2.1. Study Area
The Jequitiba River basin (area: 57,148 he) is located in the central region of Minas Gerais state (Brazil;
Figure 2), and is characterized by rural environment, except in Sete Lagoas town, which is densely urbanized and populated. The study area comprises the Marinheiro catchment (1480 he), which a small tributary headwater catchment of the Jequitiba River basin located in the neighborhood of Sete Lagoas. The Jequitiba basin is located between the geographic coordinates X = 573,198 to 594,872 m and Y = 7,859,607 to 7,836,875 m, referred to the SIRGAS 2000 geodetic datum and UTM 23 South projection, while the Marinheiro catchment fits within the coordinates X = 581,100.3 to 587,493.5 m and Y = 7,841,747.5 to 7,847,420.3 m.
According to Koppen’s classification, the climate is subtropical (Cwa), characterized by dry winters and hot summers. The mean annual rainfall was 1291.2 mm in the 2000–2018 period, while the mean temperatures ranged between 18 ºC and 24 ºC from July to January–February, and showed a mean annual value of 21.8 °C. The altitudes range from 629 to 932 m in the Jequitiba River basin and between 717 and 918 m in the Marinheiro catchment (
Figure 3a). The geologic substratum (
Figure 3b) of the Jequitiba River basin comprises an Archean crystalline basement composed of orthogneisses, granites, and migmatites (24.3%) that were overlaid by Neoproterozoic carbonate rocks of the Bambuí Group, namely calcite and dolomite limestones from the Sete Lagoas Formation (42.7%), and pelitic rocks with interlayered carbonates from the Serra de Santa Helena Formation (28.7%). Later, this sequence was covered by alluvium, colluvium, and terrace sediments (4.4%) along and lateral to the main water courses [
32,
33]. The Marinheiro catchment is dominated by carbonate rocks from the Sete Lagoas formation, with a small area of Archean rocks cropping out around the catchment headwaters. The comparison of
Figure 3b with
Figure 3c (soil map [
34]) reveals that Archean rocks as well as pelitic rocks cropping out in the catchment lowlands are overlaid with cambisols (36.7% coverage within the Jequitiba River basin), pelitic rocks cropping out in the catchment highlands are overlaid with neosols (12.2%), and limestones and terrigenous rocks are overlaid with latosols (49.6%). Accordingly, the Marinheiro catchment is mostly covered with latosols and cambisols (along the headwaters). Land in the Jequitiba River basin is mostly used for anthropogenic activities, such as livestock pasturing or agriculture, which are distributed along the drainage network (
Figure 3d). Forests occupy 15.8% of the area, and urban areas 14.4% [
35]. Agriculture is also dominant in the Marinheiro catchment, but native vegetation is kept dense in some areas, especially along the headwaters.
The hydrogeology is characterized by fractured aquifers, developed in the Archean rocks. Fractured-karst aquifers developed in the pelitic rocks interlayered with carbonates from the Serra de Santa Helena formation, karst aquifers developed in the limestones from the Sete Lagoas formation, and porous aquifers developed in the terrigenous rocks and soil layer [
36]. Recently, Monteiro et al. [
37] estimated the recharge within the Jequitiba River basin (
Figure 4) and derived average values for the various aquifer types (m
3 ha
−1 year
−1): 829.3–1357.0 (porous), 829.3 (karst and fractured-karst), and 539.3 (fractured crystalline). According to
Figure 4, average recharge within the Marinheiro catchment is 340 m
3 ha
−1 year
−1 within the fractured aquifer (≈ 30% of the catchment) and 1883 m
3 ha
−1 year
−1 within the karst aquifer (≈ 70%), which means 1420 m
3 ha
−1 year
−1 on average.
The town of Sete Lagoas has been supplied with groundwater from the karst aquifer since the 1950s [
29], when the population was less than 25,000 inhabitants and groundwater was extracted from cisterns. These sites became obsolete as the population grew, being progressively replaced by drilled wells to satisfy the public water demand. In 1993, 65 drilled wells were pumped every day for 16 h, extracting 8.0 L s
−1 each to satisfy the consumption of 150,000 people living in Sete Lagoas [
29,
38]. The situation was re-evaluated in 2008 by Botelho [
30], who counted 94 drilled wells (44% increase), keeping a similar mean abstraction (7.8 L s
−1) to satisfy the consumption of 220,000 habitants (47% increase).
2.3. Hydrologic Modeling
The hydrologic modeling was developed in the following four main steps (
Figure 5): (1) pre-processing of climatic, stream flow, and cartographic data, followed by delineation of hydrologic response units as detailed below, based on the overlay analysis of relief, soils, and land use maps. The data records comprised the 2003–2016 period; (2) hydrologic modeling of a catchment monitored for stream discharge, namely a sub-basin of Jequitiba River basin monitored at the hydrometric station 41410000. The modeling was based on the JAMS J2000 framework developed by Krause [
64] and comprehended a calibration (2003–2011) and a validation (2012–2016) period. This step was preceded by parameterization of input data and followed by the performance analysis of calibrated models based on goodness-of-fit indices; (3) “Downscaling” of stream flow discharges to the scale of an ungauged catchment enclosed in the monitored sub-basin, namely the Marinheiro catchment (study area). The “downscaling” was based on a second run of JAMS J2000, keeping the previous parameterization but adjusting the hydrologic analysis to the area, associated hydrologic response units, and climatic characterization of the Marinheiro catchment. The “downscaling” step was validated through comparison of discharge rates simulated by the JAMS J2000 framework and rates assessed in the field at the catchment outlet using a portable flow meter, as well as on the basis of a goodness-of-fit index; (4) Evaluation of reference and ecological stream flows aiming the assessment of surface water resources within the Marinheiro catchment. The four steps are detailed in the next paragraphs.
In the first step, the variables precipitation, temperature, relative humidity, hours of sunshine, wind speed, and daily stream flow data, compiled from the weather and hydrometric stations, were organized in a series of Excel spreadsheets, namely listed in ascending order of time. In turn, these spreadsheets were submitted to the online platform INTECRAL RBIS—river Basin Information System (
http://leutra.geogr.uni-jena.de/intecralRBIS)—developed by Zander and Kralisch [
65], for the conversion of the Excel spreadsheets into files in the specific format of the JAMS J2000 framework.
The cartographic data were cropped to the area of interest and verified for geometric and geographic consistency. In addition, the satellite images were submitted to supervised classification of land use or cover using the SPRING software. The classification comprised classes for cultivated area, water bodies, pastures, forest and exposed soil.
The online platform HRU-WEB (
http://intecral.uni-jena.de/hruweb-qs/) was used to delineate the homogeneous hydrologic response units (HRUs). The HRUs are used as modeling entities that have the same pedological, lithological, topographical, and land use/land cover characterizations and are heterogeneous from each other. They are connected by a topological routing scheme [
66]. The lateral water flow is simulated allowing a fully explicit spatial discretization of hydrologic response within the modeled catchment. However, the delineation of HRUs may not account for karst heterogeneity. Nevertheless, the model can be calibrated so that the observed and modeled stream flows match well. This HRU-WEB platform is embedded with tools to intersect the digital elevation model with soil, geology and land use and cover maps producing the HRUs as output. Each HRU holds a specific identification number, centroid coordinates, and connection codes to adjacent HRUs or specific water lines [
67].
In the second step, the JAMS J2000 framework model was applied to simulate hydrological processes at the scale of a sub-basin of the Jequitiba River basin located upstream of hydrometric station 41410000, which encloses the Marinheiro catchment. The J2000 is a process-based spatially explicit hydrological model, which simulates eco-hydrological processes on the river basin scale. The method is a part of a modular and object-oriented modeling system called JAMS. Altogether, the JAMS J2000 framework contains different environmental models and a plethora of components to create custom tailored models for different research questions. In addition, JAMS possesses modules for data preparation, analysis, and visualization. The full inventory of hydrologic modules used in this study is listed in
Appendix A. It is beyond the scope of this paper to describe them all in detail, especially for their mechanics. Readers are invited to consult the original works.
The JAMS J2000 framework simulated discharge rates, on a daily time step, between the calendar years 2003 and 2016. Using water balances and routing algorithms, the model also simulated flow components (surface flow, percolation in soil layers, shallow and deep underground flow). The simulated discharge rates were compared with the homologous rates measured at the hydrometric station for the sake of calibration and validation. Calibration encompassed the 2003–2011 period while validation the 2012–2016 period. Besides the measured discharge rates, other parameters were used in the hydrologic modeling, which are listed in
Table A2,
Table A3 and
Table A4 of
Appendix B. These parameters were processed in the various JAMS J2000 modules (
Appendix A), which were run until optimized values (calibrated) were obtained. The calibration procedure was based on the NSIN II algorithm (Genetic Multi-objective II) with daily time step, while adopting 5000 iterations as the stopping rule [
68]. The calibrated parameters are indicated in
Appendix C. Following the calibration procedure, the hydrological model was tested for performance through assessment of goodness-of-fit indices (a) percentage of bias (PBIAS) [
69] and (b) Nash-Sutcliffe (NSE) efficiency coefficient [
70]. The performance intervals are provided as
Appendix D. According to Gupta et al. [
69], the PBIAS estimates the percentage trend of simulated data to be higher or lower than the observed data and can be described by the following equation:
where PBIAS is the percentage of bias (%), y
i is the simulated flow (m
3 s
−1), and o
i is the observed flow (m
3 s
−1). A PBIAS = 0 occurs for a hydrological model with optimal performance. Positive or negative values indicate, respectively, that the model overestimates or underestimates the simulated flows. The NSE and LNSE coefficients are equated as follows [
70]:
where “ln” represents the natural logarithm and o
i, o
med, y
i, and y
med represent, respectively, the observed flow, the average observed flow, the simulated flow, and the average simulated flow (m
3 s
−1). The values of NSE and LNSE (dimensionless) can vary from –∞ to 1. The closer to 1, the greater the adjustment between the simulated and observed values. Results below 0 indicate that the mean observed values are more representative than the values predicted by the model.
In the third step, the calibrated stream flows (2003–2016 series) were “downscaled” to the Marinheiro catchment scale and estimated at the corresponding outlet. To accomplish the “downscaling,” the JAMS J2000 framework was run using the previously calibrated parameters (
Appendix C) and the HRUs enclosed in the Marinheiro catchment. Then, the discharge rates estimated at the Marinheiro catchment’s outlet using the JAMS J2000 framework were compared with homologous rates estimated on the basis of flow velocities and water column heights, measured with a portable flow meter in the period June of 2015 and February of 2018 (142 evaluations).
where
Q (m
3 s
−1) is the discharge rate,
A (m
2) is the wet area of a circular stream section and
v (m s
−1) is the measured flow velocity. The value of
A was estimated as follows:
with
where
h (m) is the water column height measured at the stream section’s center, while
D (m) and θ (radians) are the stream section’s diameter and the angle between the stream section central point and the stream water surface, respectively.
The performance of “downscaling” was checked through comparison of measured and simulated hydrographs and calculation of the RMSE estimator.
where
Oi and
Ei represent the observed (measured
Q) and estimated (with JAMS J2000 framework) values of a target variable, while
n represents the number of measurements and corresponding estimations (142 in the present case).
In the fourth (and last) step, reference stream flows were evaluated within the Marinheiro catchment, based on the simulated series of discharge (JAMS J2000 framework, period 2003–2016). A reference stream flow launches the upper limit of water uses in a water course. According to Harris et al. [
71], the legal application of a reference flow favors the protection of rivers, because the grants for stream water diversion in that context are based on low risk base flows. However, the setup of reference flows is also an obstacle to the implementation of a grant system [
72]. The most common reference values are [
73] Q
7,10, which is the minimum flow of seven days duration and 10 years return period, with a 10% risk of not being reached, and Q
90, which corresponds to a flow with 90% probability of being exceeded in time. They are both seen as ecological flows, meaning that any granted stream water diversion needs to permanently ensure these or larger flows in the target river [
74,
75]. In Brazil, states have adopted different criteria for setting up reference flows for granting but did not present justifications for the adoption of specific values. In Minas Gerais state, the Mining Institute of Water Management (IGAM;
http://www.igam.mg.gov.br/) defined threshold values for the Marinheiro catchment, namely Q
7,10 = 0.029 m
3 s
−1 and Q90 = 0.075 m
3 s
−1. The comparison of IGAM values with counterparts determined by the JAMS J2000 framework represents additional performance criterion.
Besides the use in model performance assessment, the reference stream flows were used in the evaluation of water resources aiming the proposed conjunctive use of karst groundwater and surface water in the supply to Sete Lagoas town. Their calculation involved the definition of a flow duration curve for the Marinheiro stream that was based on a probabilistic model [
76,
77],
where
pi is the probability of reaching or exceeding a stream flow,
i is the order of the
ith sorted flow (descending order), and
n is the number of ordered data. It also involved the definition of maximum and minimum flow values at different return periods (2, 5, 10, 50, and 100 years) using the log-normal probability distribution [
78].