Study area
The Cachoeira River Basin is located in southern Bahia (Fig. 1). It covers an area of around 4600 km² and had a mean flow rate of 21.26 m³/s between 1980 and 2015. The CRB is limited to the north by the Contas and Almada river basins; to the south, by the Pardo and Una river basins; to the west, by the Pardo River Basin; and to the east, by the Atlantic Ocean, which has altitudes that vary between 0 and 1300 m above sea level.
The CRB has two defined climate types: Tropical Rainforest Climate (Af) in its low and medium course, with a maximum annual rainfall of 2000 mm, and Sub-humid (Aw), with a mean annual rainfall of 800 mm (Silva 2016). The Tropical Rainforest Climate (Af) is characterized by the absence of a dry season, with precipitation greater than 60 mm in the driest month and a greater volume of rain between March and August, and temperatures between 24 and 25°C in the warmer months (January and February). Notably, in this unit, mean temperatures in the coldest month are above 18°C. The Sub-humid type (Aw) is characterized by a tropical climate with a dry winter, a rainy season in the summer (November to April), and a clearly defined dry season in the winter (May to October), with July as the driest month (Sousa et al. 2020).
Regarding the types of soil found in the CRB, Dystrophic Red-Yellow Argisol and Haplic Chernosol are predominant, while Red-Yellow Argisol, Dystrophic Red-Yellow Latosol, and Dystrophic Yellow Latosol appear in smaller proportions. Haplic Chernosol, which is found in most of the basin's territory, is characteristically a shallow soil consisting of mineral material from gneiss rocks from the Proterozoic era (Embrapa 2002). Red-Yellow Argisol, on the other hand, is developed from the Barrier Group of crystalline rocks, with a sandy structure on the surface and great textural change towards the B horizon, considered quite clayey (Embrapa 2006; Nacif 2003).
Figure 1
According to the Koppen classification (1948), there are two types of climates in the BHRC: tropical rainforest (Af) in the low and medium course, with annual rainfall of up to 2,000 mm, and tropical savanna (Aw) with annual rainfall of around 800 mm all year round (Silva 2016). Geologically, the basin is mainly composed of igneous rocks and is located in the alkaline province of southern Bahia, which consists of a complex of massive alkaline rocks in the southern part of the Itabuna-Salvador-Curaçá orogen (Rosa et al. 2003; Cerqueira et al. 2019).
Among the soil types found in the territory of the BHRC, dystrophic Red-Yellow Argisol and Haplic Chernosol stand out for being in greater proportions, while Red-Yellow Argisol, dystrophic Red-Yellow Latosol, and dystrophic Yellow Latosol can be found in smaller proportions. The Haplic Chernosol, found in most of the basin's territory, has shallow soil characteristics and consists of mineral material from Proterozoic gneiss rocks (Embrapa, 2002). The Red-Yellow Argisol is developed from the Barreiras group of crystalline rocks and has a sandy structure on the surface and a significant texture change to the B horizon, which is considered very clayey (Embrapa, 2002; Nacif, 2000).
Hydrological modeling
The Soil and Water Assessment Tool (SWAT) is an open-source hydrological model developed by the Agricultural Research Service of the United States Department of Agriculture (USDA-ARS) in the 1980s. It was designed to predict the impact of land management practices on water production, sediment, and agricultural chemicals in large watersheds with varying conditions over long periods. In the present model, the sub-basins are divided based on the Digital Elevation Model (DEM) and then subdivided into hydrological response units (HRUs) according to the topography, land use, and soil types of the study area (Chen et al., 2019).
The model's simulation capability has been expanded for years, which has allowed it to predict environmental impacts that were previously difficult to assess. Thus, the model presents numerous possibilities to evaluate scenarios both quantitatively and qualitatively (Bai Shen and Yan 2016). The model equation that describes the water balance is the mass conservation principle described in Eq. 1.
$${SW}_{t}={SW}_{0\left(i\right)}+\sum _{i=1}^{t}{(R}_{dia\left(i\right)}-{Q}_{\text{sup}\left(i\right)}-{E}_{a\left(i\right)}-{W}_{\text{solo}\left(i\right)}-{Q}_{\text{g}\text{w}\left(\text{i}\right)})$$
1
Where: \({SW}_{t}\) is the final water depth in the soil, mm; \({SW}_{0\left(i\right)}\) is the initial water depth in the soil on day \(i\), mm; \(t\) is time, days; \({R}_{day\left(i\right)}\) is the rainfall on day \(i\), mm; \({Q}_{\text{sup}\left(i\right)}\) is the surface runoff for day \(i\), mm; \({E}_{a\left(i\right)}\) is the evapotranspired water depth for day \(i\), mm; \({W}_{soil\left(i\right)}\) is the water depth entering the soil unsaturated zones on day \(i\), mm e \({Q}_{gw\left(i\right)}\) is the portion of channel flow from baseflow for day \(i\), mm (Neitsch, 2005).
Model dataset and configuration
For the present study, 35 years of data (1980–2015) were used as the analysis period. Daily precipitation data in the basin's area of influence were obtained from the rainfall stations of the Agência Nacional das Águas (ANA) and from the grids developed by Xavier (2015), which interpolated climate variables through grids to cover the largest area of influence of the basin. Data on solar radiation, maximum and minimum temperatures, relative air humidity, and wind speed were obtained entirely from grids by Xavier (2015). The historical flow series observed came from station No. 53180000 (bordering the BR-101) located in the municipality of Itabuna, which is a point of flow control of the basin, and obtained through the HidroWeb portal linked to ANA.
The soil classes were identified based on the soil map generated by the Superintendência de Estudos Econômicos e Sociais da Bahia (SEI) at a scale of 1:100000 from the Brazilian Soil Classification System of EMBRAPA (2006). The following 5 soil classes are found in the CRB and make up its soil cover: Dystrophic Red-Yellow Argisol (36.65%); Dystrophic Red-Yellow Argisol (10.53%); Dystrophic Red-Yellow Latosol (2.20%); Dystrophic Yellow Latosol (0.33%); Haplic Chernosol (50.29%). However, to simplify the input data in the model, all types of Argisol and Latosol were transformed into the same unit, thus leaving three types of soil classes in the basin to be modeled.
Soils were classified based on Sartori (2005) and Pereira (2013). Data referring to soils were entered into the database used by the model and consisted of the information shown in Table 1.
The CRB was delimited from the Shuttle Radar Topography Mission (SRTM) with a resolution of 30 m obtained in the USGS by the ArcGIS 10.2 software. For the study, the delimitation method was based on the following 4 steps: a) eliminating small imperfections, such as depressions and peaks; b) defining the flow directions of rainwater; c) generating the accumulated flow; and lastly, d) simultaneous processing of data from steps "b" and "c" through the Watershed tool.
The land use maps used in the study came from the MapBiomas-Coleção 5.0 Project. This project is part of a collaborative network formed by several NGOs, universities, and technology companies to map biomes and land uses in Brazil using a geographic information system (“SIG”) and remote sensing tools. For the present study, land uses from 1985, 1995, 2005, and 2015 were used to follow the evolution of use over time.
Table 1
Physical and chemical attributes of SWAT input of the different soils in the basin
Parameters
|
Argisol
|
Chernosol
|
Latosol
|
Layers
|
5
|
2
|
5
|
Hydrological Group
|
B
|
D
|
A
|
SOL_ZMX
|
1500
|
2000
|
3500
|
ANION_EXCL
|
0.47
|
0.48
|
0.50
|
C1
|
SOL_Z
|
300
|
700
|
300
|
SOL_BD
|
1.40
|
1.37
|
1.20
|
SOL_AWC
|
0.14
|
0.1
|
0.25
|
SOL_K
|
40
|
120
|
100
|
SOL_CBN
|
1.40
|
1.03
|
0.95
|
CLAY
|
19.40
|
15
|
42
|
SILT
|
9.50
|
25.50
|
26
|
SAND
|
71.70
|
59.50
|
32
|
ROCK
|
0
|
0
|
0
|
SOIL_ALB
|
0.15
|
0.1
|
0.15
|
USLE_K
|
0.05
|
0.12
|
0.13
|
C2
|
SOL_Z
|
600
|
2000
|
550
|
SOL_BD
|
1.40
|
1.49
|
1.5
|
SOL_AWC
|
0.18
|
0.09
|
0.20
|
SOL_K
|
30
|
30
|
90
|
SOL_CBN
|
0.60
|
0.60
|
0.35
|
CLAY
|
34
|
50.1
|
40
|
SILT
|
35.70
|
18.80
|
28.70
|
SAND
|
30.30
|
31.10
|
31.30
|
ROCK
|
-
|
-
|
-
|
SOIL_ALB
|
0.15
|
0.15
|
0.15
|
USLE_K
|
0.16
|
0.15
|
0.12
|
C3
|
SOL_Z
|
800
|
-
|
900
|
SOL_BD
|
1.40
|
-
|
1.70
|
SOL_AWC
|
0.18
|
-
|
0.20
|
SOL_K
|
20
|
-
|
52
|
SOL_CBN
|
0.40
|
-
|
0.40
|
CLAY
|
27
|
-
|
28.40
|
SILT
|
39.30
|
-
|
37.7
|
SAND
|
33.70
|
-
|
33.9
|
ROCK
|
0
|
-
|
0
|
SOIL_ALB
|
0.12
|
-
|
0.15
|
USLE_K
|
0.11
|
-
|
0.14
|
C4
|
SOL_Z
|
1100
|
-
|
1250
|
SOL_BD
|
1.80
|
-
|
1.50
|
SOL_AWC
|
0.14
|
-
|
0.20
|
SOL_K
|
12.50
|
-
|
40
|
SOL_CBN
|
0.40
|
-
|
0.35
|
CLAY
|
26.80
|
-
|
29.80
|
SILT
|
43.70
|
-
|
13.50
|
SAND
|
29.50
|
-
|
56.70
|
ROCK
|
-
|
-
|
-
|
SOIL_ALB
|
0.15
|
-
|
0.15
|
USLE_K
|
0.18
|
-
|
0.14
|
C5
|
SOL_Z
|
1500
|
-
|
3500
|
SOL_BD
|
1.40
|
-
|
1.50
|
SOL_AWC
|
0.14
|
-
|
0.10
|
SOL_K
|
12
|
-
|
15
|
SOL_CBN
|
0.30
|
-
|
0.20
|
CLAY
|
23.80
|
-
|
27
|
SILT
|
29.8
|
-
|
39.30
|
SAND
|
46.40
|
-
|
33.70
|
ROCK
|
-
|
-
|
-
|
SOIL_ALB
|
0.15
|
-
|
0.13
|
USLE_K
|
0.16
|
-
|
0.12
|
Identification of parameters, calibration, and validation
Experiments and data in the literature were used to identify the most significant parameters of the SWAT model (Kavian, Golshan, & Abdollahi 2017; Li et al. 2018; Woldesenbet et al. 2017; Santos et al. 2018; de Sousa et al. 2020). Thus, 15 parameters were selected for daily and monthly calibration. These parameters encompass the processes of groundwater, evapotranspiration and lateral flow, soil moisture, surface runoff, and concentration time.
The selected parameters were manually calibrated using the observed daily flow. Firstly, the historical flow series, from 1980 to 2015, was divided into four datasets to adapt the flows to the land uses of each period, as shown in Table 2. The 1st and 2nd datasets were used for calibration, while the 3rd and 4th datasets of the series were used for hydrological model validation.
Table 2
Organization of data in the calibration and validation process
Parts
|
Flow
|
Land use
|
1ª
|
1980–1988
|
1985
|
2ª
|
1989–1997
|
1995
|
3ª
|
1998–2006
|
2005
|
4ª
|
2007–2015
|
2015
|
The results obtained by the calibration and validation processes were statistically evaluated using the coefficient of determination (Eq. 2), the PBIAS index (Eq. 3), and the Nash-Sutcliffe efficiency (Eq. 4).
$${R}^{2}=1-\frac{\sum _{i=1}^{n}({S}_{i}-{\widehat{S}}_{i})²}{\sum _{i=1}^{n}({S}_{i}-{\stackrel{-}{S}}_{i})²}$$
2
$$PBIAS=\left|\frac{\sum _{i=1}^{n}({S}_{i}-{O}_{i})}{\sum _{i=1}^{n}{O}_{i}}\right|. 100$$
3
$$NSE=1-\frac{\sum _{i=1}^{n}({O}_{i}-{S}_{i})²}{\sum _{i=1}^{n}({O}_{i}-\stackrel{-}{O})²}$$
4
where:
n = number of observations during the simulated period;
\({O}_{i}\) = observed flows;
\({S}_{i}\) = flows simulated by the model;
\(\stackrel{-}{O}\) = mean of observed values;
\(\stackrel{-}{S}\) = mean of simulated values;
\({\widehat{S}}_{i}\) = obtained by linear regression between observed and estimated data.
Table 3
Classification of coefficients for evaluating the results of the SWAT model on a daily and monthly scale
Classification
|
NSE
|
PBIAS (%)*
|
R²
|
Daily
|
Monthly
|
Daily
|
Monthly
|
Very Good
|
0.80 < NSE ≤ 1.00
|
0.85 < NSE ≤ 1.00
|
PBIAS < ± 3
|
0.85 < R²≤1.00
|
0.85 < R²≤1.00
|
Good
|
0.70 < NSE ≤ 0.80
|
0.70 < NSE ≤ 0.85
|
± 10 ≤ PBIAS < ± 3
|
0.70 < R²≤0.85
|
0.80 < R²≤0.85
|
Satisfactory
|
0.50 < NSE ≤ 0.70
|
0.55 < NSE ≤ 0.70
|
± 15 ≤ PBIAS < ± 10
|
0.50 < R²≤0.70
|
0.70 < R²≤0.80
|
unsatisfactory
|
NSE ≤ 0.5
|
NSE ≤ 0.55
|
PBIAS ≥ ± 15
|
R²≤50
|
R²≤70
|
* There is no difference between monthly and daily values for PBIAS |
Source: Adapted from Moriasi et al. (2015) |
The permanence curve was used during each part of the historical series configured in the study to complement model validation, evaluate the model’s ability to reproduce the hydrological processes that take place in the CRB, and compare the simulated and observed flow data at the control point.
Configuration of simulated scenarios
In order to distinguish the impacts of climate change and land use change in the CRB, the combination of historical flow series with land uses over time was used, forming 8 possible scenarios (Table 4) for the simulations. The meteorological data from 1980 to 2015, which totaled 35 years, were divided into 4 periods: 1980–1988 (C1), 1989–1997 (C2), 1998–2006 (C3), 2007–2015 (C4), along with the land use maps from 1985 (LU1), 1995 (LU2), 2005 (LU3), and 2015 (LU4).
Moreover, 8 scenarios were configured combining the 4 climate periods (C1-4) and the 4 land uses (LU1-4). The baseline scenario (ScB) was the time series generated based on the observed climate and land use. Scenarios Sc1 to Sc4 considered the constant climate during the 4 periods of the historical series combining the 4 different land uses analyzed. For scenarios Sc5 to Sc8, the 4 land uses were considered constant and the climate periods were varied.
Table 4
Configuration of proposed scenarios for the study
Base Scenario
|
Land Use Change Impact
|
Climate Variation Impact
|
Scb
|
Sc1
|
Sc2
|
Sc3
|
Sc4
|
Sc5
|
Sc6
|
Sc7
|
Sc8
|
C1_LU1
|
C1_LU1
|
C2_LU1
|
C3_LU1
|
C4_LU1
|
C1_LU1
|
C1_LU2
|
C1_LU3
|
C1_LU4
|
C2_LU2
|
C1_LU2
|
C2_LU2
|
C3_LU2
|
C4_LU2
|
C2_LU1
|
C2_LU2
|
C2_LU3
|
C2_LU4
|
C3_LU3
|
C1_LU3
|
C2_LU3
|
C3_LU3
|
C4_LU3
|
C3_LU1
|
C3_LU2
|
C3_LU3
|
C3_LU4
|
C4_LU4
|
C1_LU4
|
C2_LU4
|
C3_LU4
|
C4_LU4
|
C4_LU1
|
C4_LU2
|
C4_LU3
|
C4_LU4
|
Analyzed hydrological indices
The Indicators of Hydrologic Alteration – IHA software aims to identify hydrological changes in the basin over time. This software calculates the characteristics of hydrological regimes by evaluating 67 statistical parameters encompassing ecological flow components and indicators of hydrological change. These parameters are divided into 2 categories called Environmental Flow Components and Hydrological Change Index. The latter has 33 parameters classified into the following 3 classes: mean values, extreme values (floods), and minimum values (droughts) (López-Ballesteros et al. 2020).
In this study, IHA software 7th version was used, in which several parameters associated with extreme flows were considered. The hydrological year configuration started on January 1st and ended on December 31st. The flows that exceed 75% of daily occurrence were considered high, the flows below 25% of occurrence were considered low, and the flows below 10% of occurrence were considered extremely low for the period.
To evaluate the hydrological changes in the different proposed scenarios, 19 indices were used. These indices can be divided into four groups based on the magnitude, duration, and frequency of the flows generated from the simulations, as shown in Table 5.
Table 5
Parameters used in the present study separated into groups according to IHA software
Group
|
Caracteristc
|
Parameter
|
1. Magnitude of monthly water conditions
|
Magnitude
|
→ Median value for each calendar month.
|
2. Magnitude and duration of anual extreme water conditions
|
Magnitude and duration
|
→ Annual minima, 1-day mean;
→ Annual minima, 3-day means;
→ Annual minima, 7-day means;
→ Annual minima, 30-day means;
→ Annual minima, 90-day means;
→ Annual maxima, 1-day mean;
→ Annual maxima, 3-day means;
→ Annual maxima, 7-day means;
→ Annual maxima, 30-day means;
→ Annual maxima, 90-day means;
→ Base flow index.
|
4. Frequency and duration of high and low pulses
|
Magnitude, frequency and duration
|
→ Number of low pulses within each water year;
→ Mean or median duration of low pulses (days);
→ Number of high pulses within each water year;
→ Mean or median duration of high pulses (days).
|
5. Rate and frequency of water condition changes
|
Rates and reversals
|
→ Rise rates: Mean or median of all positive differences between consecutive daily values
→ Fall rates: Mean or median of all negative differences between consecutive daily values
→ Number of hydrologic reversals
|
Statistical analysis
The nonparametric Mann-Kendall trend and Sen’s slope statistical tests are widely used in hydroclimatic studies (Güçlü 2020, Hurtado, Zaninelli & Agosta 2020). The Mann-Kendall method, proposed by Mann and Kendall (Mann 1945; Kendall 1975), is recommended by the World Meteorological Organization and is used in practice to assess trend changes in hydroclimatic series. According to Chen et al. (2019), this trend test does not require the sample to have a normal distribution and involves a simple calculation.
Sen’s slope calculates the magnitude of trends in time series. According to Silva et al. (2015), this method is insensitive to outliers and absent values have more rigor than the linear regression curve, thus producing more real trend measures in the series.
Therefore, these tests were applied in the series of simulated flows both in the base scenario and in the 8 simulated scenarios for the basin to determine the trends in the hydrological indices in the evaluated scenarios.