Groundwater recharge estimating in the Serra Geral aquifer system outcrop area-Paraná State, Brazil Estimativa de recarga das águas subterrâneas no sistema aquífero Serra Geral no Estado do Paraná, Brasil

Knowing the groundwater availability is necessary to ensure its sustainability. Groundwater recharge estimation is an important tool for determining the renewable reserves, contributing to the management in quantitative terms. Exploitation rates that exceed recharge may cause aquifer unsustainability. In addition to the quantitative aspects of groundwater, recharge studies also contribute on aquifer vulnerability evaluations. Methods such as DRASTIC (ALLER et al., 1987), SINTACS (CIVITA et al., 1997) and IS (FRANCÉS et al., 2001) uses recharge as an input parameter for aquifer vulnerability evaluation, since aquifers in regions with higher recharge rates are, in general, more vulnerable to contamination, as it facilitates contaminant migration into the saturated zone. Many methods are commonly used to quantify groundwater re charge (SIMMERS, 1988; HEALY AND COOK, 2002; SCANLON et al., 2002; ECKHARDT, 2005), for example: water balance, water table fluctuation, Darcy law, hydrogeologic models, chemical tracers and baseflow separation. According to Scanlon et al. (2002), groundwater recharge estimation methods has variable reliability.

ration consists in the separation of an observed hydrograph of a river in at least two components: surface flow and baseflow.There are many methods to perform a hydrograph separation, including the use of hydrochemical and isotopic tracers and the hydrograph analysis by graphical methods or digital stream-flow filtering (COLLISCHONN AND FAN, 2013).
One famous method is the Eckhardt baseflow filter (ECKHARDT, 2005), which is based in the numerical filtering of a streamflow time-series.Eckhardt's method for baseflow separation consists of a recursive digital filter, based on two parameters: a and BFImax.The parameter a can be determined directly from a graphical analysis of the hydrograph recession, but BFImax is related to the baseflow and total flow relation and can be estimated according to the local geology and the streamflow nature (perennial or ephemeral).Eckhardt (2005) stablished pre-defined values for BFImax ranging from 0.25 to 0.8.Collischonn and Fan (2013) proposed that the BFImax can be calculated through application of inverse filter or the relation Q90/Q50.
One example of Eckhardt's method application is from Mattiuzi et al. (2016), that used Eckhardt Filter to estimate groundwater recharge rates in the Ibicuí River basin in Rio Grande do Sul -Brazil through baseflow separation.The authors adopted the Collischonn and Fan (2013) proposal for the calculation of BFImax index through the relation Q90/Q50.The recharge rates varied from 88 to 378.8 mm/year.The highest recharge rates were observed in geological units with more permeable soils and higher values of transmissivity and rainfall.The lower rates occurred on crystalline geological units, with low rainfall and clayey soils.Gonzales et al. (2009) studied the baseflow in a flat land in the Netherlands.Many methods were used for the hydrograph separation, including those based on chemical tracers.The authors verified that Eckhardt's Filter with the BFImax parameter calibra ted through chemical tracer (dissolved silica) presented the best results.
In fractured aquifers, a research performed by the U.S Geological Survey (USGS) by Risser et al. (2005) in Pennsylvania state (USA) showed that the estimation using water table fluctuation should be used with caution due to the aquifer media anisotropy.The baseflow separation presented the most consistent results for the study area.
Moreover, Scalon et al. (2002) found that the choice for the appropriate method usually is a hard task.The most important considerations to be taken include the time/spatial scale.The aim of the study is also an important factor and can determine the scale.The most common studies involve groundwater quantification for management purposes, which requires large time/spatial scales in the analysis; and the evaluation of aquifer vulnerability, which requires detailed spatial scales.In this paper, baseflow separation was used for estimating groundwater recharge on Serra Geral Aquifer System (SASG) in Paraná through digital stream-flow filtering (Eckhardt's Filter).The obtained recharge values can be used as reference for the region for groundwater management and planning.

STUDY AREA DESCRIPTION AND DATA TIME-SERIES
The study area regards to the outcrop of the Serra Geral Formation in the Paraná state, which is represented by volcanic rocks resulted from Mesozoic magmatism, covering approximately 75% of the Sedimentary Basin of Paraná.In Brazil, its outcrop extends from the states of Rio Grande do Sul to Minas Gerais, as shown in Figure 1, but it also covers territorial portions of Argentina, Paraguay and Uruguay.The Serra Geral Magmatism consists in a succession of volcanic and the predominant lithotypes presents a basic composition, with interposed acid tufts (rhyolites and riodacites) (LICHT, 2013).
Regarding the hydrogeologic aspects, the outcrop area of SASG in Brazilian territory is approximately 800.000 km² (56% of its occurrence area) and the rest is covered by sediments of the Bauru and Caiuá Groups.In the Paraná state, SASG covers an area of approximately 109.000 km² and its thickness reaches 1347 meters in Cianorte -PR.It overlay the Guarani Aquifer System, which comprises sandstones of Botucatu and Piramboia Formations.In Paraná it is overlain by Caiuá Aquifer in the northwest region of the state (ATHAYDE and ATHAYDE, 2015).This aquifer has great importance for water supply on Paraná State.According to Sanepar (2015), this aquifer contributes with 57% of the total groundwater volume supplied by the company.In Paraná, 56% of the cities are exclusively supplied by groundwater and 22% by mixed systems (ANA, 2010).
According to Athayde (2012), the limits of the hydrogeological basins coincide with the hydrographic basins.The groundwater flow, in regional scale, occurs from east to west, towards the discharge regions (Paraná and Paranapanema rivers).Locally, the flow occurs towards the main drainages in the state, such as Iguaçu, Ivaí, Piquiri and Tibagi rivers.
The hydrography of the study area is represented by the basins that covers the outcrop area of SASG in Paraná, which are: Iguaçu, Piquiri, Ivaí, Pirapó, Tibagi and Cinzas.The area is limited by Rio Paranapanema in north and Rio Paraná in west.
Figure 2 presents the physical characteristics of the study area.The hydrography map presents the watersheds localization, the main rivers and the gauge stations used.The elevation map was obtained from SRTM (Shuttle Radar Topography Mission) and the model was adapted to the Brazilian reference system by Weber et al. ( 2004) and as a 90m cell resolution.The declivity map was elaborated from this model and the class intervals are defined according to the Embrapa (1979) classification.The soil type map was obtained from the Paraná Soil Map (scale 1:600,000) by Embrapa (2009).The most common soils are latosols, neosols and nitisols.Regarding the infiltration capacity of these soils, CPRM (2014) classify latosol as very good capacity; neosol as moderate capacity; and nitisol as good capacity.The physical characteristics of each basin are summarized on Table 2.
Streamflow time-series data was obtained from Hidroweb of Brazilian National Water Agency (ANA), except for the Iguaçu River, which naturalized flow data in the Salto Caxias hydroelectric power plant was obtained from National Electrical System Operator (ONS).The gauge stations are listed and described in Table 1.(2) In Eckhardt's digital filter, the A and B parameters are defined by: Therefore, the equation can be expressed as: where a is a parameter determined by the recession analysis and BFImax is related to the baseflow and total flow ratio.It was obtained from the analysis of several hydrograph recessions segments and an average value was calculated for each river.
The definition of BFImax is based on the local geology and streamflow nature (perennial or ephemeral).The values were pre-defined at 0.8 for perennial rivers with porous aquifer; 0.5 for ephemeral rivers with porous aquifer; and 0.25 for perennial rivers with fractured aquifer.These values should be used as a first approximation (ECKHARDT, 2005).According to Collischonn and Fan (2013), the greatest inconvenience presented by Eckhardt Filter consists of the difficulty of estimating BFImax, because this parameter is based on the predominant geological features of the watershed.Therefore, this value is hard to stablish due to heterogeneity of the aquifer types in a basin.Regarding this limitation, the mentioned authors proposed two different ways to estimate BFImax based on inverse filter application and in the Q90/Q50 ¬ratio.
The first proposal (IF), considers that, over long-time periods without rainfall, streamflow is maintained by baseflow and the baseflow in aquifer is linearly proportional to its storage.Therefore, reorganizing the following equation, it is possible to obtain the baseflow in a previous time step (bi-1), which would result in a present value of baseflow, given a value for the constant a.
This equation can be transformed into a backward filter which can be applied in the hydrograph in order to obtain a preliminary maximum possible flow value.
Considering that the hydrograph generated by b' presents the maximum possible value of baseflow given a recession parameter a, the BFImax estimation can be carried out by dividing the sum of b' by the sum of total flow y, according to the following equation: The authors applied the proposed methodology to estimate BFImax for daily flow data in fifteen gauge stations in South of Brazil.Results showed that the obtained values were similar to those suggested by Eckhardt (2005), considering the predominant hydrogeology in each basin.
The second proposal (QR) consists of estimating BFImax based on the ratio between flows with permanence of 90% and 50%.The authors stablished the following relation between BFImax and Q90/Q50: BFImax = 0.8344 The BFI¬max values obtained by IF and QR were compared to Eckhardt's pre-defined values with a ponderation (EP).Taking into account that all the rivers are perennial, it was considered that the values should range from 0.25 (100% of the area covered by fractured aquifer) to 0.8 (100% of area covered by sedimentary aquifer) according to the percentage of aquifer type in the basin (fractured or sedimentary).Intermediary values are observed when at the same watershed there is occurrence of both aquifer types.The different obtained values of BFImax were compared.
Recharge rates in mm/year were calculated and the obtained values were compared to average rainfall obtained from annual isohyets (time-series from 1977 to 2006) provided by Pinto et al. (2011) and physical parameters presented on topic 2.1.

RESULTS AND DISCUSSION
Table 3 presents the values of calculated BFImax for each watershed.According to Eckhardt (2005), basins in which fractured aquifers are predominant, BFImax should be around 0.25 and 0.75 for sedimentary.On Piquiri and Pirapó basins, fractured aquifers are predominant, but BFImax ¬calculated through IF and QR presents high values, close to the pre-defined values for sedimentary aquifer.This may be associated to the flow conditions of the SASG, which is a basaltic-rock aquifer.A large amount of opened fractures along with secondary porosity caused by cooling features, such as columnar joints, allow the vertical flow through the basalt, which favors the aquifer recharge (MILLER, 1999).Also, the horizontal groundwater flow in volcanic-rock aquifers is favored by horizontal discontinuity, represented by tops and bottoms of lava flows and interflow zones (SPILLER, 2005).Collischonn and Fan (2013) calculated BFImax values for several gauge stations in Brazilian rivers.Those located on predominant volcanic-rock aquifers presented higher BFImax values than those on crystalline-rock aquifers.Besides, according to the Hydrogeological Map of Brazil (CPRM, 2014b), the water productiv ity in SASG is higher than in others fractured aquifers on Brazilian territory, which can be associated to its higher recharge rates.Furthermore, Dora (2013) found that the Eckhardt's filter using pre-defined values underestimated recharge values on basaltic aquifer basins on Rio Grande do Sul state.
Figure 3 presents a graphic representation of linearized BFImax values with respect to the fractured aquifer area percent.It is possible to observe that higher occurrence of SASG in the basin results in a higher difference of IF and QR methods with respect to EP.Also, the BFImax values for IF and QR methods increases when the fractured aquifer area is higher.This suggests that the BFImax values for volcanic rock aquifers are higher than crystalline rock aquifers.
The graphic analysis showed that IF and QR methods presented positive correlation and similar values.These methods have the advantage of not needing pre-defined values, which can be inaccurate due to aquifer geology heterogeneity.The annual recharge rates considering the three methods for calculating BFImax are presented on the Table 4.The graph in Figure 4 presents a comparison between the recharge values calculated by baseflow separation.For a better understanding of recharge rates, the values were compared to physical characteristics information of each watershed presented on Table 2.  rates.These values are related to low rainfall rates, higher declivity areas the soil type.
Whereas recharge rates in mm/year is a function of rainfall and physical characteristics in a watershed, recharge values in percent of rainfall is entirely related to physical characteristics only.Therefore, the information in Table 2 can be compared to recharge in percent of rainfall in Table 4.It is not possible to stablish a relationship between recharge and fractured aquifer area percent, which means that in this study area, the aquifer type (fractured or sedimentary) may not be the most important factor in determining the recharge rates in a basin.On the other hand, declivity and soil type appear to influence the obtained results.Higher recharge rates in percent of rainfall is related chiefly to lower declivity values and latosol.The advantage in using baseflow separation is that it needs only one parameter, which is average daily flow.However, it is not possible to perform a spatial distribution of the results if there is not enough sub-watersheds stream-flow data.The existence of several fluviometric gauge stations distributed throughout watersheds is not a reality in Brazil and many other countries.Besides, it is not possible to estimate recharge for periods shorter than a year because the baseflow in each month does not refer to the corresponding recharge on that month, but to previous recharge events, since the baseflow occurs slowly, conditioned by the rock pores and fractures.
Average recharge rates were calculated by each method as an approximation for the region.These values are presented on Table 5.The three methods presented similar average values, but IF and QR presented slightly higher values than the EP values.
The lower value presented by EP method is a result of the underestimation of the pre-defined values for volcanic-rock aquifers.

Figure 1 -
Figure 1 -Localization of the study area and Paraná Basin occurrence in Brazil

Figure 2 -
Figure 2 -Physical characteristics of the study area: hydrography (a), elevation (b), slope (c) and soil type (d)

Figure 3 -
Figure 3 -Linearized BFImax values for each method

For a better
representation of the baseflow separation results, Figures 6 and 7 presents hydrographs for 2007 of the Laranjinha and Pirapó Rivers, which have different flow behavior.In the Laranjinha River hydrograph, the flow undergoes a fast response to rainfall events and presents several flow peaks throughout the year and a low baseflow.The average flow calculated in Laranjinha River for the historical series of 2000-2008 is 61 m³/s and the baseflow calculated by Eckhardt's Filter using Inverse Filter is 17 m³/s.In Pirapó basin, the river flow (57 m³/s average) is sustained mostly by the baseflow (41 m³/s average).

Table 3 -
BFImax values calculated for each basin

Table 4
presents recharge rates values in mm/year, obtained by dividing baseflow by the watershed area, and in percent of annual rainfall, obtained by dividing annual recharge by annual rainfall.

Table 4 -
Annual recharge rates for each basin The highest annual recharge rates calculated by baseflow separation refers to Piquiri River basin.These values can be explained by the high rainfall rates, low declivity and high infiltration rates soils.On the other hand, Cinzas River basin (which includes Laranjinha River basin) presented the lowest recharge