1. Introduction
Groundwater occurring in the form of a thin lens floating on denser seawater is the primary source of freshwater in most small islands [
1]. This limited groundwater resource is highly vulnerable to saltwater intrusion due to natural causes such as droughts, storm surges, sea level rise, and anthropogenic activities such as increased groundwater withdrawals [
2,
3]. Many small islands in the Pacific face groundwater shortages during droughts associated with El Niño—Southern Oscillation (ENSO) events [
2] which are predicted to be frequent in the future due to global warming [
4]. Public water supply in small raised coral islands of the Pacific including Tongatapu and Niue [
5] relies on multiple vertical wells. Large-scale localized and unplanned abstraction can expedite the salinization of the freshwater lens [
2]. Over-pumping can severely contaminate the aquifer, and hence, it is essential to modify and manage the pumping rates, especially during droughts. Management of groundwater withdrawal requires knowledge about the extent of freshwater lens and identification of wells that are under high risk of saltwater intrusion [
6]. Hence, regional and wellscale modeling of freshwater lens are essential for planning water management strategies in small islands.
Various numerical models are available to assess the freshwater resources in small islands [
7]. However, quantitative assessment of freshwater resources and seawater intrusion impacts have been performed in only a small number of real-world islands [
8,
9]. Numerical models, which consider the dispersion zone between the freshwater lens and underlying seawater, can closely represent typical island groundwater processes [
10,
11,
12]. With fine discretization of space and time and sufficient monitoring data, dispersion models can provide a quantitative assessment of freshwater lens including the local effects of pumping [
7,
11,
13]. Proper monitoring of well salinities is often scarce, if not absent, in many real islands [
2] and it is difficult to obtain a sufficient quantity and quality of monitoring data for calibration [
14,
15]. Hence, due to the high computational demands and lack of sufficient monitoring data for calibration of heads and salinity, field-scale modeling of regional and well scale freshwater lens have been rarely attempted using dispersion models [
7,
11], and use of empirical models are suggested for well scale modeling [
16]. As a result, use of a single numerical model to evaluate both the island-wide extent of freshwater lens and the local effects of pumping remains a challenge when the size of an island is not so small.
Numerical models that assume freshwater and saltwater to be immiscible and separated by a sharp interface have been used for the evaluation of risks of saltwater intrusion to freshwater resources when dispersion models are difficult to be implemented in regional scales [
17]. Sharp interface numerical models have lesser data and computational requirements and have been used to predict salinity at pumping wells [
18,
19]. Hence, a sharp interface approach is used in this study for regional and well scale modeling of the freshwater lens on an island. It ensures a single numerical model is capable of providing a first-hand prediction of the extent of freshwater lens and the local effects of pumping.
Sharp interface two-phase numerical modeling was applied for Nantucket Island [
20] to assess the impact of projected pumping on the freshwater lens. The model predicted the general possibility of saltwater upconing into the well field when the interface was within the intake of deepest well, though no attempt was made to quantify individual well salinization. Freshwater heads were used for calibration of the model, and no island-wide comparison of the model predicted freshwater thickness with the field observations were done. The midpoint of freshwater-saltwater mixing zone was matched to the sharp interface and hence the encroachment of water with salinity higher than the freshwater limit (chloride content of 250 mg/L) into well screen could not be identified.
When a numerical model for groundwater flow is developed, the model is generally calibrated with observed groundwater heads [
21]. Some issues associated with the monitoring data of groundwater heads in small islands make it practically difficult to be used for calibration. The highest freshwater heads in small islands are typically of the order of 0.2 to 0.5 m above the mean sea level [
2,
22] due to the high hydraulic conductivity of the aquifer. Since the depth of the interface estimated by Ghyben-Herzberg’s approximation [
23,
24] is 40 times freshwater heads, highly accurate groundwater heads in the order of centimeter accuracy may be required to predict freshwater thickness. Additionally, due to the insufficiency of salinity monitoring data [
2,
7], the vertical profile of groundwater density is ambiguous, and so conversion of observed heads to equivalent freshwater heads is also challenging. In some islands like Tongatapu, which is the case study selected here, accurate land surface elevations are not available at monitoring locations.
Freshwater thickness is the most critical parameter to be assessed for groundwater management in small islands [
7]. In this study, the freshwater thickness is selected for calibration of the groundwater model to overcome the challenges associated with the use of freshwater heads in calibration. The two-fluid sharp interface approach used in this study [
25] represents groundwater flow in terms of freshwater and saltwater heads. Though it is unable to predict the nature of the transition zone, it can predict the response of the interface [
26] and pumping wells [
18,
27] to applied stresses. By adopting the upper limit of freshwater electrical conductivity as the interface, it is possible to predict the extent of freshwater thickness in the island.
The objective of this work is to develop a methodology based on a sharp interface numerical model to conduct regional and well scale modeling of island freshwater lens under long-term stresses. The methodology is applied to Tongatapu, a small Pacific Island. A numerical model is developed and calibrated for the average recharge and the current pumping condition and is then used to assess the effect of long-term stresses of recharge and pumping variations on saltwater intrusion and well salinization. This work expands the capacity of the sharp interface numerical model to regional and well scale modeling of freshwater lens in real-world small islands. The methodology developed can be applied to other similar island systems for the prediction of freshwater lens and to quantify the local effects of pumping. The impact of long-term stresses on saltwater intrusion along with well salinization is evaluated using the groundwater system of Tongatapu and generalized outcomes are relevant to other similar island systems with concentrated pumping. Tongatapu is the largest and the most populated island of the Kingdom of Tonga, a Pacific island nation. There has been an increasing trend of migration to Tongatapu, more specifically, to its capital Nuku’alofa from outer islands [
28]. Groundwater extracted from vertical wells meets the public water supply demands for Nuku’alofa due to the lack of surface water sources [
29]. With the increase in groundwater demands and expected climate changes in the future, there has been an increase in public concern over groundwater availability. Previous numerical modeling studies on Tongatapu [
30,
31] used a large grid size (about 1 km) for simulations and ignored the effects of groundwater pumping and future change in recharge patterns. For small islands like Tongatapu where groundwater pumping is concentrated over a small area, the risk of well salinization is high due to saltwater upconing [
2]. The results of this study can help to identify the wells with a high risk of salinization, which was rarely attempted in earlier modeling studies and can contribute to planning and management of groundwater development.
4. Discussion
The numerical model developed in this study revealed the quantitative behavior of freshwater lens and replicated observed freshwater thickness with a root mean square error of 1.44 m. Island aquifers are characterized by high heterogeneity and lack of data [
7]. Previous modeling studies on real islands using dispersion models reported scaled-RMSE of 10% for Bonriki Island model with an area of 9.5 km
2 [
12] and 21% for Kish Island model with an area of 112 km
2 [
10] for modeled versus measured salinity. Scaled RMSE for modeled versus measured freshwater thickness is 17.5% in this study (
Table 2). Given the large size of the island (259 km
2) and simplifying assumptions, the model-measured difference that is not unacceptable.
The total volume of freshwater lens is estimated to be 637 MCM for an annual recharge of 147.6 MCM. Current groundwater development is estimated at 6.1 MCM/year. The volume of the freshwater lens with the thickness greater than 10 m is less than 300 MCM, and exist in less than one-third of the total area of the island (
Table 3). It is seen that the current pumping rate of about 4.1% of annual recharge can be sustained by a freshwater lens with no saltwater intrusion under the assumption of uniform recharge throughout the year. However, its impact on predevelopment groundwater lens appears to be significant. The area and volume of the freshwater lens with thickness greater than 10 m are reduced by 9.4% and 11.9%, respectively, as compared with those for the predevelopment lens. Current pumping reduced freshwater thickness in an area of 12.3 km
2 around well field by over twometers, as compared to the predevelopment state of freshwater lens. Jakovovic et al. [
51] adopted an interface rise of two meters to delineate a region around a pumping well as saltwater upconing zone of influence (SUZI). A limit of two meters is about 25% of the average freshwater thickness of Tongatapu. Considering the above criteria of the saltwater upconing zone of influence, an area over 20 times well field size indicates a very significant reduction in freshwater thickness in Tongatapu.
Scenarios considered in the study investigated the effect of drought and increased pumping. For a scenario with a 40% decrease in average annual recharge, 86% of freshwater lens volume with thickness more than 10 m is lost. For pumping at 8.2% of the average annual recharge, loss of freshwater lens volume with thickness more than 10 m is 13%. Groundwater development in Tongatapu is concentrated over a smaller area which leads to gradient reversals (ocean to land) resulting in saltwater intrusions (
Figure 7). Salinization risk of each well is influenced by location from the center of the pumping well field and also the distance from the lagoon (
Figure 6 and
Figure 7). For scenarios with a reduction in recharge (0.69 R, 0.65 R, and 0.6 R), lateral saltwater intrusions are prominent, whereas saltwater intrusion due to upconing under wells is significant for scenarios with increased pumping rates (1.6 P, 1.8 P, and 2 P). However, wells with the highest risk of salinization are almost common for scenarios of reduced recharge and increased pumping. Identification of high-risk wells helps in the planning and management of groundwater pumping.
Among the scenarios considered in this study, the worst cases of 0.6 R and 2 P cause over 50% of public wells to be intruded by saltwater (
Table 4), even though total freshwater volume change is 24% for 0.6 R and 3% for 2 P, as compared to the base case (
Figure 5). Hence, evaluation of freshwater volume change alone may not indicate the potential threats due to pumping. In scenario 2 P, though the pumping rate is about 8.2% of the average annual recharge, 50% of wells are intruded by saltwater. Well salinization of over 50% of number of public wells indicates a reduction in freshwater production by 50%. This further emphasizes the need for groundwater management and planning.
To evaluate the relative impact of pumping and recharge, the total volume of water extracted is quantified as the reduction in recharge [
22]. This approach can also evaluate the relative impact of pumping distributed across total area and localized pumping. As base pumping rates are about 4% of recharge, scenario 1.8 P represents a 7% decrease in recharge. Well salinization, in terms of the number of wells intruded, under this scenario, is comparable to that of 0.65 R where effective recharge reduction is 39%, including pumping. Similarly, scenario 2 P and 0.6 R are quite comparable in terms of the number of wells salinized. It is seen that pumping impacts on well salinization are about five times that of recharge for Tongatapu Island. The high relative contribution of pumping to well salinization can be attributed to the high concentration of wells in the small area in Tongatapu.
Since seawater intrusion into pumping wells is a local phenomenon, sustainable groundwater development from a limited number of pumping wells is much smaller than the estimated volume of freshwater. White et al. [
22] estimated island-wide sustainable groundwater development to be 20% of groundwater recharge and total sustainable groundwater yield considering areal pumping rates as 54,000 to 72,000 m
3/d. The numerical model, considering the current distribution of pumping wells, indicates that extractions at about 6.5% of the annual average recharge, and 50% of the lower range of areal pumping rates, as mentioned above, can cause saltwater intrusions.
Even though the precipitation exhibits seasonal and yearly patterns, the steady-state assumption was not too unrealistic as the groundwater system responded quickly due to large hydraulic conductivity. The salinity monitoring wells in Tongatapu use multi-nested tubes resulting in different values of heads in each tube due to the density differences in the tubes. The mean water table elevation for Tongatapu from observations in village wells for 1971–2007 was 0.41 m above mean sea level [
22]. Error in deducing head from multi-nested monitoring wells might be large, as compared to the head above the MSL at the location. Hence, careful attention is required to make full use of the water level observations from multi-nested tubes.
Spatial heterogeneity of the aquifer and recharge distribution have not been considered. The tidal effects were not included in the study. The model developed considers the sharp interface at the freshwater limit. Hence, the fluid above the interface is freshwater, but in reality, the fluid below is not completely saltwater. Error of neglecting mixed saltwater below the interface on general flow pattern needs to be investigated further. While all the public wells for urban water supply have been included in the model, only 56 village wells were represented out of the over 200 village wells known to exist [
36] as there was no information about the location of other wells. Estimated total groundwater development rate of the island was applied uniformly to all wells in the well field as individual pumping rates were not known. A more accurate representation of wells, in terms of locations, pumping rates, and screen length could result in better estimates of well salinization.
5. Conclusions
Prediction of the extent of freshwater lens and quantification of the local effects of pumping is essential for the design and development of effective water management strategies in small islands. In this study, a steady-state sharp-interface numerical model of fresh and saline groundwater flow is developed to evaluate saltwater intrusion and well salinization in Tongatapu Island. The model has a comparatively fine grid size of 25 m × 25 m in public well field and 100 m × 100 m in other areas, which is able to better represent the freshwater distribution and saltwater intrusion characteristics at pumping wells. Freshwater thickness is used as the calibration target in this study with the sharp interface corresponding to the upper limit of freshwater electrical conductivity adopted in the Kingdom of Tonga. Such an approach has not been reported in previous applications of sharp interface numerical models for islands. Thus, the current study is the first, to the best of our knowledge, to conduct a comprehensive analysis of freshwater lens and well salinization using sharp interface approach in real islands.
Model results indicate that at present conditions, 46% of the total freshwater volume of Tongatapu, estimated around 630 MCM, is from freshwater lens thicker than 10 m and concentrated in two regions towards the center of the island. This may indicate sufficient availability of resources for groundwater development. However, the concentration of pumping wells in a single well field increases the risk of saltwater intrusion. Various scenarios of decreased recharge and increased pumping were considered to evaluate saltwater intrusion and well salinization. More than 50% of public wells are intruded by saltwater for 40% decreased recharge (0.6 R) or groundwater pumping at 8% of average annual recharge (2 P). It is seen that pumping impacts on well salinization are about five times that of recharge for Tongatapu Island. The higher relative contribution of pumping to well salinization can be attributed to the high concentration of wells in the small area. The results suggest that the risk of salinization for individual well depends on the distance from the center of well field and the distance from lagoon or coast. However, the closeness of the well to the center of well field results in higher saltwater ratios and the extent of upconing. Major factor leading to well salinization was saltwater upconing rather than lateral saltwater intrusions. Hence, it is imperative to include pumping wells in a groundwater model, especially when pumping is localized.
While the model is currently limited to steady state analyses, it can help water managers of the island to assess potential impacts on groundwater resources caused by changes in some important hydrogeologic parameters, such as pumping rates and recharge rates. It also helps to identify wells with a high risk of salinization. Currently, an unsteady model is being developed to simulate drought conditions. An enhanced model will be linked with an optimization method to identify the best groundwater management schemes under drought conditions. The study highlights the applicability of a sharp interface numerical model calibrated for freshwater thickness to assess the extent of freshwater lens, saltwater intrusion, and well salinization in real islands. It can be used as the first stage analysis or when sufficient data and computational resources are not available for implementation of dispersion models.