1. Introduction
Climate change impacts present challenges of different dimensions to the development and management of water resources. In many parts of the world, especially Africa, water supply systems are already stressed and impacts of climate change will further complicate management of most of the systems. Across Africa, decrease in annual discharge will significantly affect present surface water supplies in large parts of the continent by the end of the century [
1]. Southern Africa, due to its dependency on rain-fed water systems for food production, has been projected to be one of the regions of the world that will be negatively affected by climate change [
2]. It follows therefore that water resource projects ought to consider climate change impacts for future planning, and management. Moreover, much of southern Africa depends on hydropower as a main source of electricity such that any changes in the water resources may result in changes in electricity supply. Timmermann et al. observed that climate change impacts are rarely explicitly considered in water resources planning, operations and management [
3]. Although the available water resources in the Zambezi Basin, in general, exceed the demand at present, this situation may change as a result of the increase in population, more industrial and mining development, increased irrigated food production, a higher standard of living of the population, including the environmental water demand of the system. There are more than 28 relatively large dams with a storage capacity in excess of 12 million m
3 in the Zambezi River Basin, built for domestic, industrial, and mining water supply, irrigation and power generation. Kariba is the largest (160,000 million m
3) and Cahora Bassa the second largest (52,000 million m
3).
Historical observations of rainfall and temperature show that Africa in general is warming at the rate of about 0.05 °C per decade, with slightly larger warming in the June–November seasons than in December–May season [
4]. The future projections for temperature in the region show increasing temperatures over the entire region. Temperature is expected to increase by 2–5 °C [
5,
6,
7] by the end of the 21st century. These higher temperatures will increase the rate of evapotranspiration. Hewitson and Crane observed drying trends for the months of October–December (western side of southern Africa) and for January–March period. However an increase in precipitation to the south eastern part of the region was observed [
8].
The Climate Systems Analysis Group (CSAG) and the University of Cape Town (South Africa), has developed comprehensive future climate projection scenarios for the southern African region. The future climate scenarios are based on GCMs used in the fourth report of the International Panel on Climate Change (IPCC—AR4), and recently using the IPCC—AR5 GCMs. The general direction of change—increasing temperatures and reduced precipitation—appears to be consistent [
9,
10]. These projections are likely to impact the water resources and hence the hydropower systems in the region.
There have been a few studies around the southern African region regarding climate change and its impact on water resources. Some of the most recent studies focused on the Zambezi River Basin [
11,
12], on the Okavango delta in Botswana [
13,
14,
15], the Pungwe River Basin in Zimbabwe and Mozambique [
16] and on the entire Zambezi River Basin, a risk assessment of the river system [
17] and more recent Water Supply and Demand Scenarios for the Zambezi River Basin [
18]. The Southern African Development Community (SADC) through the Global Carbon Capture and Storage Institute provides hydropower and hydrology simulations with quantification of elasticity [
19]. Most of these studies indicate that the temperature is rising while precipitation is likely to reduce in the 21st century. Most of the studies highlight that impacts of these changes are reduced water resources (river flows in most rivers) in the southern African region as indicated through the following publications [
2,
20,
21,
22,
23,
24,
25,
26,
27,
28,
29]. The southern African region has not yet been extensively studied as far as the climate change impacts on hydropower production potential are concerned.
On a global scale, the impacts of climate change on hydropower have been analyzed and the results indicate that at a global scale the impacts are minimal and slightly positive [
30,
31]. However, on the regional level of the southern Africa, there are negative impacts [
2,
30,
31]. This is mainly due to declining river flows as observations already indicate. Other studies have suggested that also flow regulation and irrigation can alter local freshwater conditions. This can result in consistent effect like increasing evapotranspiration and decreasing temporal runoff variability from flow regulation [
32]. In the southern African region, power is beginning to be pooled together through the Southern African Power Pool (SAPP) and more interconnections are planned to strengthen and improve the power exchange [
33,
34].
The objective of this study is to evaluate impacts of climate change on water resources in Zambezi River Basin and its implications on hydropower production potential. The following major hydropower projects were included: Victoria Falls, Kariba, Kafue, Cahora Bassa and Shire River. The Kariba hydropower system comprises of two hydropower plants on both sides of the dam wall. The hydropower plants lie in two countries, Zambia and Zimbabwe. In this study, these plants are analyzed as a unit. The study investigates the impacts of changes based on the changes in river flows. As such, the climate change is the main driver of change. The landscape drivers of change like land and water use and water storage can counter the effect of climate change [
32], highlighting the importance of knowing the land and water use for future developments in the region from these related activities.
The process of assessing impacts of climate change involves selecting or defining possible future climate. Likely, future climate scenarios are generated by General Circulation Models (GCMs) and are the main tools for researchers [
31,
35,
36,
37]. The procedure in general is that GCMs simulations of future climate are used as inputs into hydrological models in order to study the impacts on river flows. As GCMs cannot provide the required spatial resolutions for hydrological model it is, most times, necessary to downscale to finer resolution suitable for hydrological modelling [
38,
39,
40,
41,
42,
43]. Downscaling is usually carried out through two different methods; dynamic downscaling, or statistical downscaling.
2. Study Area
The Zambezi River lies in south central Africa within 8°42′ and 21°35′ south and 18°11′ and 36°17′ east. It is Africa’s fourth largest river after the Nile, the Congo and the Niger Rivers. The basin has a total area of about 1,390,000 km
2. The Zambezi River Basin is the largest of the African river systems flowing into the Indian Ocean (
Figure 1). It is shared by eight countries and supports a population of more than 40 million people. The main economic activities within the riparian states are mining, agriculture, tourism, fisheries, and manufacturing. Most of these activities depend mainly on the electricity produced in the hydropower plants of the basin, as well as on other sources of energy (primarily coal and oil).
2.1. Climate and Hydrology
The Zambezi basin lies in the unimodal rainfall zone and therefore there is not much difference in rainfall pattern between the different parts of the basin except the reduction in amounts from north to south. Generally rainfall start in the September—November (SON) season, peaks in the December–February (DJF) season and ends in the March–May (MAM) season as illustrated in
Figure 2. The June–August (JJA) season is dry though some of the northern parts of the basin may receive rainfall, also as early as August. The average annual rainfall over the upper catchment is 1100–1300 mm, with considerably higher rainfall near the Zambezi source area in Angola while in the southern low rainfall areas; it is as low as 500 mm/year. The seasonal distribution of rainfall is shown in
Figure 2.
Table 1 contains a list of selected meteorological stations and their precipitation patterns.
The mean temperature over the basin is highest (26 °C) during the SON and DJF seasonal as shown in
Figure 2. The hottest month is October, and sometimes November–December just before the rains begin. The dry season of JJA is also the cold season, and mean temperatures can be as low as 15 °C, while the MAM season is cool.
The largest natural lake in the basin is Lake Malawi (28,750 km2). The largest artificial lakes (reservoirs) are Kariba (5180 km2) and Cahora Bassa (2660 km2). Other important reservoirs with large surface areas are the Kafue Dam (89 km2) and the Itezhi-tezhi Dam (865 km2). There are five major swamps, the Barotse, the Eastern Caprivi, the Kafue, the Busanga, and the Lukanga, covering an area of 20,000 km2 at high flood periods. The mean annual discharge at the mouth of Zambezi River is 4200 m3/s (130 km3/year) as it enters the Indian Ocean. The main contributions to Zambezi river flow are from the tributaries grouped as; Upper Zambezi upstream of Victoria falls (25%), Kafue River (9%), Luangwa River (13%), and Shire River (12%) adding to a total of 60% of the Zambezi river discharge.
Figure 2 shows the climate (rainfall and evaporation regimes) on the top plot. The lower plot shows the annual runoff pattern in upper Zambezi, and the effect of the flood plains on flows. Runoff in the upper Zambezi is highest in period of March–May. Mean annual runoff from the region is about 26.8 × 10
9 m
3 providing an average annual flow of 850 m
3/s.
The peak runoff typically reaches Lukulu by February-March but this runoff takes one and half months to pass through the Barotse Flood plains and peak discharge near the downstream outlet (Senanga) is often delayed until April or early May. Flood-waters recede slowly from the Barotse flood plains during the six-month dry season, with high evaporation losses throughout the year.
2.2. Hydropower
The hydropower facilities are listed in
Table 2. At present, the basin has 4833 MW of installed hydropower generation capacity along the main trunk Zambezi River and the two main tributaries Kafue and Shire. In addition there are a few smaller hydropower plants not included in this analysis and in
Table 2. Potential plans for additional power plants and upgrading or expansion shows that the current average hydropower production of 31,598 GWh/year could be increased.
3. Methodology
Assessment of climate change impacts on water resources and hydropower can be carried out at various levels of detail with different approaches. The main steps in the analysis we have done here for the Zambezi case are shown in
Figure 3.
Global Climate Models (GCMs) driven by future emission scenarios are the main tools used to develop future climate scenarios. It is common to use results from several different GCMs each with different emission scenarios to develop future climate scenarios. Next, by the process of downscaling, local future climate scenarios can be established for specific climate stations in the catchment. Statistical downscaling involves regression between GCMs outputs and local observations, resulting in projected future local climate for specific stations.
The first step was to access the data from several global circulation models. GCM simulations are produced by large international climate research centers worldwide (
Table 3) and the resulting data are published on servers where free downloads can be made. In total there were 24 GCMs available during the fourth assessment report (AR4) by IPCC. This number of GCMs was too large for practical use, so it was necessary to select a few GCMs to be used during this analysis. The selection process employed the Taylor diagram method to compare the GCM data and observed data from Climate Research Unit (CRU) [
18], and the five models Coupled Global Climate Model.
(CGCM3.1), Commonwealth Scientific and Industrial Research Office (CSIRO3.0), Max Planck Institute for Meteorology global Model, (ECHAM5), Community Climate System Model (CCSM3.0) and Hadley Centre Coupled Model (HACDM3) models were selected for further analysis. For more information about these models, see
Table 3. Though there were many emission scenarios available, only three scenarios (A2, A1B, B2) were selected. These scenarios cover a wide range, both low (B2), middle (A1B) and high (A2) emission scenarios. This decision also reduced the number of future ensembles of the climate variables generated.
In the Rest of This Paper the Middle Emission Scenario A1B is Used, Unless Otherwise Stated
Future climate within different regions in the basin were projected by downscaling the GCM results using the Empirical Statistical Downscaling (ESD) method of statistically downscaling [
41,
42]. Here, linear multiple regression is used to establish a statistical relationship between monthly values from station observations and the gridded GCM data outputs.
The simulated data from the five GCMs were used for downscaling to stations within the catchments and the mean/median of these results used as the future climate variables. The downscaling method used mainly daily data, in some cases data with monthly time step. For downscaling, the clim.pact package [
41,
42] was used to evaluate the expected changes on temperature and precipitation. The results of the downscaling were derived for the three future periods of 30 years. The reference period sometimes referred to as current period, is the 1961–1990 and the three future periods are called 2020s, 2050s and 2080s, representing 2011–2039, 2040–2069, and 2070–2099 respectively.
In order to provide corresponding scenarios of future runoff, results from downscaling of GCMs were applied in hydrological models to transform climate into runoff. The projected future climate scenarios were computed using change factors (see
Section 5.2) between the historical period and future periods. This was done to reduce some systematic biases. In this approach, differences in relevant climate variables—typically precipitation, temperature and evapotranspiration—were extracted from the control and scenario simulations of the climate model and processed before being transferred onto an observed time series. The change factors (sometimes called the delta changes) are a common transfer method used [
20].
Hydrological modelling was then used to transform the future climate scenarios to future river flows. The hydrological model was calibrated using the observed runoff data during the period representing 1961–1990. The study used five GCM models and the SRES emission (A2, A1B, B2) scenarios to project climate scenarios of temperatures and precipitation. The downscaling was performed at a monthly time step such that the generated output was monthly time series of downscaled mean monthly temperature (°C) and monthly precipitation amount (mm/month).
The HBV model was used for translating the climate scenarios, temperature, evaporation and precipitation, to hydrological changes. The HBV is a conceptual lumped rainfall runoff model originally developed for operational runoff forecasting [
44,
45]. It has also been used extensively to perform impact studies for climate change assessments [
46]. The model, depicted in
Figure 4 uses precipitation, air temperature and potential evaporation as input and is usually runs on a daily time-step. The model contains routines for snow accumulation and melt, soil moisture, and groundwater response. Potential evaporation was estimated based on temperature and precipitation series by the Hargreaves method, which is a modified Thornthwaite method [
36,
47,
48].
Typically calibration of HBV involves running the model and comparing the simulated results against river flow observations to obtain optimal performance. The model parameters were calibrated to fit the observed runoff at Victoria Falls gauging stations for the period 1962–2010 (hydrological years). The calibration was performed with a combination of manual calibration (manual adjustment of parameters and weighting of precipitation stations) and automatic calibration. Inflows to the ungauged catchments constituting (mid-Zambezi) were estimated based on the calibrated model for the upper-Zambezi using scaling techniques.
In order to analyze the impact of changed flows on hydropower productions, a model that describes the hydropower system is required to simulate the system with future flows. While it is sometimes tempting to assume that the changes in runoff directly relate to changes in hydropower production, this assumption should be used only in large (regional or global) areas analysis or run-of-river systems. However, where there is storage and other user demands it is necessary to carry hydropower simulations to ascertain the changes that may result from computed changes in runoff. Since the basins selected all have reservoirs with varying sizes, the hydropower production simulations were deemed necessary. In this case the results of the hydrological simulations (river flows) were inputs into the hydropower stimulations by an energy model. The energy model was used to highlight changes that are likely to occur given the changes in the river flows.
The nMAG hydropower simulation model was used to assess the impacts of climate change on hydropower potential. The nMAG model setup for Zambezi hydropower system is shown in
Figure 5. The nMAG model was developed at NTNU [
49] from 1984–2004 and was primarily intended for operation simulation to estimate the production and economic benefit of the hydropower system under varying hydrological condition. In addition, it is capable of simulating reservoir operation strategies for an integrated water resources system that includes water supply, irrigation, and flood control projects. The model contains nodes from four different module types where all or some are contained in a system at a time. These are termed as: Regulation reservoirs, Power plant, Water transfer (Diversions) and Control point. Input data including system reservoir, power plant, bypass, and operation strategy (reservoir operation rules) are used to describe the hydropower system for each site. Reservoir evaporation and environmental requirements are specified as well. A monthly time step was selected for the runoff time series.
4. Data
The GCMs results for the five different models (
Table 3) and three different emission scenarios (B2, A1B and A2) was obtained from an InterGovernmental Panel on Climate Change IPCC data centre, the Program for Climate Model Diagnosis and Inter-comparison (PCMDI) [
2,
50]. The data access is free through its data portal [
50] by user registration.
Most of the observed historical precipitation and temperature datasets were obtained from Global Historical Climatological Network (GHCN) [
51]. Additional datasets, Zambian station data was obtained from the Zambian Meteorological Department. All the stations used in the analysis lie within the Zambezi River Basin. Observed temperature and precipitation data for the different climate stations (
Table 1) were used as observed data (predictands) in the downscaling process.
The discharge data were obtained from the Department of Water Affairs, Ministry of Energy and Water Development in Zambia. The annual and monthly discharge data were inspected and selected based on the continuity in data and position within the Zambezi Basin. The discharge data was used in the calibration of the hydrological model. This data set came with some gaps and therefore required some filling in for the missed values. The main gauging stations along the Zambezi River with data were Zambezi River at Kabompo, Zambezi River at Senanga and Zambezi River at Victoria Falls.
The hydrological description of the basin was derived based on Geographical Information systems (GIS) analysis of the basin. The data required was the basin area, sub basins, slope, and elevation zones among many parameters. Data for hydropower plants and their description was obtained from Zambia Electricity Supply Corporation (ZESCO), the electricity utility company in Zambia and from various reports. The hydropower system data required included reservoir size (area, volume), installed capacity, efficiency and other parameters.
6. Conclusions
The hydropower production system of southern African region is likely to be strongly affected by changes in climate. These results show that the future climate within and around the Zambezi catchment will get drier with temperatures higher than those in the current period. The temperature projections in the basin indicate an increase up to 2.7 °C by end of the century.
The rainfall analysis and projections show a decrease in the future precipitation which seem to continue the trend seen for current observations. The resulting effect of these climate changes on water resources is a gradual decrease in river flows, from 14% to 26%, towards the end of the century, depending on emission scenarios.
As a result of these changes, the water resources available for hydropower generation also decrease, and this will have a significant impact on the hydropower production potential. The results show that there could be a decrease up to 15%–31% in the hydropower production potentials towards the end of the century, again depending on emission scenario. Increased reservoir evaporation is a significant driver of the change in generation, in addition to the change in inflows. Sedimentation is another challenge for many reservoirs, but for the major reservoirs in Zambezi catchment, Lake Malawi, Kariba and the Cahora Bassa, the sedimentation rate is not as high and due to their sizes, it will take very long time for the reservoir volume to be affected by sedimentation.
This change is an indication that there will be large negative impacts on the hydropower system, and power deficits are likely with the current generation capacities, unless measures are taken. The scenarios presented in this study should serve as indications of direction of change and how large these could be, rather than exact predictions of the impacts of climate change in the basin.
This analysis has given a basis on which further detailed impact study on particular hydropower sites on the Zambezi basin can be evaluated. This is especially important for new hydropower development of sites that are planned within the Zambezi river basin.
There is significant uncertainty in the projections that are a result of many factors such as the lack of long climatic observations within the basin, uncertainty regarding future GHG emissions, the GCMs’ ability to adequately simulate the future climate, and the adequacy of downscaling from global to local climate change.
We feel that these results, like many other impact studies, highlight the fact that planning, development and operation of water resource projects, especially the hydropower stations, need to take into account the impact of a changing climate and the changing land use and water use in this region. Ignoring the impacts resulting from changing climate could result in uneconomical and unsustainable development and operation of water related projects.