile flows and irrigation in Egypt, although problematical, would be nowhere near as severe as for Lake Chad. Even under the most severe drought conditions the yield from Lake Nasser is predicted to fall only from 55.5 km3 design yield to 48.4 km3, or less than 15%, although the reliability of supply would fall from a design yield of 96 to 88% (Table 2.6)
Figure 2.9. Departures from mean annual White Nile flows at Mogren (1912-1989)
Figure 2.10. Departures from mean Blue Nile flows at Khartoum (1911-1989)
Table 2.6. Summary of Lake Nasser reservoir simulations
|
Sequence type |
Total number of years |
Release from Lake Nasser (km3) |
Number of years with reductions |
Reliability yield(%) |
Minimum annual release (km3) |
Average yield obtained (km3) |
||
|
1st lowest |
2nd lowest |
3rd lowest |
||||||
|
Moderate drought sequence |
400 |
52 |
52 |
87 |
39.6 |
40.1 |
41.3 |
51.2 |
|
Severe drought sequence |
100 |
49 |
12 |
88 |
38.9 |
39.8 |
40.4 |
48.4 |
Options do remain, however, for increasing supplies by reducing drainage outflow and operational losses to the sea, but these will largely be needed to meet increased municipal and industrial demands. Conway et al. (1996), in a recent study on future water availability in the Nile basin, utilized GCM outputs and modelled runoff incorporating land cover changes. The precipitation changes ranged from +18 mm to +39 mm and Nile flows were calculated to range from a 3% decrease to a 10% increase. These changes are much smaller that those recorded over the last two decades.
The Nile basin is a closed system and in irrigation terms is very efficient (>70%). Improvements in efficiently are therefore limited. Because it is a closed system increased groundwater abstractions do not represent extra water in the long term and provide only short-term temporary benefits. There is potential to save water by selecting less water demanding crops and the use of short-term varieties (5% savings). However, significant enhanced supplies could be obtained from the Sudd and other wetlands in Sudan but these all have major political or ecological constraints. Ignoring the political and ecological implications there is still scope for further water development within the Nile basin. However, when these are set against the population explosion in the region and possible problems of climatic change, the outlook cannot be viewed with optimism.
CASPIAN SEA LEVELS
Background
Another region which has experienced a major change in water resources is in Russia. The River Volga, which rises near the Baltic Sea and discharges into the Caspian Sea, is the largest river in Europe, 3 530 km long. Its catchment area of 1.5 million km2 makes it one of the major rivers in the world. Its river basin is larger than that of the Zambezi, the Indus or the Ganges. Around 60% of its discharge is derived from melting snow. Three main rivers, the Volga, Ural and Terek, account for 88% of the total river inflow to the Caspian Sea and, of this, over 50% arises from the Volga river. It is connected by canals to Moscow, Volgograd, St. Petersburg and the Arctic Ocean for river traffic and, with hydropower stations and extensive irrigation systems around the north Caspian Sea coast and the Azov Sea, the Caspian basin is crucial to the economy of Russia, Azerbaijan, Kazakhstan and the Ukraine.
The Caspian Sea has been called the greatest salt lake in the world, although its salt content is only one-third of that of the main oceans. It has no natural outlet. It covers an area of 372 000 km2 with its surface at a level of -28.5 m below mean sea level, whilst the total basin area extends to some 3.7 million km2. An extensive sedimentary plain surrounds the outlet of the River Volga where the sea is shallow with depths of only 4-6 m. The total capacity of the sea is 66 960 km3, compared with an annual inflow of around 310 km3. The annual volume of inflow would raise levels by 0.8 m if evaporation is ignored. Net evaporation (evaporation - rainfall) is therefore close to this figure. Being an enclosed sea, levels fluctuate with inflow and historically have shown a wide variation. Also, developments within the Caspian Sea basin have seriously reduced River Volga flows. Levels fell dramatically from 1930 reaching a low in 1977, although surprisingly since this date levels have started to rise and plans are in hand to mitigate flood damage. However, without predictions of future sea levels and a knowledge of the processes causing the fluctuations, attempts to mitigate the damages resulting from too high or too low levels are likely to be both ineffective and uneconomic.
Water development
The River Volga plays a vital role in the economic life of the states through which it passes. It is used for:
· navigation (151 000 km of waterways connecting the Arctic Ocean to the Caspian Sea, with the Moscow canal, the Volga-Don canal and the Volga-Baltic waterway all with a 4.5 m navigation depth);· power generation (a cascade of 11 hydroelectric schemes with a capacity of 11000 MW);
· irrigation (in the Caspian-Azov basins some 7.5 million ha of irrigation have been developed; 4 million ha using sprinkler systems).
Relative water consumption over the years has been compiled and forecasts made for the future based on the construction of the Volga-Don and Volga-Chograi irrigation canals (Table 2.7) (Berezner, 1987).
For withdrawals of 37 km3 and a residual flow of 266 km3/year the Caspian Sea was in equilibrium at a level around -28.5 m amsl. This is considered to be about the optimum level for the valuable fisheries. Over the last millennium the highest level was thought to be -22 m recorded in the 1600s, and the lowest levels, -31 m in the 1400s, and -29.0 in 1977, a range of some 9 m (Figure 2.11).
Because of the extensive and dramatic water developments undertaken by the former USSR in the Volga basin since the 1930s, the first step in any evaluation is to naturalize the flow to remove the influences of human activities (Vali-Khodjeni, 1991). But for the massive water developments, the present-day levels of 27.7 m amsl would now have returned to the 1929 high of -26.2 m amsl (Figure 2.12). The larger abstractions from the Volga river are therefore contributing significantly to reducing sea levels, albeit fortuitously, and so preventing more serious damage to irrigation areas and urban development adjacent to the Caspian Sea (most of which have been developed in recent years). There has been therefore a dramatic reversal from the problems of falling sea levels to one of rising levels and how to contain increasing flood waters.
Table 2.7. Abstractions from Volga/Caspian Sea, post-1930 (km3/yr) (after Berezner, 1987)
|
Water consumption |
1980 |
1990 |
2000 |
2010 |
|
Water supply |
5.6 |
7.3 |
9 |
10.5 |
|
Irrigation |
18.6 |
23 |
33 |
36 |
|
Reservoir evaporation |
6 |
7 |
8 |
9 |
|
Filling of reservoirs |
3 |
2 |
0 |
0 |
|
Inflow to Kara-Bogaz-Gol |
0 |
2 |
2 |
2 |
|
Water consumption in Iran |
3.5 |
4 |
5 |
6 |
|
TOTAL |
37 |
44 |
57 |
63 |
Figure 2.11. Variations in levels in the Caspian Sea (1839-1987) (after Vali-Khodjeni, 1991)
Figure 2.12. Recorded and naturalized levels in the Caspian Sea (1930-1990) (after Vali-Khodjeni, 1991)
Between 1931-1960 and 1961-1990 rainfall on the Volga catchment has increased by over 10% whilst naturalized inflow to the Caspian Sea has increased by a similar amount.
The Intergovernmental Panel on Climate Change (IPCC, 1990) suggests increases in precipitation of between 100 and 200 mm over the winter period for a doubling in greenhouse gases. If this prediction proves correct then it would produce a very significant increase in surface runoff in the River Volga which is predominantly snow-melt fed. The additional precipitation will fall mainly as snow on frozen ground. There will therefore be little increase in losses and most of the additional snow will be converted to surface runoff and will increase River Volga discharges substantially. River Volga flows could increase by between 25 and 50%.
If the present change in climate with increased precipitation over Russia continues over the coming decades, serious problems in terms of drainage and flooding of low-lying irrigation lands are likely to be experienced. However, in the longer term the additional water would be extremely beneficial for agricultural development in the Caspian Sea sedimentary basin where 100 000 km2 of land lies below sea level. Alternatively, surplus water could be used to replenish all of the damage inflicted on the Aral Sea: an annual flow of around 50 km2 would be required to restore the Aral Sea to its pre-1960 condition. A surplus of 100 km3/year of freshwater would certainly find customers in the Near East, although the task of transferring it would be formidable.
1. The effect of climate change on available water resources in most regions, while significant, will be small compared with demands generated by population growth, industrialization, urbanization, land-use changes and improved standards of living. In many countries resources are already fully committed and water will become a scarce commodity.
2. Although agriculture is the most prolific user of the world's water resources, it is more vulnerable to water shortage because of the higher priority given to potable water supply and other high value users.
3. GCMs represent the only plausible method for predicting the effect of global warming on the hydrological cycle. Unfortunately, existing GCMs are not capable of producing realistic precipitation outputs which are required in water resource planning. This is due in part not only to the way clouds and precipitation are modelled, and to coarse resolution, but also to the inadequate representation of the land phase and ocean systems in existing equilibrium GCMs and in the new transient GCMs. For example, lateral water transfer and artificial influences are not taken into account. Because of the interactive relationships between the atmosphere, land and oceans, such simplifications are likely to affect outputs from current GCMs. As a result, even the sign of the change in runoff over regions and some continents cannot be predicted with any confidence.
4. Until transient GCMs can simulate sea surface temperature patterns and regional precipitation changes, observed this century (and the changes are significant), future precipitation changes cannot be predicted with confidence.
5. Predictions from the large number of impact studies undertaken to date to evaluate the effect of climate change on the hydrological cycle can largely be discounted. Their value at present lies in improving impact study techniques rather than assisting in water development planning. However, much of the research has been directed at trying to convert inadequate outputs from GCMs to more realistic values for use in the impact studies rather than trying to target specific problem areas.
6. Deterministic models, which adequately represent the hydrological cycle and are calibrated, should be used in impact studies. Black box, simple water balance and stochastic models have a limited role to play where catchment characteristics are changing rapidly with time.
7. All catchment models used in impact studies should be adequately calibrated. Research into the development of remote sensing algorithms for satellite data and its interfacing with GIS systems, together with better methods of applying small-scale physically based techniques at larger scales, will eventually reduce the present dependence on calibration. The calibration requirement for river basin studies applies equally to the land phase of GCMs.
8. Under conditions of rising temperature, vegetation cover of basins will change significantly and hydrological models used in impact studies should allow for this.
9. Most impact studies have found that surface runoff is influenced more by changes in precipitation than by changes in evapotranspiration. This has arisen in part due to ignoring parameters other than temperature, such as windspeed, net radiation and humidity, in calculating evapotranspiration; realistic values of these are difficult to obtain from GCM outputs. Recent studies indicate that increased evapotranspiration may be more significant than previously thought, although reduced stomatal conductance and reduced leaf area will partially limit its effect.
10. Because the relationships between rainfall and runoff are non-linear, the response of runoff generated by rainfall is magnified. In addition, the rainfall over a large basin is integrated into a single runoff value. As such, flow data represent valuable means of monitoring climate change.
11. Predicted GCM global precipitation increases for 2 x CO2 scenarios range from 8 to 15%, whilst the water holding capacity of the atmosphere for a global 4°C rise will increase by 30%. Under transient conditions the land will warm more rapidly than the oceans, guaranteeing that ocean evaporation (global precipitation) will not keep pace with potential evapotranspiration over the continents. This situation is likely to persist for a century or more. In broad terms then, runoff would appear likely to decrease due to greater aridity and increased frequency of droughts over the next century, providing precipitation forecasts are correct.
12. Recent research supports the claim that changes in climate recorded over the present century are due to increases in greenhouse gases in the atmosphere. If this is the case, then valuable information should be obtained from recorded changes in flow regime in major river basins. It is essential that transient GCMs are capable of simulating these changes.
13. Case studies for the River Nile, Lake Chad and Caspian Sea demonstrate how hydrology is highly susceptible to climate change.
14. Data collection of hydrometric data is essential for operating water resource systems. It is also essential for planning water development. It is unfortunate that in many developing countries data collection has almost been abandoned. Water resource assessment agencies are invariably small units within large sectorial ministries and perceived to be of little value. It is essential that this situation is remedied as a matter of urgency.
15. Hydrology should be treated on a regional, continental and global scale, in the same way, not nationally in isolation. As with climatic parameters, river flows are related on a regional, continental and global scale. All measures which will promote this change should be supported by governments and international agencies.
16. Increased storage in water-critical regions is desirable for agricultural development and sustainability. Worldwide climate change may exacerbate the differences between water-rich and water-poor areas. Water transfer will therefore assume increased urgency and importance.
17. Special water resources design criteria taking into account climate variability and change should be developed particularly for vulnerable areas of the world. Hydrological design for water development projects should continue to be based on the past hydrological records. However, the design should consider both extremes from the past and scenarios from the GCMs in order to achieve the necessary flexibility to accommodate possible future changes.
Arnell, N.W. and Reynard, N.S. 1993. Impacts of Climate Change on River Flow Regimes in the United Kingdom. Institute of Hydrology, NERC, Water Directorate, DOE, July 1993.130 p.
Arnell, N.W., Brown, R.P.C. and Reynard, N.S. 1990. Impact of Climatic Variability and Change on River Flow Regimes in the UK. Report No. 107. Institute of Hydrology, Dec. 1990. 154 p.
Beran, M.A. 1986. The water resources impact of future climatic change and variability. In: Effects of Changes in Stratospheric Ozone and Global Climate. Vol. 1. J.G. Titus (ed.). US EPA/UNEP, Washington DC. pp. 299-328.
Berezner, A.S. 1987. Probability prediction of the Caspian Sea level with consideration of the development of water consuming industries in its basin. Water Resources 14 (1): 34-40.
Charney, J.G. 1975. Dynamics of deserts and droughts in the Sahel. Quart. J. R. Met. Soc. 101 (428): 193-202.
Conway, D., Krol, M., Alcamo, J. and Hume, M. 1996. Water availability in the Nile Basin: the interaction of global, regional and basin scale driving forces. Ambio (in press).
Cusbach, W.P, Hasselmann, K., Hoeck, H. et al. 1992. Time-dependent greenhouse warming computations with a coupled ocean-atmosphere model. Climate Dynamics, 8: 55-69.
Folland, C.K., Palmer, T.N. and Parker, E.D. 1986. Sahel rainfall and world-wide sea temperatures 1901-1985. Nature 320: 251-254.
Gleick, P.H. 1993. Water in Crises - A Guide to the World's Freshwater Resources. Oxford University Press. 473 p.
Grotch, S.L. 1989. A Statistical Intercomparison of Temperature and Precipitation Predicted by Four General Circulation Models. DOE Workshop on Greenhouse-Gas Induced Climate Change. US Dept. of Energy, Univ. Mass, May 1989.
Hadley Centre. 1992. Transient Climate Change Experiment. Hadley Centre, UK Met. Office, Aug. 1992. 20 p.
Hadley Centre. 1995. Modelling Climate Change, 1860-2050. Hadley Centre, UK Met. Office, Feb. 1995. 12 p.
Hall, M.J. and Sarenije, H.H.G. 1993. Climate and Land Use: A Feedback Mechanism? Conference Water and Environment - Key to Africa's Development, IHE Delft, Report Series 29, June 1993.
Hansen, J., Lacis, A., Rind, D. et al. 1984. Climate sensitivity: analysis of feedback effects. In: Climate Processes and Climate Sensitivity, Geophysical Monograph 29: 130-163.
Howell, P.P. and Allan, J.A. 1994. The Nile - sharing a scarce resource. In: Global Climate Change and the Nile Basin. N. Hume. Cambridge University Press, Cambridge. pp. 139-162.
Hulme, M. 1994. Regional climate change scenarios based on IPCC emissions projections with some illustrations for Africa. Area 26 (1): 33-44.
Hulme, M., Marsh, R. and Jones, P.P. 1992. Global changes in a humidity index between 1931-1960 and 1961-1990. Clim. Res. 2: 1-22.
Hurst, H.E. (1965). Long-term Storage. Constable, London.
IPCC [Intergovernmental Panel on Climate Change]. 1990. Climate Change - The IPCC Scientific Assessment. Cambridge University Press, Cambridge.
Kite, G.W. 1993. Application of a land class hydrological model to climate change. Water Resource Research 29 (7): 2377-2384.
Kite, G.W., Dalton, A. and Dion, K. (1994). Simulation of streamflow in a macro-scale watershed using GCM data. Water Research Paper 30 (5): 1547-1559.
Klemês, V. 1974. The Hurst Phenomenon: a puzzle. Water Resources Research 10 (4): 675-688.
Lewis, M.R. 1989. The variegated ocean: a view from space. New Scientist 1685 (7 Oct. 1989): 37-40.
Lockwood, J.G. 1985. World Climate Systems. Edward Arnold, London. 292 p.
MacDonald, Sir M. & Partners. 1965. Survey of Canje Reservoir Scheme, British Guyana. BRG/3, SF4/5. FAO/UN Special Fund. FAO, Rome.
MacDonald, Sir M. & Partners. 1988. Rehabilitation and improvement of water delivery systems in Old Lands, Egypt. Annex A: Water Resources. IBRD/UNDP Project No. EGY/85/012. Washington DC.
Maslin, M. 1993. Waiting for the Polar meltdown. New Scientist, 4 Sept. 1993: 36-41.
Milankovitch, M. 1930. Handbuch der Klimatologie l. Teil, A. Ed Koppen and Geiger, Berlin.
Mitchell, J.F.B., Manabe, S., Meleshko, V. and Tokioka, T. 1990. Equilibrium climate change - and its implications for the future. In: Climate Change, the 1990 Scientific Assessment, J.T. Houghton, G.J. Jenkins and J.J. Ephraums (eds.). Cambridge University Press, Cambridge, UK. pp. 131-172.
Mott MacDonald Ltd. 1992, 1993. Scenario Planning and Risk Analysis. Anglian Water Services Ltd, October 1992 and October 1993. Cambridge. UK.
Newman, S. and Fairbridge, W. 1986. The management of sea-level rise. Nature 320: 319-321.
Nicholson, S.E. 1986. The nature of rainfall variability in Africa south of the equator. J. Climatology 6: 515-530.
Rind, D. 1995. The tropics drying out. New Scientist, 6 May 1995: 37-40.
Rind, D., Goldberg, R., Hansen, J. et al. 1990. Potential evapotranspiration and the likelihood of future drought. J. Geophysical Res. 95 (D7): 9983-10004.
Rodda, J.C., Pieyns, S.A., Sehmi, N.S. and Matthews, G. 1993. Towards a world hydrological cycle. J. Hydrological Sciences 38 (5): 373-378.
Rowntree, P. 1989. The Needs of Climate Modellers for Water Runoff Data. Workshop on Global Runoff Data Sets and Grid Estimation, 10-15 Nov. 1988, Koblenz. World Climate Programme, WMO, June 1989. Geneva.
Schlesinger, M.E. and Zhao, Z.C. 1989. Seasonal climatic change introduced by doubled CO2 as simulated by the OSU atmospheric GCM/mixed-layer ocean model. J. Climate, 2: 429-495.
Schnell, C. 1984. Socio-economic impacts of a climate change due to a doubling of atmospheric CO2 content. In: Current Issues in Climate Research. Reidel, Dordrecht. pp. 270-287.
Semtner, A.J. 1984. The climate response of the Arctic Ocean to Soviet river diversions. Climate Change 6: 109-130.
Semtner, A.J. 1987. A numerical study of sea ice and ocean circulation in the Arctic. J. Phys. Ocean. 17: 1077-1099.
Shukla, J. and Minz, Y. 1982. Influence of the land surface evapotranspiration on the earth's climate. Science 215: 1077-1099.
Street-Perrot, A.E. and Perrot, R.A. 1990. Abrupt climate fluctuations in the tropics: the influence of Atlantic Ocean circulation. Nature 343: 607-611.
Thomson, D.J. 1995. The seasons, global temperature and precession. Science 268: 59-68.
Tyson, P.D. 1987. Climatic Change and Variability in Southern Africa. Oxford Univ. Press, Captetown. 220 p.
Vali-Khodjeni, A. 1991. Hydrology of the Caspian Sea and its problems, hydrology of man-made lakes. Proc. of Vienna Symposium, August 1991. IAHS Publ. No. 206. IAHS, Waterloo, Ontario.
Viner, D. and Hume, M. 1993. The UK Met. Office High Resolution GCM Transient Experiment (UKTR). Tech Note No. 4. CRU/Hadley Centre, Oct. 1993.
Wetherald, R.T. and Manabe, S. 1986. An investigation of cloud cover change in response to thermal forcing. Climatic Change, 10: 11-42.
Wigley, T.M.L. and Jones, P.D. 1985. Influences of precipitation changes and direct CO2 effects on stream flow. Nature 314: 149-152.
Wilson, C.A. and Mitchell, J.F.B. 1987. Similated climate and CO2-induced climate change over Western Europe. Climate Change, 10: 11-42.
World Bank/UNDP/ADB/EC/French Government. 1993. Sub-Saharan Hydro-logical Assessment. July 1993. Washington DC. 17 p.
