Abstract
This study assessed the impacts of climate change on streamflow in the data-scarce Upper Ruvu River watershed (URRW). The Long Ashton Research Station Weather Generator (LARS-WG) was employed for generating the future ensemble-mean climate scenario based on six global circulation models (GCMs), under two representative concentration pathways (RCPs: RCP4.5 and RCP8.5). The future projections were made in two periods (2041–2060 and 2081–2100), and the baseline period (1951–1978) was used as a reference. The watershed hydrology was represented by the Hydrologic Engineering Center's Hydrologic Modeling System (HEC-HMS) model, which was calibrated and validated by using 5 and 4 years of streamflow data, respectively. Results indicate that the rainfall and minimum and maximum temperatures will increase in both periods, under both scenarios. This will potentially affect the streamflow that is projected to increase from March to August and decrease from September to February. The mean annual streamflow could potentially change from 48 m3/s in the baseline period to 45.6 and 56.5 m3/s during 2041–2060, and 52.4 and 67.4 m3/s during 2081–2100, under RCP4.5 and RCP8.5, respectively. The minimum and maximum streamflows are also predicted to change in both periods, under both scenarios. Considering these results, the climate change will have significant impacts on the streamflows of the URRW.
HIGHLIGHTS
The climate change impacts on streamflow were assessed in two future periods (2050s and 2080s) under RCP4.5 and RCP8.5.
The rainfall and temperature are projected to increase in both periods under RCP4.5 and RCP8.5.
The climate change will have significant impacts on the streamflow of the Upper Ruvu River watershed.
Graphical Abstract
INTRODUCTION
Climate change is one of the most delicate subjects in many catchments around the world (Niang et al. 2014). Increased temperature, rising sea levels, altered precipitation patterns, and changes in snow cover are all indicators of climate change (Ayalew et al. 2022). These changes are largely caused by the emission of greenhouse gases including carbon dioxide (CO2), methane (CH4), and nitrous oxide (N2O), which are linked to the usage of fossil fuels (IPCC 2014a, 2014b). In most cases, greenhouse gas emissions increase surface temperatures, which alter evapotranspiration and precipitation (Bates et al. 2008). According to several studies, hydrological systems are extremely susceptible to climatic changes (IPCC 2007; Bates et al. 2008; Zereini & Hötzl 2008; Toulmin 2009; Niang et al. 2014). Changes in temperature and precipitation affect catchment ecosystems, which in turn affect water availability and increase the frequency of hydrological extremes such as drought and floods (Rotich & Mulungu 2017; Shagega et al. 2019; Awotwi et al. 2021; Gurara et al. 2021).
Several studies have reported a decreasing rainfall trend in many parts of the Sub-Saharan region over the past years (Niang et al. 2014). Also, the near-surface temperature in the region has risen by more than 0.5 °C over the last 50 years (IPCC 2007). In contrast to the maximum temperature, the minimum temperature has risen more rapidly (Niang et al. 2014). According to Niang et al. (2014), the mean annual temperature across the majority of the sub-Saharan region is predicted to rise by 2 °C in the middle of the twenty-first century, and by 4 °C in the late twenty-first century, under the representative concentration pathway (RCP)8.5. Similarly, Almazroui et al. (2020) predict an increase of the mean temperature by 1.2, 1.5, and 1.8 °C under weak (SSP1-2.6), moderate (SSP2-4.5), and strong (SSP5-8.5) forcing, respectively. Additionally, persistent droughts and floods are predicted in areas bordering the western Indian Ocean, including most parts of eastern Africa (IPCC 2007).
In Tanzania, climate change and its impacts on catchments hydrology have been the subjects of interest in recent years (Gulacha & Mulungu 2017; Luhunga et al. 2018; Shagega et al. 2018; Ayugi et al. 2021). A study by Gulacha & Mulungu (2017) in the Wami-Ruvu basin reported a change of precipitation by −68 to 328% and −56 to 199% during the 2050s, under SRES A2 and B2 scenarios, respectively, with respect to the baseline period of 1961–1990. The study further reports a change of 0.2–7.5 °C and −0.4 to 1.5 °C from the 2020s to the 2080s for maximum and minimum temperature, respectively. Similarly, Näschen et al. (2019) in the Kilombero catchment predicted an increase in temperature, and rainfall change of approximately −8.3 to 22.5% in the 2060s, under RCP4.5 and RCP8.5. Also, a streamflow change of 61.6–67.8% was predicted. Another study by Shagega et al. (2019) in the Ngerengere River catchment predicted an increase of temperature between 0.2 and 2.6 °C in the 2050s, and between 2.7 and 4.4 °C in the 2080s, under the SRES A2 scenario, but the rainfall was predicted to decrease by 3–58%. According to these studies, changes in temperature and precipitation will have a considerable impact on the hydrology of catchments. However, because the majority of studies were carried out at the regional level, they are less helpful for planning and managing water resource infrastructures at the catchment scale (IPCC 2014b). This is partially due to the region's diverse climate, which varies from one catchment to another (Mbungu 2016). Therefore, climate prediction needs to be made at a finer spatial scale in order to provide reliable results (Toulmin 2009). In addition, past studies have paid less attention to the high-humidity tropical region where the Upper Ruvu River watershed (URRW) is located. This region is unique due to high rainfall variability resulting from Intertropical Convergence Zone and El Nino-Southern Oscillation (Mbungu 2016). As a result, the prediction of future climatic conditions and their effects on streamflows is critical for water resource planners and watershed stakeholders to enable effective water resource management.
Furthermore, global circulation models (GCMs) have been the widely used tools for the prediction of future meteorological conditions around the globe (Eisner et al. 2017; Wang et al. 2018; Ahmed et al. 2019; Givati et al. 2019). Most frequently, hydrologic models rely on the output of the downscaled GCMs to predict future hydrologic conditions (Chen et al. 2012). For climate projections in eastern Africa, particularly Tanzania, it has been a common practice for researchers to use a single GCM (Shagega et al. 2018). However, studies indicate that the amount of uncertainty associated with a single GCM prediction is fairly high (Wang et al. 2018; Ahmed et al. 2019), hence multi-model ensembles are strongly recommended in order to reduce these uncertainties (Deb et al. 2018). In this context, integrating data from multiple GCM outputs is particularly desired in order to generate reliable estimates (Rotich & Mulungu 2017; Deb et al. 2018).
Therefore, the main objective of this study is to assess the impacts of climate change on streamflows in the high-humid tropical URRW. The specific objectives are: (1) generating future climate scenarios in the URRW based on multiple GCMs under RCP4.5 and RCP8.5; (2) developing, calibrating, and validating the Hydrologic Engineering Center's Hydrologic Modeling System (HEC-HMS) model to represent the catchment hydrology; and (3) quantifying the potential impacts of climate change on the streamflows of the URRW.
Study area
The map of the URRW showing (a) topography and hydro-meteorological stations and (b) slopes.
The map of the URRW showing (a) topography and hydro-meteorological stations and (b) slopes.
The map of the URRW showing (a) land use/land cover and (b) soil types.
The plots of (a) mean monthly rainfall and (b) the mean monthly streamflow of the study area.
The plots of (a) mean monthly rainfall and (b) the mean monthly streamflow of the study area.
Data
The digital elevation model (DEM), hydro-meteorological, and land use/cover data were used for developing, calibration, and validation of the HEC-HMS model for this study. The observed climatic data were used for calibration and validation of LARS-WG, while the projected temperature and rainfall data for RCP4.5 and RCP8.5 were used for generating the future streamflows. In this study, the hydro-meteorological data for the baseline period (1951–1978) were obtained from the Wami-Ruvu Water Basin office, which has daily records for 11 rainfall stations (Table 1), 2 climatic stations, and 10 streamflow stations. For the calibration of the hydrologic model, one streamflow station (1H10) was used. The selected streamflow gauging station has relatively complete streamflow records and represents a significant area of the watershed. The daily outputs of rainfall and maximum and minimum temperatures from six CMIP5 GCMs under RCP4.5 and RCP8.5 (Table 2) were extracted from World Climate Research Programme (WCRP) website (https://esgf-node.llnl.gov/search/cmip5/). Moreover, the Space Shuttle Radar Topographical Mission (SRTM) DEM with 30 m spatial resolution was downloaded from USGS Earth Explorer. The land use/cover data were also downloaded from USGS Earth explorer and processed by using ArcGIS software. The processed land use/cover maps were further verified by using site surveys, existing maps from the Tanzania Ministry of Water, and satellite imagery. The soil data for the study area were obtained from the International Soil Reference & Information Center (ISRIC) database and were verified using locally available maps.
A summary of meteorological stations used in this study
S/N . | Station ID . | Station name . | Lat . | Long . | Alt . | Rainfall . | Temperature . |
---|---|---|---|---|---|---|---|
1. | 09737006 | Matombo Mission | −7.08 | 37.77 | 388 | x | |
2. | 09738016 | Mikula | −7.25 | 38.25 | 84 | x | x |
3. | 09637045 | Mondo | −6.95 | 37.63 | 1,285 | x | |
4. | 09637052 | Morogoro Hydromet | −6.82 | 37.65 | 512 | x | |
5. | 09737000 | Duthumi Estate | −7.38 | 37.82 | 91 | x | |
6. | 09737024 | Kibungo Mission | −7.07 | 37.68 | 975 | x | |
7. | 09737015 | Bunduki | −7.03 | 37.62 | 1,281 | x | |
8. | 09737008 | Kisaki | −7.47 | 37.60 | 183 | x | |
9. | 09737011 | Kikeo Mission | −7.22 | 37.55 | 610 | x | |
10. | 09737016 | Mizungu Mgeta | −7.07 | 37.58 | 1,097 | x | |
11. | 09637076 | Morogoro Met. | −6.82 | 37.65 | 512 | x | |
12. | 09737013 | Chenzema Mission | −7.12 | 37.60 | 1,676 | x |
S/N . | Station ID . | Station name . | Lat . | Long . | Alt . | Rainfall . | Temperature . |
---|---|---|---|---|---|---|---|
1. | 09737006 | Matombo Mission | −7.08 | 37.77 | 388 | x | |
2. | 09738016 | Mikula | −7.25 | 38.25 | 84 | x | x |
3. | 09637045 | Mondo | −6.95 | 37.63 | 1,285 | x | |
4. | 09637052 | Morogoro Hydromet | −6.82 | 37.65 | 512 | x | |
5. | 09737000 | Duthumi Estate | −7.38 | 37.82 | 91 | x | |
6. | 09737024 | Kibungo Mission | −7.07 | 37.68 | 975 | x | |
7. | 09737015 | Bunduki | −7.03 | 37.62 | 1,281 | x | |
8. | 09737008 | Kisaki | −7.47 | 37.60 | 183 | x | |
9. | 09737011 | Kikeo Mission | −7.22 | 37.55 | 610 | x | |
10. | 09737016 | Mizungu Mgeta | −7.07 | 37.58 | 1,097 | x | |
11. | 09637076 | Morogoro Met. | −6.82 | 37.65 | 512 | x | |
12. | 09737013 | Chenzema Mission | −7.12 | 37.60 | 1,676 | x |
A summary of selected GCMs for this study
S/N . | Model . | Resolution . | Institution . |
---|---|---|---|
1. | ACCESS1-3 | 1.25° × 1.25° | Commonwealth Scientific and Industrial Research Organization/Bureau of Meteorology, Australia |
2. | Bcc-csm1-1 | 2.8° × 2.8° | Beijing Climate Center, China Meteorological Administration, China |
3. | CanESM2 | 2.8° × 2.8° | Canada Centre for Climate Modelling and Analysis, Canada |
4. | IPSL-CM5A-MR | 2.5° × 1.25° | Institut Pierre Simon Laplace, France |
5. | HadGEM2-ES | 1.875° × 1.25° | Met Office Hadley Centre, UK |
6. | MIROC5 | 1.4° × 1.4° | Atmosphere and ocean research institute (the University of Tokyo), National Institute for Environmental Studies, and Japan Agency for Marine-Earth Science and Technology |
S/N . | Model . | Resolution . | Institution . |
---|---|---|---|
1. | ACCESS1-3 | 1.25° × 1.25° | Commonwealth Scientific and Industrial Research Organization/Bureau of Meteorology, Australia |
2. | Bcc-csm1-1 | 2.8° × 2.8° | Beijing Climate Center, China Meteorological Administration, China |
3. | CanESM2 | 2.8° × 2.8° | Canada Centre for Climate Modelling and Analysis, Canada |
4. | IPSL-CM5A-MR | 2.5° × 1.25° | Institut Pierre Simon Laplace, France |
5. | HadGEM2-ES | 1.875° × 1.25° | Met Office Hadley Centre, UK |
6. | MIROC5 | 1.4° × 1.4° | Atmosphere and ocean research institute (the University of Tokyo), National Institute for Environmental Studies, and Japan Agency for Marine-Earth Science and Technology |
METHODS
The hydrometric and meteorological data were first checked for missing data and consistency which would otherwise cause biases in the study results (Ayalew et al. 2022). The data consistency was checked by the double mass curve method. In addition, a number of common techniques for filling in missing data including the inverse distance weighting, normal ratio, and multiple linear regression methods were considered and analysed in this study. The normal ratio method was found to provide relatively accurate estimates and therefore was adopted for filling in missing rainfall data. Filling of missing data was only applied at the Mondo rainfall station which had a significant length of missing values. Also, the change point analysis was conducted in order to establish the baseline period for this study. The abrupt changes were analysed by using Pettit's, Buishand's, Hubert Segmentation, and Bayesian method of Lee and Heghinian tests (Rougé et al. 2013). These tests were carried out by using the Khronostat software developed by the French Institute of Research for Development (IRD).
The climate projections were carried out by using the outputs of six CMIP5 GCMs (Table 2) under RCP4.5 and RCP8.5. The RCP4.5 scenario was chosen because it represents the on-going global collaborative effort to reduce the consequences of climate change, whereas the RCP8.5 scenario was chosen because it represents the highest radiative forcing among the four RCPs (Moss et al. 2008). Both scenarios are widely used by water resources planners for developing mitigation and adaptation strategies owing to climate change impacts (Niang et al. 2014). The CMIP5 GCMs were first ranked based on their spatial resolution, and then we selected the commonly used GCMs in the tropical region that has a finer spatial resolution. The generation of climate scenarios was carried out by using LARS-WG. The HEC-HMS model was then developed, calibrated, and validated by using the observed hydro-meteorological data. The HEC-HMS model performance was evaluated by using Nash–Sutcliffe efficiency (NSE) and the coefficient of determination (R2), as described in Table 3. The LARS-WG generated climate scenarios were combined to generate the ensemble-mean (EM) climate scenario, which was then incorporated into a calibrated HEC-HMS model for simulating the future streamflow in the watershed.
A mathematical description and interpretation (according to Moriasi et al. 2007) of statistical criteria used for evaluating HEC-HMS model performance
. | NSE . | R2 . |
---|---|---|
Mathematical equation | 1 − ![]() | ![]() |
Performance measures | ||
Very good | >0.75 | > 0.85 |
Good | 0.65 < NSE ≤ 0.75 | 0.75 < R2 ≤ 0.85 |
Satisfactory | 0.5 < NSE ≤ 0.65 | 0.60 < R2 ≤ 0.75 |
. | NSE . | R2 . |
---|---|---|
Mathematical equation | 1 − ![]() | ![]() |
Performance measures | ||
Very good | >0.75 | > 0.85 |
Good | 0.65 < NSE ≤ 0.75 | 0.75 < R2 ≤ 0.85 |
Satisfactory | 0.5 < NSE ≤ 0.65 | 0.60 < R2 ≤ 0.75 |
Note. Qo and Qs are observed and simulated streamflow values, respectively. ‘–’ on the top of the variable denotes the mean value.
Calibration and validation of LARS-WG
For each climatic variable, P0 and Pn are set to 0 and 1, respectively, with corresponding values of Z0 = min {Zobs} and Zn = max {Zobs}. The downscaling process begins with the selection of a random value from one of the intervals, followed by the selection of a random value by using the uniform distribution (Hassan et al. 2013). In order to correctly assess the climatic variables, Pi is assigned close to 1 for extremely high values and close to 0 for extremely low values. The remaining intervals are evenly distributed in the probability scale (Chisanga et al. 2017).
In this study, the observed rainfall, and minimum and maximum temperatures data were used for the calibration of the LARS-WG. The performance of LARS-WG was validated by using Kolmogorov–Smirnov (K–S) test as well as t and F-tests to compare means, standard deviations, and probability distributions. These tests search for differences between the simulated and observed data. Each test calculates a p-value which is used to determine whether the simulated and observed data have the same distribution. A very low p-value (below 0.01) indicates that the observed and simulated values are unlikely to have the same distribution, hence poor model performance (Chisanga et al. 2017). Additionally, the simulated versus observed values of mean monthly rainfall, daily rainfall maxima, minimum daily minima and maxima, and maximum daily minima and maxima were also used to assess the goodness of fit. The calibrated LARS-WG was then used to generate the synthetic data of rainfall and the maximum and minimum temperatures for two periods (2041–2060 (the 2050s) as near future and 2081–2100 (the 2080s) as far future) under RCP4.5 and RCP8.5.
Calibration and validation of the HEC-HMS model
Sensitivity analysis is an important aspect of hydrological modelling (Song et al. 2015). It is used to rank model parameters according to their impacts on model output and helps in identifying the optimal parameters during model calibration (Pechlivanidis et al. 2011). The sensitivity analysis for the HEC-HMS model was performed manually by changing one parameter at a time as described by Ayalew (2019). The value of the model parameter was increased and decreased by 40% at a 5% interval while observing the value of NSE. In this study, a total of 17 HEC-HMS model parameters were subjected to sensitivity analysis.
RESULTS AND DISCUSSION
Calibration and validation of the LARS-WG
The change point analysis for meteorological data (1951–1978) was done at the annual time scale using a 0.05 significant level. The results indicated no significant break in the distribution of the annual rainfall series; hence, the baseline period (1951–1978) was selected based on the available data.
The results of the K–S test for seasonal wet/dry SERIES distribution
Season . | Wet/dry . | N . | K–S . | p-Value . | Performance . |
---|---|---|---|---|---|
DJF | Wet | 11.5 | 0.003 | 1.000 | Perfect fit |
DJF | Dry | 11.5 | 0.075 | 1.000 | Perfect fit |
MAM | Wet | 11.5 | 0.115 | 0.996 | Very good fit |
MAM | Dry | 11.5 | 0.061 | 1.000 | Perfect fit |
JJA | Wet | 11.5 | 0.032 | 1.000 | Perfect fit |
JJA | Dry | 11.5 | 0.058 | 1.000 | Perfect fit |
SON | Wet | 11.5 | 0.163 | 0.892 | Very good fit |
SON | Dry | 11.5 | 0.095 | 1.000 | Perfect fit |
Season . | Wet/dry . | N . | K–S . | p-Value . | Performance . |
---|---|---|---|---|---|
DJF | Wet | 11.5 | 0.003 | 1.000 | Perfect fit |
DJF | Dry | 11.5 | 0.075 | 1.000 | Perfect fit |
MAM | Wet | 11.5 | 0.115 | 0.996 | Very good fit |
MAM | Dry | 11.5 | 0.061 | 1.000 | Perfect fit |
JJA | Wet | 11.5 | 0.032 | 1.000 | Perfect fit |
JJA | Dry | 11.5 | 0.058 | 1.000 | Perfect fit |
SON | Wet | 11.5 | 0.163 | 0.892 | Very good fit |
SON | Dry | 11.5 | 0.095 | 1.000 | Perfect fit |
The results of the K–S test for daily rainfall and minimum and maximum temperature distributions
. | . | Daily rain . | Daily Tmin . | Daily Tmax . | |||
---|---|---|---|---|---|---|---|
Month . | N . | K–S . | p-Value . | K–S . | p-Value . | K–S . | p-Value . |
Jan | 12 | 0.065 | 1.000 | 0.053 | 1.000 | 0.053 | 1.000 |
Feb | 12 | 0.195 | 0.726 | 0.053 | 1.000 | 0.053 | 1.000 |
Mar | 12 | 0.065 | 1.000 | 0.053 | 1.000 | 0.106 | 0.999 |
Apr | 12 | 0.065 | 1.000 | 0.053 | 1.000 | 0.033 | 1.000 |
May | 12 | 0.067 | 1.000 | 0.105 | 0.999 | 0.000 | 1.000 |
Jun | 12 | 0.137 | 0.972 | 0.053 | 1.000 | 0.053 | 1.000 |
Jul | 12 | 0.124 | 0.990 | 0.053 | 1.000 | 0.053 | 1.000 |
Aug | 12 | 0.1 | 1.000 | 0.053 | 1.000 | 0.053 | 1.000 |
Sep | 12 | 0.08 | 1.000 | 0.053 | 1.000 | 0.053 | 1.000 |
Oct | 12 | 0.065 | 1.000 | 0.053 | 1.000 | 0.106 | 0.999 |
Nov | 12 | 0.112 | 0.997 | 0.053 | 1.000 | 0.053 | 1.000 |
Dec | 12 | 0.064 | 1.000 | 0.053 | 1.000 | 0.053 | 1.000 |
. | . | Daily rain . | Daily Tmin . | Daily Tmax . | |||
---|---|---|---|---|---|---|---|
Month . | N . | K–S . | p-Value . | K–S . | p-Value . | K–S . | p-Value . |
Jan | 12 | 0.065 | 1.000 | 0.053 | 1.000 | 0.053 | 1.000 |
Feb | 12 | 0.195 | 0.726 | 0.053 | 1.000 | 0.053 | 1.000 |
Mar | 12 | 0.065 | 1.000 | 0.053 | 1.000 | 0.106 | 0.999 |
Apr | 12 | 0.065 | 1.000 | 0.053 | 1.000 | 0.033 | 1.000 |
May | 12 | 0.067 | 1.000 | 0.105 | 0.999 | 0.000 | 1.000 |
Jun | 12 | 0.137 | 0.972 | 0.053 | 1.000 | 0.053 | 1.000 |
Jul | 12 | 0.124 | 0.990 | 0.053 | 1.000 | 0.053 | 1.000 |
Aug | 12 | 0.1 | 1.000 | 0.053 | 1.000 | 0.053 | 1.000 |
Sep | 12 | 0.08 | 1.000 | 0.053 | 1.000 | 0.053 | 1.000 |
Oct | 12 | 0.065 | 1.000 | 0.053 | 1.000 | 0.106 | 0.999 |
Nov | 12 | 0.112 | 0.997 | 0.053 | 1.000 | 0.053 | 1.000 |
Dec | 12 | 0.064 | 1.000 | 0.053 | 1.000 | 0.053 | 1.000 |
The scatter plots of LARS-WG-simulated and observed: (a) mean monthly rainfall; (b) minimum temperature; and (c) maximum temperature in the URRW.
The scatter plots of LARS-WG-simulated and observed: (a) mean monthly rainfall; (b) minimum temperature; and (c) maximum temperature in the URRW.
Calibration and validation of the HEC-HMS model
The observed and HEC-HMS-simulated streamflow during (a) calibration and (b) validation.
The observed and HEC-HMS-simulated streamflow during (a) calibration and (b) validation.
Projected changes in meteorological conditions
The projected monthly rainfall changes in (a) the 2050s and (b) the 2080s periods, under RCP4.5 and RCP8.5 scenarios.
The projected monthly rainfall changes in (a) the 2050s and (b) the 2080s periods, under RCP4.5 and RCP8.5 scenarios.
The baseline and projected annual rainfall at various stations during (a) the 2050s and (b) the 2080s, under RCP4.5 and RCP8.5 scenarios.
The baseline and projected annual rainfall at various stations during (a) the 2050s and (b) the 2080s, under RCP4.5 and RCP8.5 scenarios.
The projected monthly changes of minimum temperature in (a) the 2050s and (b) the 2080s, under RCP4.5 and RCP8.5 scenarios.
The projected monthly changes of minimum temperature in (a) the 2050s and (b) the 2080s, under RCP4.5 and RCP8.5 scenarios.
The projected changes of monthly maximum temperature in (a) the 2050s and (b) the 2080s, under RCP4.5 and RCP8.5 scenarios.
The projected changes of monthly maximum temperature in (a) the 2050s and (b) the 2080s, under RCP4.5 and RCP8.5 scenarios.
The plots of baseline and projected streamflow hydrograph during (a) the 2050s and (b) the 2080s, under RCP4.5 and RCP8.5 scenarios.
The plots of baseline and projected streamflow hydrograph during (a) the 2050s and (b) the 2080s, under RCP4.5 and RCP8.5 scenarios.
To gain more insight into projected future meteorological conditions in the region, the results of this study were compared with a number of studies from neighbouring catchments. The study results were found to be consistent with Shagega et al. (2018) in the Ngerengere catchment which projected an increase in rainfall, and minimum and maximum temperatures in the 2050s and 2080s under the SRES A2 scenario. Similarity is also found in several studies in the region (Gulacha & Mulungu 2017; Kishiwa et al. 2018; Näschen et al. 2019). Consistency in projections essentially reduces uncertainties, allowing for more efficient planning and management of water resource infrastructures.
Analysis of future streamflow changes
The plots of percentage change of month streamflow in (a) the 2050s and (b) the 2080s, under RCP4.5 and RCP8.5 scenarios.
The plots of percentage change of month streamflow in (a) the 2050s and (b) the 2080s, under RCP4.5 and RCP8.5 scenarios.
The plots showing the change of monthly minimum flows in (a) the 2050s and (b) the 2080s, under RCP4.5 and RCP8.5 scenarios.
The plots showing the change of monthly minimum flows in (a) the 2050s and (b) the 2080s, under RCP4.5 and RCP8.5 scenarios.
The plots showing the change of monthly maximum flows during (a) the 2050s and (b) the 2080s, under RCP4.5 and RCP8.5 scenarios.
The plots showing the change of monthly maximum flows during (a) the 2050s and (b) the 2080s, under RCP4.5 and RCP8.5 scenarios.
The FDCs for baseline and projected streamflow during (a) the 2050s and (b) the 2080s, under RCP4.5 and RCP8.5 scenarios.
The FDCs for baseline and projected streamflow during (a) the 2050s and (b) the 2080s, under RCP4.5 and RCP8.5 scenarios.
The flow duration curves are useful when assessing the streamflow indices that are widely used by planners in planning, designing, and managing water resources infrastructures. In most cases, the low flows (less than Q70), medium flows (Q20–Q70), and high flows (greater than Q20) are usually used (Saez et al. 2018), hence they have been analysed in this study (Figure 15). The results indicate that high flows will be significantly higher in the 2050s and 2080s under both scenarios. For comparison, the value of Q5 is predicted to be 273 and 315 m3/s in the 2050s, and 200 and 276 m3/s under RCP4.5 and RCP8.5, respectively, while the Q5 for the baseline period is 143 m3/s. This is equivalent to a 90 and 120% increase in the 2050s, and a 40 and 93% increase in the 2080s under RCP4.5 and RCP8.5, respectively. Also, the medium flows are generally expected to be higher except for Q50–Q70 and Q65–Q70, which are predicted to be slightly lower in 2050s and 2080s, respectively. For comparison, the projected Q50 in the 2050s are 25 and 26 m3/s, and 31 and 31.5 m3/s in the 2080s, under RCP4.5 and RCP8.5, respectively, whereas the Q50 for the baseline period is 26.6 m3/s. This is equivalent to a decrease of approximately 6 and 2% in the 2050s, and an increase of 17 and 18% in the 2080s. Moreover, the low flows (Q70–Q90) are projected to be lower in both periods, but very low flows (less than Q90) are projected to be higher. For comparison, the Q75 for the baseline period is 13.7 m3/s, while the projected Q75 are 10.7 and 10.9 m3/s during the 2050s, and 11.1 and 10.6 m3/s during the 2080s under RCP4.5 and RCP8.5, respectively. On the other hand, the Q95 for the baseline period is 8.6 m3/s, while the projected Q95 in both periods is 10 m3/s, under both scenarios.
Conclusions and recommendations
In this study, the impacts of climate change on streamflow in the URRW were evaluated. First, the meteorological data were analysed for abrupt changes in order to establish the baseline period. The Khronostat software was used for analysing the abrupt changes in the observed rainfall distributions. The results indicated that there was no break in meteorological data. As a result, 1951–1978 was selected as the baseline period. The climate projections were made from six CMIP5 GCMs outputs, which were then downscaled by the calibrated LARS-WG at each station. The EM scenario was then computed and incorporated into the calibrated and validated HEC-HMS model for future streamflow projections.
The LARS-WG was found to be a reliable tool for generating climate scenarios for the URRW. The rainfall and minimum and maximum temperatures were well simulated with R2 above 0.726 for all variables. Moreover, the daily streamflow observed at the 1H10 streamflow station was used for calibration and validation of the HEC-HMS model. The calibration and validation results indicated that the HEC-HMS model is suitable for the streamflow simulation in the watershed. The model yielded the NSE and R2 of 0.76 and 0.76 during calibration, and 0.68 and 0.76 during validation, respectively. This performance is good considering the data scarcity in the watershed.
Furthermore, the EM scenario derived from multiple GCMs was used for projecting the watershed streamflows. This approach is well described by Ma et al. (2021) and provides reliable estimates since direct averaging of the GCMs outputs smooths the climatic variability. The findings indicate that future climate changes will have significant meteorological impacts on the streamflow of the URRW. It was found that the monthly rainfall could potentially increase by 6.7–78% during the 2050s, while an increase of 8–80% is predicted during the 2080s, under RCP4.5 and RCP8.5. The minimum temperature could potentially change between −0.43 and 1.02 °C during the 2050s, and 0.16 and 3.1 °C during the 2080s, while the maximum temperature could potentially change between −0.74 and 1.28 °C during the 2050s, and −0.05 and 3.0 °C during the 2080s, under RCP4.5 and RCP8.5.
The meteorological changes will have impacts on streamflow which is anticipated to change by −65.1 to 89.7% during the 2050s, and −62.9 to 120.9% during the 2080s, under RCP4.5 and RCP8.5. On a monthly scale, a decrease in streamflow is projected for January (30%), February (36%), September (49%), October (39%), November (63%), and December (65%) months, while the increase in streamflow is projected in March (4.2%), April (34%), May (10%), June (33%), July (69%), and August (3%) months. Additionally, a comparison of baseline and projected minimum and maximum streamflow revealed significant changes. The results indicate that a significant minimum streamflow change of approximately 20–41 m3/s during the 2050s, and 22–97 m3/s during the 2080s could potentially occur in April under RCP4.5 and RCP8.5. Maximum streamflow is also expected to change significantly in April for both periods. Moreover, a comparison of flow duration curves (FDCs) indicates an overall increase of high flows (larger than Q20), very low flows (smaller than Q95), and moderate flows (between Q20 and Q70) under both scenarios. However, the low flows (between Q70 and Q90) will be slightly lower.
The results of this study are based on six GCMs and only two scenarios (RCP4.5 and RCP8.5). As described by Ma et al. (2021), incorporating more GCMs and emission scenarios for hydrological studies is important for a better understanding of associated uncertainties. Therefore, it is recommended to incorporate additional GCMs and emission scenarios in order to generalize the study findings. Moreover, the HEC-HMS model was calibrated and validated by using 5- and 4-year streamflow data, respectively. Although the model was able to simulate well the watershed streamflow, it is recommended to recalibrate the model once the new data are available.
Additionally, the impacts of future land use/cover change were not incorporated in this study. Despite the fact that the majority of the watershed area is currently undisturbed, changes in land use/cover are expected to be significant in the near future due to the growing population in the watershed. Therefore, streamflow projections in future studies should consider both climate and use/cover changes in order to increase the accuracy of the results.
Overall, the results of this study are important for providing an insight into the future climate conditions and surface water availability on URRW. However, they should be used with care by considering the study limitations. Also, this study is a starting point for future studies in URRW and other high-humid tropical catchments.
ACKNOWLEDGEMENT
We acknowledge the support offered by the Department of Water Resources Engineering of the University of Dar es salaam during data collection.
AUTHOR CONTRIBUTIONS
N.T. conceptualized, investigated, did data curation, performed the methodology, did formal analysis, visualized, and wrote the original draft. D.M.M.M. conceptualized, investigated, validated, supervised, wrote, reviewed, and edited the article. F.I. investigated, validated, supervised, wrote, reviewed and edited the article.
DATA AVAILABILITY STATEMENT
Data cannot be made publicly available; readers should contact the corresponding author for details.
CONFLICT OF INTEREST
The authors declare there is no conflict.