For many of the low islands of the tropical Pacific, freshwater is a scarce resource. Water catchment areas are small and groundwater storage is a shallow fresh water lens. The high hydraulic conductivities of the coral and sand substrate means surface water is limited. Realization of the possible impact of climate change has highlighted the sensitivity of island communities to the availability of water. However, impact evaluation requires specialized data as well as appropriate sensitivity assessment methodologies. This is the second of a two part study. The first addressed the data problem by assembling and validating a suitable database. The second develops an island water balance model and applies a sensitivity assessment. Data are at a 2.5° × 2.5° latitude–longitude grid resolution for the Pacific bounded by coordinates 30°S to 30°N and 155°E to 120°W. Output is in the form of Climate Change Sensitivity Index maps that show the impact on the spatial redistribution of climate-determined freshwater resources under various climate scenarios. The method allows for estimation of water deficits or surpluses for low islands located in any part of the study area. Areas of high sensitivity to climatic change are those that sit between margins of very wet and very dry zones. Their extent is determined by the gradients at the margins. Steep gradients define small areas of high sensitivity, whereas gentle gradients appear as large areas of high sensitivity. Adjustments to the model for differing local surface conditions on different islands can be easily made, which allows a sensitivity assessment of individual islands, even for islands with no climate station data. The approach could be a powerful tool to gain useful information on the influence of climate change on freshwater resources of low islands. Planning decision-making is possible without knowing precisely the magnitude of climate change that might occur.
The Pacific Islands region consists of nearly 30 000 islands scattered over 30 million km2. Approximately 1000 are inhabited small islands, most of which are located in the tropical and sub-tropical zones of the central and southern Pacific Ocean (White and Falkland, 2010; de Freitas, 2011). Many of the so called ‘low islands’ are atoll reef islets composed of unconsolidated coral-derived substrate, others are low limestone islands comprising lithified reef geology; but both types face similar issues with vulnerable groundwater resources (Terry and Chui, 2012). Water catchment areas are small and groundwater storage is in the form of a shallow fresh water lens floating on salt water, as shown schematically in Figure 1 (de Freitas, 2009; Chui and Terry, 2011; Terry and Chui, 2012). They are particularly vulnerable to extended droughts or periods of below average precipitation because of their limited storage capacity (Wheatcraft and Buddemeier, 1981; Giambelluca et al., 1988; Nullet and Giambelluca, 1988; White et al., 1999; Chui and Terry, 2011; Terry and Chui, 2012). Surface water resources are extremely rare. The high hydraulic conductivity of coral sand and permeability of the soil and underlying regolith cause a high infiltration rate, which impedes surface runoff and an accumulation of water on the surface. Roof catchment of rainwater is therefore an important supply of freshwater on small islands. But the exploitation of the island's groundwater resources is frequently used in conjunction with rainwater collection to meet the water demand (Lloyd et al., 1980).
Freshwater resource management focuses both on groundwater resources as well as on the soil water availability that is crucial for plant growth and thus for rain-fed agriculture on these islands. Freshwater resources of Pacific islands are coming under increasing pressure as populations grow and rates of development increase. Already, in some of these islands, population densities are as high as 12 000 people/km2 (White and Falkland, 2010). This pressure is superimposed on spatial and temporal variation in the availability of precipitation which depends largely on climatic conditions. Realization of the possible impact of climate change, natural or anthropogenic, has highlighted the sensitivity of island communities to the availability of freshwater (Chui and Terry, 2011; Terry and Chui, 2012).
The size and surface area of the island and the aquifer geology determine the upper volume limit of the freshwater lens (White et al., 2007), whereas the climate determines the replenishment of the lens. This means the climate defines the sustainable yield of water that can be pumped out of the freshwater lens. But only a few islands hold specialized climatic data and information on groundwater resources, which hinders a realization of sustainable water management (Falkland, 1999). Thus, the first problem is availability of suitable data. The second problem is managing uncertainties in climate.
Projections of future climate variability and change are required for informed resource management decision-making, but such projections are plagued by uncertainties on the regional scale and these uncertainties are likely to persist in the near future (Barnett, 2001). All the same, policymakers need to work towards adaptation strategies to mitigate impacts and exploit changed opportunities. Sensitivity methodologies address the problem of uncertain projections of future climates and provide useful results for decision-makers. The sensitivity approach copes with uncertainties by identifying regions of high susceptibility to change if a change in climatic were to occur (de Freitas and Fowler, 1989).
The overall aim of the current research project is to address these two problems, namely: (1) the lack of island-specific climatic data to assess sensitivity; and (2) an appropriate methodology to assess the sensitivity of freshwater resources to climatic change and variability on atolls and other small low islands. Helbig et al. (2011) addressed the first problem. The research identified data required for water balance assessments on low islands of the tropical Pacific region, then assembled and validated that database. In the second part reported here, a scheme is developed to assess sensitivity using that data.
2. Background to the methodology
Any assessment of adaptation to climate change effects on water resources must first consider the impact of climate change on hydrology. This latter requires knowledge of both future climate as well as methods capable of transforming this knowledge into likely hydrological effects. There are two ways of approaching this, the scenario or the sensitivity assessment approach. The first is by far the most common and is driven by our inability to adequately forecast future climate. Scenarios are effectively ‘what if’ statements that represent plausible future climate states from a range of possibilities based on the current state of knowledge of how the global climate system works using general circulation models (GCMs) of global climate along with estimates of future rates of greenhouse gas emissions and atmospheric concentrations. Scenarios project greenhouse gas emissions, which are then entered into the GCMs to project climate change.
In the scenario approach, a future climate state is identified and impacts evaluated. But this method is hampered by the unreliability of GCMs, especially at the regional level. Moreover, we do not know which climate change scenario to use. Clearly, the consequences of using inappropriate scenarios could have serious implications for planning adaptive responses. Sometimes there is an implicit assumption that a specific changed climate condition is predicted, reinforced by the fact that a GCM is limited to calculating an equilibrium response condition. The heart of the problem of using scenarios lies in the scale mismatch between global climate models and data required for decision-making at the local level. GCMs provide information at spatial resolutions of several tens of thousands of square kilometres. Planners at the local level require data on at least daily scales and at a resolution of perhaps a few square kilometres. There are advocates for the use of regional downscaling from GCMs to assess impacts on future water resources, but an underlying problem is that the prediction skill of the downscaling is no better than that of the GCMs. No matter how they are used, climate scenarios remain the single greatest source of uncertainty in hydrological impact assessments (Bergström et al., 2001). In the alternative approach using sensitivity assessment, many of these problems can be circumnavigated.
The aim of climate change impact assessment is to determine how the availability of climatic resources will change and which regions will lose or gain from these changes. The impact potential of a given change in climate is related to overall sensitivity of a particular water supply or demand unit to those aspects of climate that do change; or it may be related to the particular climate type or climate regimes in which change occurs (de Freitas, 2009). For example, a 10% increase or decrease in rainfall may be of little consequence in an equatorial climate region where there are already extended periods of high rainfall throughout the year. On the other hand, sub-humid environment may be highly sensitive and respond dramatically to even the smallest decrease or increase in precipitation, which will reduce water availability and further increase water demand, respectively. There are a variety of ways of identifying sensitivity. In theory, sensitivity of a region to changes in climate does vary depending on climate type or regime. Climatic types can be characterized and assessed on the basis of this sensitivity since a given change will perturb some climatic regimes more than others (de Freitas and Fowler, 1989).
By identifying the sensitivity to climate and evaluating it in terms of the adaptive capacity of the exposure unit, vulnerability of water resources to climate change may be determined and assessed. With this information, planning decisions would be possible without knowing precisely the magnitude of climate change that will occur. Research is needed to develop and test sensitivity assessment methodologies to cope with this, but some methods for water resource assessment are currently in use (de Freitas and Fowler, 1989; Fowler and de Freitas, 1990; van Minnen et al., 2000; Lynch et al., 2001; Semadeni-Davies, 2004). The results of sensitivity assessments give a birds-eye view of regional sensitivity in terms of key environmental responses to a range of possible changed climate conditions. The procedure provides a means by which a variety of climate change scenarios can be used to identify regions that are more or less sensitive to certain changes. This facilitates evaluation of impact potential for policy planning purposes. A ‘what if’ or scenario approach to climate change predictions can then be used to assess the desirability of societal adjustments in sensitive regions designed either to amplify beneficial or dampen undesirable effects of change. An advantage is its flexibility. A wide range of new or changed scenarios can be easily handled, which avoids the need to rerun the transfer function, thus facilitating use by non-climate specialists such as planners and policy makers wanting to reassess impacts on water resources.
Sensitivity analysis in water resource assessment can be rendered as the sensitivity of a specific response variable, soil moisture for example, to change in the two controlling climate features or processes, potential evapotranspiration and precipitation for example. The relationship between the response variable and climate determined from a pre-tested set of relationships, usually in the form of an empirical model, called a transfer function (de Freitas and Fowler, 1989; Fowler and de Freitas 1990; de Freitas, 2009). Changes might be simply percentage adjustments to the each of the driving variables. The response is a value relative to the unamended baseline data representing no climate change. The result is a direct measure of sensitivity. For example, a 20% response to a 10% change in a controlling climate variable is clearly an example of impact amplification. Conceptual models to visually demonstrate these relationships are given by de Freitas and Fowler (1989).
Earlier work dealing with the spatial dimensions of climate change sensitivity focussed on the agricultural sector by examining the possible shift of margins of zones of high productivity for individual crops (e.g. Newman, 1980; Parry, 1985). Agricultural margins are defined as regions ‘where activities are at the edge of their ideal climatic region’ (Parry, 1985, p. 352). Locations within biophysical margins of regions of the same agricultural or climatic conditions show highest sensitivity. The spatial shift of these areas caused by changing climatic conditions gives an indication of the type and location of impact. In the current work, the aim to produce what we call Climate Change Sensitivity Index (CCSI) maps that provide a concise summary of the impact on the spatial redistribution of freshwater resources of low Pacific islands due to changing climatic conditions.
The influence of climate on freshwater resources can be modelled by a water balance or budget of an atoll or low island. A water balance describes how precipitation inputs are distributed to different reservoirs (e.g. to the atmosphere via evapotranspiration, to the groundwater aquifer through deep percolation) in a defined region and time period. The procedure can be used to account for the volume of water moving through a system in terms of inputs, outputs and storage terms for a particular area and time period. The water balance can be drawn for two different reference zones. The upper zone represents the soil root zone, and is influenced mainly by climatic processes, soil and vegetation properties. The lower zone includes the groundwater table. The output of the soil water balance at the lower boundary can be considered as the input in the groundwater system. Because of the limited size and storage volumes of groundwater lenses, rapid turnover times are typical for low island hydrological systems (Falkland, 1999). In this study, the soil water budget is modelled for average monthly climatic conditions.
Once rainfall reaches the surface it enters the soil or runs off as surface flow. The latter is rare on atolls due to porous coral-derived soils (Fosberg and Caroll, 1965). If energy is available, water evaporates from the surface and is transpired by plants. Precipitation minus evapotranspiration determines water that is available for soil moisture replenishment. Soil moisture is the storage term in the water budget. The amount of water that can be stored is defined by the field capacity of the soil. If the soil moisture exceeds the field capacity, groundwater recharge occurs. If evapotranspiration exceeds precipitation, water is taken from the soil to meet the water demand, until the available water is completely removed from the soil and high suction pressure prevents further removal of soil water. In this case, a soil moisture deficit occurs. A soil's available water capacity (AWC) determines the maximum amount of water that is available for plants. A water surplus is considered an input to the water balance of the freshwater lens. Since freshwater has a lower density than salt water, the accumulated recharge builds up a floating freshwater lens in the porous rock of the island. By and large, there is an overall increase in lens thickness with increasing recharge rates Bailey et al. (2009). The soil water budget, or balance, can be expressed as:
where SD is either soil water surplus (S) or deficit (D), P is precipitation, Ea is actual evapotranspiration and ΔSm is change in soil moisture storage. If estimates of potential evapotranspiration (Ep) are available, water balance models based on the Thornthwaite–Mather water balance method (Thornthwaite and Mather, 1955, 1957)—described by Mather (1978)—can be used to estimate Ea. An example is given in Table 1, which shows monthly data for the 160°W and 20°S grid-square. Based on literature, AWC is taken as 80 mm, which is typical of atoll soils (Alley, 1984; Nullet, 1987; Nullet and Giambelluca, 1988).
Table 1. Monthly climatic water balance for the region of 160°W and 20°S, where P is precipitation, Ea is actual evapotranspiration and Sm is soil moisture storage, D is soil water deficit and S is soil water surplus. All data in mm
The Priestley and Taylor (1972) method is used to estimate Ep given as:
where Eeq is equilibrium evaporation, Δ is the slope of the saturation vapour pressure curve that is a function of air temperature, γ is the psychrometric constant, Q* is net allwave radiation and λ is the latent heat of vaporization of water. Derivation of Q* as used in the current research is discussed by Helbig et al. (2011). Eeq is the lower limit of Ea, but in most circumstances, Eeq is enhanced by advection. Priestley and Taylor (1972) suggest a constant coefficient of α = 1.26 to adjust Eeq. The reliability of this value of the coefficient is discussed in detail by Eichinger et al. (1996). Brutsaert (1982) showed it to be appropriate for general use, while De Bruin (1983) and McAneney and Itier (1996) confirmed it suitability for use in the humid tropics. The Priestley–Taylor method has been successfully used in studies of Pacific atolls (De Bruin and Keijman, 1979; Nullet, 1987; Giambelluca et al., 1988; Nullet and Giambelluca, 1988).
Low islands have little or no orographic influence upon cloud and precipitation patterns due to their low elevation (Lavoie, 1963); thus, synoptic climatic conditions of small low islands are almost identical with that over the surrounding ocean (Giambelluca et al., 1988). Because of this, datasets based on areal projections of satellite and surface observations and extended over large areas for latitude-longitude grid squares may be used (Uppalla et al., 2005). The first part of the current project assessed data availability, suitability and reliability (Helbig et al., 2011). Reanalysis ERA-40 data (1962-2000) are used comprising solar and longwave radiation, air temperature and precipitation at a 2.5° × 2.5° latitude–longitude grid resolution for that part of the Pacific bounded by coordinates 30°S to 30°N and 150°E to 120°W, as evaluated in the first part of the research reported by Helbig et al. (2011).
Employing the above approach, one can estimate the monthly D and S under average climatic conditions. Groundwater recharge (S) rates are relevant to replenishment of the freshwater lens and D to likely demands on it for irrigation or as an indicator of plant water stress or drought. The critical level for S varies depending on the state of development of the water management systems, population size and water using activities, which in the present circumstances is set at 100 mm. In the case of D, the critical level is assumed to be three months per year of severe deficit, which is defined in the current research as D being larger than 50% of the monthly P. This is to take into account that the negative impact on plant growth depends on the magnitude of D and varies between different plant types (Jackson, 1989). The impact of changed climatic conditions is assessed by running the water balance model under different climatic conditions. The results are mapped as shifts of isopleths on maps accounting for different freshwater resource questions. The areas of high sensitivity of freshwater resources on atolls are simulated using the water balance.
Modelled outputs are generated on a monthly basis; however, mainly annual means are reported here given that generalized results of the methods used are the focus. The aim is to produce CCSI maps that identify regions or zones showing different magnitudes of sensitivity to climate change. The focus is on the magnitude of spatial shifts in patterns. The first example is shown in Figure 2 in which an annual zero SD (i.e. the sum of annual S and annual D equals 0 mm) is taken to represent a critical threshold relevant to freshwater resources on atolls. Figure 2 illustrates the spatial impact of a change in P visualized as a shift of the annual zero SD isoline. An increase of P results in a shift towards the drier regions representing an improvement in freshwater resource conditions on atolls located in this area. A decrease of P has an inverse effect. The magnitude of this shift depends not only on the magnitude of change of P, but also on the area's sensitivity to changes in P. An increase or decrease in P smaller than 20% needed to bring about a zero SD is taken to characterize a region of high sensitivity to variability and change in P. The higher the required increase or decrease, the lower is the sensitivity to change of low islands in the area.
Three zones which are affected by a change of P of less than 20% can be identified in Figure 2. They differ in their spatial extent. The first is an elongated belt extending from the north western part of the study area to the eastern part and is extensive north of the Marshall Islands and narrows to the east. The second is the belt east of Kiribati and French Polynesia, which is uniformly narrow. The third and most extensive is south of Vanuatu, Fiji and the Cook Islands. The shaded area bordered by +20% and the −20% isolines is the zone of high sensitivity to change in rainfall. In the future, should rainfall decline, atolls in this zone would suffer the most from soil moisture deficits and, in a worst case scenario, they would experience an increase in the frequency of droughts. On the other hand, should rainfall increase, atolls in this zone would stand to benefit the most.
Two large very dry areas in the east require an increase of more than 100% in P to bring about an annual zero SD (Figure 2). Areas outside these three zones can still be affected by change in P, but larger and less likely deviations from the average are needed. Annual CCSI maps conceal short terms variability, which might impact water resource availability of island communities for extended periods of the year. For example, months of D can be followed by months of S and an annual sum would result in a zero SD. For practical applications and more detailed analyses, CCSI maps of monthly rather the annual data should be used.
Figure 3 is another CCSI map for P, but in this case showing areas of sensitivity with regard to drought conditions (i.e. more than three months of severe soil water deficit). In this case, isolines indicate percentage change in annual P required to bring about 3 months of severe soil water deficit. The isoline representing three months of severe D in a year is subjectively defined as the critical threshold concerning the soil moisture variations during the year and thus concerning rain-fed agricultural purposes on atolls. A severe deficit is defined in the current research as D being larger than 50% of the monthly P. Since monthly averages are used, a deficit does not necessarily occur on all days of a month with a deficit. Figure 3 illustrates the effects of two scenarios: a change in P of ±20% and ±40%. The results show that the islands of New Caledonia are highly sensitive to an increase in P of 20% and the Phoenix Islands to a decrease in P of 20%. Other locations, such as parts of Vanuatu and Fiji, Tonga, Niue, the Cook Islands and Kiribati, would experience worsened soil moisture conditions in years with 40% less P than average if average precipitation were to decrease by 40%.
Figure 4 illustrates the shift of the isoline representing a water surplus (S) of 100 mm/year assuming an increase or decrease in Q* of 10 and 20% (with effect on evapotranspiration). This critical level is set subjectively and varies due to different factors; for example, the state of development of water management systems, population size and type of water use. Areas of high sensitivity are for the most part less spatially extensive than for the same change in P indicating a smaller influence of Q* on the S rates. The area of high sensitivity in the northern part of the study area is clearly larger than the narrow belts in the vicinity of the Phoenix Islands and New Caledonia in the south.
Given that any change in climate is unlikely to affect only one of P and Q*, a combined scenario of change is explored. Precipitation can only occur when clouds are present and cloud cover reduces Q*. That implies an increase of P with decreasing Q* and vice versa (Liu and Scott, 2001). Four different scenarios are illustrated in Figure 5. The same method is applied as used for Figure 4. The negative correlation between P and Q* causes an amplification of the effect. The greater extent of highly sensitive areas than in the previous assessments demonstrates this. Many islands are located within the highly sensitive areas, including those of New Caledonia, Tonga, parts of the Cook Islands, Tubuai, Pitcairn Island, Phoenix Islands and Kiribati.
The sensitivity assessments so far were based on surplus and deficit on an annual basis. Other assessment methods can be applied by using monthly time steps. Here effects of decreased P or Q* on S and D are evaluated for selected months. However, the same method can be applied to examine impacts of changed Q* or combined changes.
Assuming a decrease of 40% in monthly precipitation, CCSI maps for the months of January (Figure 6) and July (Figure 7) show the effects of the change. Areas where groundwater recharge is very likely to persist under the new climatic state are marked by the symbol ‘+’. These are exceedingly wet regions. Areas where a soil moisture deficit is likely to remain under the new climatic state in the month are marked by the symbol ‘−’. These are exceedingly dry regions. The regions are separated by a zone rather than a boundary. Adjacent to the exceedingly dry region is a belt of sensitivity in which days of soil moisture deficits will occur at some time during the month (SZD in Figures 6 and 7). Adjacent to the exceedingly wet region is a belt of sensitivity in which days of soil moisture surplus disappear (SZR). In the centre of the transition zone is a neutral belt in which available soil moisture (AWC) is reduced but remains between 0 mm and water holding capacity (i.e. 80 mm) without recharge to the groundwater (NZ in Figures 6 and 7).
A decrease in P results in a shift of the neutral zone (with soil moisture ranging between zero and maximum water holding capacity) towards the wet region (Figure 6). With the change, some islands that were previously located in this neutral belt where AWC remains between 0 mm and 80 mm now experience soil moisture deficits. These lie in the sensitive belt (SZD in Figure 6), which includes New Caledonia, Hawaii, Kiritimati, Pitcairn and Tubua. The neutral zone includes the Marshall Islands, Kiribati, southern parts of Vanuatu and the Cook Islands and Tonga (NZ in Figure 6). Those islands that previously experienced a water surplus during January are now likely to lack recharge water since water holding capacity is not reached any longer. This belt (SZS in Figure 6) includes the northern part of Vanuatu, Fiji, Western Samoa, the northern Cook Islands and parts of French Polynesia. Islands located outside this transition zone are not severely affected by a 40% decrease in precipitation since their climate is either exceedingly wet or dry. In fact, there are few inhabited low islands in the exceedingly dry zone because of the lack of freshwater.
Figure 7 shows the CCSI map for July when the ITCZ has moved northwards generating the wet season in the Northern Hemisphere. Compared with January, the impacts of decreased P are limited to a smaller area in the northern part of the study area; whereas the opposite is true for the southern part of the study area where a water surplus ends in large areas during July. Some islands that in January experienced water surplus, now lie in a neutral belt (NZ in Figure 7), for example, some islands in the Vanuatu group, Fiji, the northern islands of Tonga and parts of French Polynesia. These results highlight the importance of changes in the seasonality of an island's climate when assessing the impacts on freshwater resources.
Generally speaking, assuming a decrease in precipitation, the area covered by SZD and SZR are different in January and July. SZR covers a larger area in latitudes south of 10°S in the dry season on the Southern Hemisphere (July) than in the wet season (January). The area of SZD is larger north of 10°N in the dry season on the Northern Hemisphere (January) than in the wet season (July). The pattern of SZD and SZR in equatorial regions are similar in January and July. Zones of sensitivity are smaller in areas with steep gradients of P or Q*. The largest sensitive areas are those that sit between the margins of the very wet and very dry regions that exhibit only small gradients, which is the case for the northwestern and the southern part of the study area. It may be worth reiterating that the results apply only to low islands. High islands such most islands of Vanuatu, Fiji, Western Samoa and Hawaii are not the subject of this study.
Future research could attempt ‘on-the-ground’ validation of the results of sensitivity analysis with a case study. To do this would require data on groundwater recharge rates from water table height monitoring and rates of actual evapotranspiration made using lysimeters along with corresponding climatic data from several low islands over a reasonable period of time. Clearly, these requirements are demanding and the challenge in acquiring the data needed is formidable. As far as one knows, there are no atoll islands that provide robust, long-term climatic and groundwater recharge data. Usually recharge is not measured since it is a difficult variable to determine directly. One option is to use total volumes of the freshwater lens, but this may be affected by groundwater pumping, which means the data are not useful. Additionally, site-specific soil properties and surface albedo will have to be taken into account. There is the exceptional case of Bonriki Island of the Tarawa atoll in Kiribati for which average long-term average recharge data are available (Lloyd et al., 1980; White et al., 2007). The model output for the current research (Tarawa, Kiribati: 1°25′N, 173°00′E) has recharge at 1108 mm. This fits reasonably well with the recharge values by Lloyd et al. (1980) of 1302 mm for the year 1977 and by White et al. (2007) of 980 mm for the long-term average. This comparison indicates that our model produces water resource indicators that are in the correct range for sensitivity analysis.
The generalized nature of the method employed means that the precision of the results needs to be carefully considered. Boundaries of the zones of sensitivity are approximate. Model parameter AWC has to be adjusted to match soil and surface conditions of atolls to which the generalized result are applied. Knowledge of the patterns of change and variability of P and Q* is also necessary to interpret the results. Helbig et al. (2011) found that change and variability of P was greater than change and variability of Q* indicating the crucial importance of current and future patterns of P in the study area. Climatic changes in the study region are also expected to more strongly alter the precipitation patterns and magnitude than the patterns and magnitude of evapotranspiration rates (Bates et al., 2008).
Clearly, the finer points of decision making and planning will depend on resource specific considerations, the most important of which will relate to the level of climate-based freshwater resource use. For example, vulnerability of society to a change in climate will depend on the demand for the resource and levels of use. If there is a high level of resource use, the society is more sensitive to climatic change that leads to reduced water availability. The implications for water resource planning are clearly quite serious. In contrast to this, where there is low resource use there is a larger margin of safety, it provides a buffer against both sporadic dry periods or a trend in decreasing recharge to the groundwater lens resulting from, say, gradual climatic change leading to reduced precipitation.
Another consideration is that not only changes in average conditions can have severe impacts on freshwater resources and communities depending on them. For example, Katz and Brown (1994) and Smit et al. (2000) highlight the greater importance of climate variability for climate impact assessments. Nicholls (1988) stated that actual precipitation amounts for individual years can differ largely from average values in the tropical Pacific. This interannual climate variability related to the Southern Oscillation is especially relevant to the Pacific region and can have severe impacts on communities there (Smit et al., 2000). Similarly, a significant proportion of the annual precipitation in the tropical Pacific is delivered by large rainfall events associated with storms or tropical cyclones in the wet season (McGregor and Nieuwolt, 1998). Hydroclimatic sensitivity of low islands would be influenced by possible shifts in tropical storms patterns that might accompany large scale regional climate change.
An additional consideration relates to the assumption that low atoll islands have no orographic effect on rainfall production. In fact, some low barrier reef islands may experience orographic rainfall if lying close offshore of the windward coast of a neighbouring high island. On the other hand, they may experience reduced rainfall from a rainshadow effect if lying near the leeward coast of an adjacent high island. Low islands influenced in this way may be more or less sensitive to hydroclimatic change than remote atolls.
On many of about 1000 populated low islands in the tropical Pacific freshwater is a scarce resource. Water resources are under increasing pressure as populations grow and rates of development increase. Realization of the possible impact of climate change has highlighted the sensitivity of island communities to the availability of water. However, impact evaluation typically requires specialized data, reliable projections of future climate as well as appropriate sensitivity assessment methodologies. All the same, policymakers need to work towards an adaptation strategies to mitigate impacts and exploit changed opportunities. Sensitivity methodologies address the problem of uncertain projections of future climates and provide useful results for decision-makers. The sensitivity approach copes with uncertainty by identifying regions of high susceptibility of a particular reference variable (i.e. freshwater availability) to change if a change in climate were to occur. The current research project addresses the problem of the lack of island-specific climatic data to assess sensitivity and develops an appropriate methodology to assess the sensitivity freshwater resources to climatic change and variability on atolls and other small low islands.
The approach used converts changed climate conditions into impacts on freshwater resources. It provides a simple but powerful method to identify low islands that lie in areas of high sensitivity, even for regions with no data from climate stations.
Output is in the form Climate Change Sensitivity Index (CCSI) maps that provide a concise summary of the impact on the spatial redistribution of climate-determined freshwater resources due to changing climatic conditions. The method allows for estimation of water deficits or surpluses for low islands located in any part of the study area. Output gives an indication of the size and location of impact. A particularly useful index is per cent change in climatic variables (i.e. precipitation and evapotranspiration) required to bring about a certain threshold value of the reference variable (e.g. zero recharge/surplus). Because zones of relatively high precipitation amounts generally coincide with zones of low potential evapotranspiration, and vice versa, the negative correlation amplifies the climatic effects on freshwater availability. The margins of very wet and very dry zones and the area delimited by them are areas of high sensitivity to climatic change and variability. Their areal extent is determined by the gradients at the margins of the very wet and very dry zones. Steep gradients generally define small areas of high sensitivity, whereas gentle gradients appear as large areas of high sensitivity.
Adjustments to the model parameters for differing local surface characteristics (such as soil type and depth) on different islands can be easily made, which allows a sensitivity assessment of specific island groups or individual islands, even for islands with no climate station data. This sensitivity approach is suitable for a variety of climate studies and represents a powerful tool to gain essential information on the influence of climate on freshwater resources of atolls. It is shown how sensitivity to climate change may be assessed. With this information, planning decision-making is possible without knowing precisely the magnitude of climate change that might occur.
The authors are grateful to the European Centre for Medium-Range Weather Forecasts for providing ERA-40 reanalysis climate data. The support of the Friedrich Ebert Foundation that supported M. Helbig through a travel stipend is also gratefully acknowledged.