1. Introduction
Groundwater is an important source of water supply for domestic, agricultural, and industrial use in Thailand. An increase in population, urbanization, industrial, and agricultural growth has led to increasing demand for groundwater resources in many parts of the country where surface water supply is limited. Chiang Mai is one of the fastest-growing cities where groundwater has been utilized extensively to meet the demand of the population, agriculture, and tourism economy [
1,
2,
3]. The Chiang Mai basin (
Figure 1) is an intermontane basin covering parts of the Chiang Mai and Lamphun provinces, Northern Thailand. Outside the municipal areas, most populations mainly rely upon the availability of groundwater. The Department of Groundwater Resources (DGR) of Thailand [
3] has forecasted an increasing demand at a groundwater usage rate of 4.6 Mm
3/y, causing a decrease in groundwater storage and associated hydraulic heads.
There has been a significant decline in groundwater levels especially in the south (Hang Dong and San Pa Tong districts) and north and northeast (San Kamphaeng and Mae Rim districts), as a result of agricultural and population growth [
1,
2,
3]. Land use in the Chiang Mai basin has also changed recently. Urbanized and agricultural areas have increased due to population growth causing a significant change in groundwater extraction. Moreover, the Royal Thai Government had initiated nationwide groundwater exploration and development projects to mitigate drought, resulting in a drastic increase use of groundwater resources [
4].
Few studies were conducted to assess the groundwater potential, as well as the impact of groundwater exploitation of the Chiang Mai basin [
5,
6]. Groundwater potential, in this work, refers to (1) the ability of the aquifer to store and transmit groundwater, which can be assessed using field estimates of hydraulic conductivities and storage coefficients from pumping tests, (2) the amount of natural recharge that replenishes the groundwater aquifer annually, and (3) the total pumping rate that does not cause a significant decrease in hydraulic heads (i.e., safe-yield). The last two quantities for groundwater recharge assessment, natural recharge and safe-yield, will be obtained from groundwater modeling results. It should be noted that the quality of groundwater, which is beyond the scope of this paper, is not included in groundwater potential assessment.
This paper presents pioneering work that integrates large-scale field survey hydrogeological data to assess groundwater potential, including safe yield using the long-term transient groundwater flow model of unconsolidated aquifers of the Chiang Mai basin. Observed hydraulic head from 49 piezometers during 2006–2016 is used for calibration, whereas the 2016–2019 data is reserved for model validation. The numerical model, MODFLOW, will be coupled with the PEST [
7,
8] computer code to assist model calibration. The model will be used to assess yearly recharge variability for evaluating groundwater potential and to delineate the areas where groundwater usage is critical.
4. Results and Discussion
4.1. Steady-State and Transient Model Calibration
Steady-state groundwater flow model was simulated under equilibrium conditions, such as representing long-term average hydraulic balance, and/or conditions where aquifer storage changes are not significant [
22]. The model was also used to investigate hydraulic properties, boundary conditions, long-term (or average) behavior of evapotranspiration and recharge rates, and sensitivity analysis of different model parameters.
Figure 10a shows a scatter diagram for comparison of observed (measured in August 2006) versus simulated hydraulic heads. The error bars shown on the plot indicate the minimum and maximum possible heads observed during 2006–2016. Although some wells showed large discrepancies between modeled or simulated heads vs. measure heads (e.g., G265 and MW187), this regional model appears to be relatively reliable with the majority of observation wells fall within a 95% confidence interval limit. The values RMSE and NRMSE are 0.86 m and 0.42%, respectively.
Figure 10b shows a steady-state hydraulic head distribution, indicating that groundwater generally flows from the north, east, and west toward the central plain and southward with the major cone of depression located at southern parts of the basin.
Water budget calculation from steady-state simulation indicated that the major source of groundwater was recharge of approximately 140.69 Mm3/y, which corresponded to an average recharge of 134 mm/y over the entire basin. The main groundwater outflows were due to evapotranspiration, discharge to main rivers. Groundwater usage through pumping wells in unconsolidated aquifers was 22 Mm3/y or approximately 60,275 m3/d. River leakage had outflow more than inflow indicating the majority portion of rivers were gaining streams. General-head boundary inflow (in the north) had a high influx of groundwater, as expected because the northern part of the basin is mainly a groundwater recharge area.
The calibrated steady-state model was subsequently used as a starting point (i.e., initial conditions) of a long-term (10-year) transient simulation.
Figure 11 illustrates the result of the transient model calibration; the time-variable hydraulic heads from August 2006 to July 2016 of some selected wells. The model could capture the transient behavior of this groundwater system where groundwater level fluctuated seasonally (e.g., G135 and MW988). The final parameter values after calibration are shown in
Table 3. Spatially variable parameters such as hydraulic conductivities, recharge, and evapotranspiration rates include, geometric mean, minimum and maximum values, and their values were within reasonable range compared with initial parameters estimated from field tests and water budget calculations from previous studies. The true values of the riverbed and general-head conductance, on the other hand, were not available for comparison. Sensitivity analysis indicated the most sensitive parameters were hydraulic conductivities, recharge rates, evapotranspiration, and riverbed conductance, respectively (see
Table 3).
Final parameters distributions (K, S
s) are illustrated in
Figure 12 and
Figure 13 indicating that the Chiang Mai aquifer systems are highly heterogeneous.
Figure 14 and
Figure 15 illustrate the recharge and evapotranspiration, which vary temporally from year to year, respectively. Zones of high recharge rates are located in the northern, eastern, and parts of southwestern areas of the basin. These areas should be reserved for recharge protection areas. The transient model was investigated further to evaluate the effect of yearly precipitation variation on groundwater natural recharge. Based on the water budget analyses, it was confirmed that the values of natural recharge vary temporally and have a strong dependency on the precipitation pattern (see
Figure 16), with a Pearson’s correlation coefficient of 0.79. In recent years, the overall precipitation has been declining significantly and so has the groundwater recharge. This is also confirmed by observed groundwater heads at several locations where groundwater table or potentiometric surface, though seasonally varying, are continuously lowered.
4.2. Model Validation
After the transient calibration was completed, the model was executed in a forward mode to demonstrate its validity and applicability in prediction using an independent set of hydraulic head data (2016–2019). In the validation simulation, actual pumping rates and river stages were input. The recharge rates during 2016–2019, on the other hand, were approximated from annual precipitation (based on the recharge trend shown in
Figure 16, which is approximately 11–12% of the annual precipitation). As shown in
Figure 17, the model was able to capture the trends of measured hydraulic heads, which decreased over time.
4.3. Groundwater Potential of the Unconsolidated Aquifers
The groundwater potential of the basin has several definitions depending on the methods of evaluation. Traditionally, the groundwater potential of an aquifer system is assessed based on hydraulic parameters such as hydraulic conductivity (K) and storage coefficient (S) obtained from pumping test analyses. Aquifers with high K and S can be said to have high potential because of high flow and storage potentials. Another aspect of groundwater potential can be referred to as the ability of the aquifer to recover from hydraulic head drops. If the water table or potentiometric surface from the observation well(s) rebounds (or recovers) quickly within a reasonable time (e.g., dry and wet seasons or during pumping scheme), such an aquifer can be regarded as high potential. A more quantitative assessment of groundwater potential refers to an estimation of groundwater recharge and groundwater storage (or ‘change’ in groundwater storage). Total or net recharge estimation of a basin has been conducted using several methods such as groundwater level fluctuations, chloride balance, flow net analysis, water budget, etc., although the spatial and temporal distribution of recharge is not easy to evaluate. Analysis on groundwater storage or ‘change’ in groundwater storage, on the other hand, cannot be assessed straightforwardly and, thus, a groundwater model can be applied.
The groundwater model does not only provide the aquifer’s hydraulic parameters, but it also gives quantitative information on net recharge and aquifer storage. In this study, groundwater potential can be assessed as follows.
Hydraulic conductivity is one of the most important hydrogeological parameters needed for managing groundwater resources. It describes an ability of an aquifer to transmit water per unit area. Although a high value of hydraulic conductivity alone may not guarantee a large quantity of extractable groundwater, it can be used to assess how fast groundwater can flow and recover after pumping. The average hydraulic conductivities of the unconsolidated aquifers (Qfd, Qyt, Qot) of model layers 1–3 were in the range of 10−3 to 10−1 m/d (three orders of magnitude), indicating poor to fair aquifer conditions.
Storage coefficient is the volume of water that the aquifer releases or takes into storage per unit surface area of the aquifer per unit component of the head normal to that surface. In an unconfined aquifer, the storage coefficient corresponds to its specific yield. The value of the storage coefficient for typical water table aquifers (i.e., specific yield or S
y) varies from 0.01 to 0.35, while for well-confined aquifers, specific storage varies from 10
−5 to 10
−3 [
22]. In the Chiang Mai basin, field studies and model calibration results show that specific yield (S
y) varied from 10
−2–10
−1 whereas specific storage (S
s) shows a wider range of 10
−6–10
−4 m
−1. Values indicate that the storage coefficient of the Qfd, Qyt, and Qot aquifers of the Chiang Mai basin is generally lower than the normal range, so the groundwater potential is probably not in good condition.
Groundwater recharge and storage:
Figure 18 illustrates transient groundwater budget, recharge, and storage during the simulation period. It is clear that, in recent years, groundwater recharge and storage of the unconsolidated aquifers (Q’s) is declining. This is partly due to the lowering of annual precipitation, as discussed earlier. Based on the above assessment, it can be deduced that the groundwater potential of the unconsolidated aquifers (Qfd, Qyt, Qot), where the average total thickness is ~125 m, is not in good condition. Careful groundwater usage must be taken into consideration because groundwater storage is declining according to increasing groundwater usage and decreasing recharge. This observation is confirmed by a continuous decline in the hydraulic head which is indicated in long-term groundwater monitoring over the past fifteen years.
4.4. Evaluating Groundwater Safe Yield in Unconsolidated Aquifers
Sustainable or safe yield refers to the amount of groundwater that can be extracted from an aquifer on a sustained basis, economically and legally, without deteriorating the native groundwater quality or creating an undesirable effect, such as environmental damage. In this work, the safe yield is defined as the amount of extractable groundwater through pumping wells that may cause the drawdown, on average, to drop not more than the specified values. The pumping rate from Q units of 24.0 Mm
3 in 2015 was used as a base case. Then, the pumping rates were incrementally increased until the drawdown in all active extraction wells dropped, on average, to the specified preset criteria.
Table 4 shows that the values of yield (i.e., groundwater abstraction or pumping rates) corresponded to the targeted average drawdown of the basin for 1-, 2-, 3-, 4- and 5-m drawdown. For example, if the targeted drawdown was set at 2-m below current heads, groundwater abstraction should increase from 24.0 to 51.2 Mm
3/y. That is, a cumulative pumping rate of 51.2 Mm
3/y is the value of safe yield for an average of a 2-m drawdown.
5. Conclusions
This work represents a long-term, up-to-date, and comprehensive groundwater modeling study of the Chiang Mai basin. The model was set up in the simplest, most effective, manner because of data limitations, especially on the availability of hydraulic conductivity distribution and actual groundwater wells. A transient finite-difference groundwater flow simulation of the Chiang Mai basin was setup using the MODFLOW program. The model primarily focused on unconsolidated aquifers (Qfd, Qyt, and Qot), where the majority of local or domestic groundwater wells are extracted. The model calibration was achieved based on an automated parameter estimation scheme using the PEST program. Parameter sensitivity analysis during calibration was obtained using the perturbation method. The sustainable or safe yield based on specified average drawdown, as well as the water budget or groundwater potential, was assessed.
The numerical model consisted of 71 columns, 80 rows, and 4 layers. A uniform horizontal grid size of 1000 × 1000 m2 with variable thickness for all model layers was used. Major aquifers of this basin are unconsolidated to semi-consolidated aquifers. Model simulation results indicated that hydraulic heads in the north, east, and west regions were high and continuously decreased toward the center of the basin, and groundwater was discharged to main rivers and flowed southward. In some areas, groundwater level dropped significantly below ground surface, especially in the central plain where groundwater abstraction was high due to agricultural, domestic, and industrial needs. The model was calibrated using an automatic parameter estimation program namely PEST with the use of the pilot point technique to evaluate heterogeneous or spatially variable hydraulic conductivities, recharge rates, maximum evapotranspiration rates. Parameters’ sensitivity analysis from PEST allowed the assessment of the relative importance of the model parameters. The results showed that the model outcome was sensitive to, in order of importance, hydraulic conductivities, recharge rates, and evapotranspiration rates. The root mean square (RMSE) and normalized root-mean-square (NRMSE) errors of the steady-state model, based on 49 observation wells, were 1.26 m and 0.62%, respectively.
The transient model was calibrated and validated to demonstrate the capability of the model to simulate the seasonal variation of the hydraulic head (2006–2019). The model also indicated the prior assumption where recharge was expected to occur from July to September. Based on the water budget analysis from the transient model, it was confirmed that the values of natural recharge vary temporally and has a strong dependency on precipitation pattern. In recent years, the overall precipitation is significantly declining and so has the groundwater recharge. This is also confirmed by observed groundwater heads at several locations, where groundwater table or potentiometric surface, though seasonally varying, are continuously lowered.
Safe yield calculation suggested that the basin can sustain abstraction rates up to 214% (~51.2 Mm3/y for Qfd, Qyt, Qot) of the current extraction rates for a 2-m drawdown threshold. Overall, the groundwater potential of unconsolidated aquifers (Qfd, Qyt, Qot) in the Chiang Mai basin is generally not in good condition, indicating a crisis may occur if uncontrolled use of groundwater of the basin is not properly managed.
Although the numerical model of the regional groundwater system presented in this study was satisfactorily calibrated and validated and able to capture the transient head trends, it was clear that the model did not produce a perfect match in some wells where discrepancies between observed and simulated heads still exist. The model should then be used with caution. While the average or overall response of this regional groundwater system such as storage and recharge can be evaluated, there are still uncertainties associated with heads in individual wells, and interpretation or application must be handled with care.