Estimating groundwater recharge uncertainty for a carbonate aquifer in a semi-arid region using the Kessler's method

Highlights

• Estimation of recharge rates in carbonate aquifers in a Mediterranean region.

• Provides the confidence interval and the quantification of uncertainty.

• It is a cost-free graphical tool for the estimation of recharge and quantification of uncertainty.

• Does not require large amounts of data to be applicable.

• The graphic tool to estimate the groundwater recharge and the associated uncertainty is easily updated with new data.

• It is applicable to the every semi-arid region with the same characteristics of a Mediterranean climate.

Abstract

Groundwater recharge is an important hydrological parameter for the quantification of water budgets in order to better achieve a sustainable groundwater management. Such management is paramount in the southern region of Portugal because groundwater represents more than 75% of the public water supply, the water demand for irrigated agriculture and tourism industries is rapidly growing, and the region is severely threatened by droughts and extreme dryness. Nevertheless, the estimation of recharge to carbonate aquifers in semi-arid regions remains challenging. Uncertainty analyses are still scarce and the natural variability of recharge is unknown in most cases. The present work aims at estimating the recharge rate in carbonate aquifers of southern Portugal based on the Kessler method, identifying the confidence interval with known variability and reduced uncertainty. The outcome is a cost-free graphical tool that calculates an annual recharge rate per year, proving to be a reliable tool for annual recharge estimation and allowing to overcome the limitations of the chronological variability and data acquisition associated with other current methods. Acknowledging the confidence intervals of the annual recharge estimations can improve greatly the decision making process for the regional sustainable groundwater resources management.

Introduction

Under normal circumstances in temperate and semi-arid regions, groundwater recharge occurs mostly in direct response to the infiltration of rainfall (Flint et al., 2004; Alcalá et al., 2011). The infiltrating precipitation and surface water that enters an underlying phreatic aquifer through the water table - sometimes after the infiltration event - is called the aquifer recharge. The fraction of infiltration that remains is transpired by plants, evaporated from the soil or returned to the surface as interflow or shallow ephemeral perched aquifers discharge (Scanlon et al., 2006; Custódio et al., 1997). The total recharge (R_T), which is referred to the water table aquifers, may be transferred vertically to deeper aquifers through aquitards or by horizontal groundwater flow. This recharge has to be separated from what is the net recharge, which refers to what remains in the water table aquifer after discounting what is used by phreatophytes (de Vries and Simmers, 2002). Net recharge is then added to aquifer storage and moves toward the aquifer system discharge points and areas. Groundwater recharge to shallow, unconfined, fractured and carbonate aquifers in semi-arid/arid regions is mainly a concentrated recharge (R_C). This recharge occurs by rainfall infiltration, from watercourses, near-surface bedrock fractures and sinkholes (high bedrock permeability), affecting a small fraction of the territory and producing local hydrological and hydrogeochemical responses in groundwater storage and quality (Flint et al., 2004; Alcalá et al., 2011). When groundwater recharge occurs in porous media through the rainfall infiltration then is called diffuse recharge (R_D). This type of mechanism gains importance to R_T when it takes place on well-developed soils in the plain regions with temperate to humid climates (Keese et al., 2005; Maréchal et al., 2009; Healy, 2010).

In well-developed karstic systems the response of the precipitation can be perceived in discharge points very rapidly, which is a clear indication of karstic aquifers with conduit flow behavior (Kessler, 1965; Andreo et al., 2008; Li et al., 2008). In this case, or in fissured and bare bedrock areas in arid regions, R_C represents a very significant fraction of R_T (Wood et al., 1997; Simmers et al., 1997).

Depending on variations in climate, lithology, soil type, fracturation, vegetation and land use, slope, etc., most large semi-arid carbonate regions and in particular in karstic aquifer systems located in a temperate climate, conditions are established to the combined and variable contribution to R_T of both diffuse and concentrated recharge mechanisms (Lerner et al., 1990; Wood et al., 1997; de Vries and Simmers, 2002; Alcalá et al., 2011). Even if the quantification of recharge by each mechanism is still difficult to calculate separately and accepting that R_C is a very important fraction of R_T, at these conditions karstic aquifer system and high bedrock permeability in a temperate climate and semi-arid/arid region - it may be assumed that the contribution of evapotranspiration and surface runoff to R_T is negligible (Custódio et al., 1997; Flint et al., 2004).

Some of the most important challenges in estimating groundwater recharge in carbonate aquifer are the spatiotemporal variability of recharge, the assessment and regional hydrological consequences of localized and indirect recharge, the extrapolation of localized data to a wider area, the determination of representative data (de Vries and Simmers, 2002) and finally, the uncertainties associated to every method (Scanlon et al., 2002; Flint et al., 2002). Additionally to these methodological challenges, global climate models consistently predict increases in drought and drier future soil conditions that raise concerns regarding climate change throughout the Mediterranean regions, particularly in Portugal and Spain (Wang, 2005; Heinrich and Gobiet, 2011; Sheffield and Wood, 2008). In the case of Portugal, the southern region is the most affected by droughts and extreme dryness, with clear evidence of land degradation and soil erosion, with significant trends to have extreme rainfall events in the winter months (Costa et al., 2008; Durão et al., 2010; Costa and Soares, 2012). This situation leads to significant water deficits with strong impacts on the environment, agriculture, industry and drinking water supply, all greatly dependent on groundwater (Cunha et al., 2006; Ribeiro and Cunha, 2010). From the combination of these conditions, with increasing demand for the quantification of water budgets in support of management decisions comes an increased need for practical methods to quantify recharge rates in semi-arid regions (Scanlon et al., 2002; Alcalá et al., 2011; Andreu et al., 2011).

Recharge estimation methods focus on local rainfall events or on measuring yield long-term areal values. Most studies are carried out at the aquifer scale and the recharge rates are applied directly to the outcrop area of the aquifer. According to the available data and conceptual recharge model applied in each case study the most appropriate method is selected in order to reduce the uncertainty (Martos-Rosillo et al., 2015). The simultaneous applicability of several methods is not always possible due the unavailability of hydrogeological data. According to Martos-Rosillo et al. (2015) when the time series of hydrogeological data was scarce the recharge was assessed using aquifer water budget in some aquifers, adequate for a delimited aquifer when its saturated volume at the beginning and the end of the period is the same. However, in many aquifers it is not possible to apply this method, and in this case the soil water balance method is used mostly the times, In fact, the soil water balance is the most widely used method to estimate R_T applied from arid to humid regions at different spatiotemporal scales (Lerner et al., 1990; Simmers et al., 1997; Scanlon et al., 2002). Nevertheless, the results from this method are usually associated to high uncertainty and biased values when used in dry climates (Wood et al., 1997; Keese et al., 2005; Alcalá et al., 2011). The chloride mass balance (Eriksson and Khunakasem, 1969) has been used in previous research to estimate the total recharge originating from both diffuse and preferential flow components in continental Spain (Alcalá and Custodio, 2014) and especially in carbonate aquifers of the Betic Cordillera (Andreu et al., 2011; Alcalá et al., 2011; Martos-Rosillo et al., 2013).

The inverse calibration in numerical modeling is commonly used to predict recharge rates during calibration to heads and groundwater flow rates (Scanlon et al., 2002). However, the lack of an accurate spatial and hydraulic characterization of karstic channeling makes the estimation of a precipitation-recharge relation really complex since the channels supplying recharge to groundwater are rather isolated from the evapotranspiration processes occurring in the surrounding matrix (Weiss and Gvirtzman, 2007; Zhang, 2014). In semi-arid regions, the water tables are usually deep and only long dry or wet periods will affect the net value of R_T.

Additionally, carbonate aquifers are likely to have turbulent flow components meaning that numerical methods applied in the saturated zone, such as those based on Darcy's Law, may be limited for estimating short-term R_T (Custódio et al., 1997; Simmers et al., 1997). One of the most accepted approach to model karst groundwater flow is the triple porosity approach (matrix, fracture and conduit), because storage is often dominant in the rock matrix and the fissure system whereas flow is achieved mainly through conduits. Palmer et al. (1999) expressed the opinion that ‘the heterogeneity of karst aquifers is so severe that it is virtually impossible to acquire sufficient information to construct a predictive digital model trustworthy enough to allow extrapolation of heads and flow conditions from known to unknown locations, let alone into the future’. In Portugal, one of the major problems when a study on the estimation of groundwater recharge in carbonate aquifer is carried out is the lack of data series, in specific, data from spring discharges and hydraulic heads. In most of the cases the carbonate aquifer discharges are not monitored at all and, when they are, the data is available only for a few years, presenting significant time gaps that considerably decrease the quality of the information. Thus, the piezometers installed on the aquifers are in most cases insufficiently representative for the characterization and comprehensive analysis of the systems (Nunes et al., 2004; Nascimento et al., 2013), limiting the knowledge on the evolution of the hydraulic heads and its spatial variability. Having this into account, some methods for the estimation of recharge have been used by several authors in order to encounter reliable results without requiring great amounts of data. One of the most widely used empirical method for the estimation of groundwater recharge in carbonate aquifers in the semi-arid Mediterranean region is APLIS (Andreo et al., 2008; Martos-Rosillo et al., 2015). This method estimate the mean recharge rate in carbonate aquifers, expressed as a percentage of precipitation, based on the variables altitude (A), slope (P), lithology (L), infiltration landforms (I) and soil type (S). Still, these methods have to be validated by other more formal methodologies to compare long-term estimates (de Vries and Simmers, 2002; Scanlon et al., 2006). One of the empirical methods used to study concentrated groundwater recharge in carbonate aquifer in semi-arid and temperate regions is the Kessler's method (Kessler, 1965). This method was first applied in karst aquifers located in Hungary in order to calculate the potential recharge rate (Rc) in carbonate rock via unsaturated zone. Since then it has been used successfully in Mediterranean environments, showing good correlation with values estimated by other mass balance calculations (Almeida, 1985; Silva, 1988; Andreo et al., 2008), and for our studied region, previous works have shown a good adaptation to the particular climatic conditions (Silva, 1988; Monteiro, 2001; Stigter et al., 2009). Some research has been done very recently on the quantification of the uncertainty of groundwater recharge estimated by other methods, namely by water and energy balance model (Xiea et al., 2018) and by the use of regression kriging and the chloride mass balance method (Crosbie et al., 2018). However, no research has been published on the quantification of uncertainty associated to the estimation of groundwater recharge using Kessler's method. In order to improve the reliability of the Kessler's method and its results when applied in the Mediterranean (semi-arid regions) the objective of this work was to identify a confidence interval of estimation, with known variability and uncertainty. This interval could then be used to indicate the reliability of the estimated value, applicable to water resources management. Hence, it becomes possible to provide a simple but consistent estimation of present and future annual groundwater recharge.