#05 2020

Continuing drought in Algoa water source areas: climate and streamflow trends in the Kromme catchment

By Julia Glenday, SAEON Fynbos Node

Wash your hands well and often, for 20 seconds each time, and keep your surroundings clean, but be sure not to use more than 50 L of water per day…

It is never a good time for a drought, but with the ongoing Covid-19 pandemic a water crisis is a double whammy for the Algoa region of the Eastern Cape.

This article takes a look at some recent and historical climate and estimated streamflow statistics for the Kromme catchment, one of the major water supply catchments in the area where SAEON postdocs and students have been conducting research.

Algoa drought and water supply – keeping up with an uncertain future

While much of the Western Cape has emerged from the infamous 2017–2018 ‘Day-Zero’ drought over the last year, other parts of South Africa have experienced continued or renewed water scarcity. As of 20 September 2020, the five water supply reservoirs of the Algoa Water Supply System had a combined storage of 18.7%.

The Algoa Water Supply System supplies the city of Port Elizabeth in the Nelson Mandela Bay Metropole (NMBM) and several surrounding towns, as well as the Coega Industrial Development Zone and large areas of commercial irrigated agriculture. Water restrictions have been in place for several months, with those listed in July 2020 allowing agricultural users 20% of their normal allocation and the NMBM 70%. From September 2020, a 50 L per person per day restriction was placed on domestic users in the NMBM, with flow restrictors to be widely installed.

Even without drought conditions, the Algoa Water Supply System was recognised to be insufficient to meet projected demand, last thought to be in theoretical balance in 2009. Efforts to reconcile supply and demand have led to investment in river transfer infrastructure, bringing in water all the way from the Orange River via the Fish River into the Sundays River, at no small cost.

The reconciliation strategy also calls for addressing system leaks and water recycling in the NMBM and acknowledges the positive impacts of clearing of invasive alien plants in catchment areas of the reservoirs, an ongoing effort largely through the Working for Water programme.

This, and other water supply-demand reconciliation strategies across the country, rely on assumptions about the level of climate variability that can be expected in areas feeding their supplies. There is growing recognition of the need to take climate change into account going forward, urgently underlined by the current prolonged dry periods.

The Algoa region lies in the a-seasonal rainfall zone of South Africa, between the winter rainfall zone to the west and the summer rainfall zone to the east. It has inherently high interannual climate variability, sometimes catching the edge of winter and/or summer weather systems and sometimes not. This poses an extra challenge for predictive climate modelling.

However, looking across the suite of downscaled global climate models’ (CMIP3 and 5) predictions across different greenhouse gas emissions scenarios (RCP 4.5 and 8.5) presented on the Climate Information Platform of the University of Cape Town Climate Systems Analysis Group (https://cip.csag.uct.ac.za/), the majority of models suggest a future decrease in winter rainfall and an increase in summer rainfall on average, with increasing average temperatures and likelihood of extremes (2040–2060 model predictions relative to 1980–2000). The magnitudes of predicted change vary notably across the models and scenarios, but there is less discrepancy about the direction of change.

Upper Kromme catchment and available data

Water supply shortfalls are the outcome of many factors which change over time. Prolonged dry and warm weather conditions will impact the supply-vs-demand scarcity differently depending on properties of the catchment areas that impact streamflow generation and groundwater recharge (e.g. land cover and use), the water supply infrastructure and management (e.g. storage and leakage) and the magnitude of the demand.

Analyses presented here for the Kromme catchment focus on the climate and the catchment response. They provide a preliminary exploration into a few key questions: How extreme is the current climatological drought compared to the long-term climate record for the area? Is there evidence of a long-term trend in annual and seasonal rainfall and streamflow? If there is a shift towards less winter and more summer rainfall, what could this mean for water supply?

The upper catchment of the Kromme River feeds the Kromrivier Dam (also referred to as the Churchill Dam), one of the major reservoirs in the Algoa Water Supply System, contributing close to 30% of NMBM’s normal water allocation. The 360 km2 catchment area lies in the Tsitsikamma and Suuranys mountains and is dominated by fynbos vegetation. It is also known for its palmiet wetlands, however much of the area is highly invaded with black wattle (Acacia mearnsii).

The area supports fruit orchards and dairy and small stock farming. There are a few rainfall and temperature gauges from the South African Weather Service (SAWS) and Department of Water and Sanitation (DWS) in operation in and around the catchment (Figure 1), some with data going back to the 1950s or earlier.

Unfortunately, there is no long-term streamflow gauge in the catchment. However, the water balance from the Kromrivier Dam, i.e. recorded reservoir water levels, outflows, rainfall and evaporation available from DWS, can be used to back-calculate an estimate of the streamflow entering it starting in 1957, with useable data until late 2017 when an outflow gauge malfunctioned.

Being so mountainous, rainfall is unevenly distributed over the catchment. Rainfall spatial surfaces derived from station data by Lynch, 2003 (Figure 1), accounting for elevation, aspect, and distance from sea, were used to upscale station data to estimate a catchment-scale average rainfall timeseries for the Kromme for the 1960–2018 water years.

Temperature data was similarly scaled using surfaces from Schulze and Maharaj (2004) and was used to estimate potential evapotranspiration (PET) using the Hargreaves and Samani (1985) method. PET is a measure of the potential for water to evaporate or be transpired by plants if there is water available.

Data from SAWS, the Agricultural Research Council (ARC), DWS and SAEON were used to patch gaps in records for the long-term stations. ‘Water years’ are used instead of calendar years in hydrological analyses to avoid splitting seasons down the middle. In this case, an April to March water year (i.e. April 1960 to March 1961 as the 1960 water year) was applied because average streamflow was found to be higher in winter.

Starting the water year in October is more typical, but less appropriate when there is a winter streamflow peak because this would place the flow recession from a winter high into the next ‘water year’ (Figure 2).

The area supports fruit orchards and dairy and small stock farming.

Climate and streamflow trends

Annual rainfall and runoff for the Kromme catchment by water year are shown in Figure 3. The volume of streamflow is normalised by the catchment area to give a runoff depth value in millimetres for comparison to rainfall. The annual average rainfall for this period was 638 mm, but there is high variability with values ranging from 322 mm (50% of average) in 2008 to 1 035 (165% of the average) in 1971.

On average, there is a fairly even split between rain falling in winter versus summer (52% vs 48%), but again this is highly variable, with winter rain ranging from 22% to 77% of the annual total in different years. The datasets used did not extend through the full 2019 water year, however catchment rainfall for the 2019 calendar year was 453 mm, well below average. The data shows that the current hydrological drought is not a product of the lowest rainfall years on record in recent decades, however it is due to continuous below-average rainfall for multiple years in a row.

In the Kromme, streamflow is more variable than the rainfall. This is typical of semi-arid environments because they reach wetness thresholds for producing streamflow less often. The ratio of runoff to rainfall shows that on average, 18% of annual rainfall becomes streamflow, with values for individual years ranging from as low as 1% in 2017 to 48% in 1971.

Years with the lowest rainfall were not always those with the lowest runoff or runoff ratio: seasonal distribution, rain intensity and the duration of a dry period made a difference. Several below-average rainfall years in a row were found to result in progressively declining runoff ratios as seen in 2008–2010 and 2016–2017, such that a year with higher rainfall occurring later in a prolonged dry period can produce less streamflow than a lower-rainfall year occurring earlier in the drought.

Winter rainfall tended to produce a higher streamflow response, likely because evaporative demands are lower, as shown in Figure 4 illustrating monthly rainfall vs PET. The average winter runoff ratio was 20% versus 13% in summer. Years with more rainfall had higher runoff ratios (R2=0.52, p<0.1), but winter rainfall was a stronger predictor than total annual rainfall (R2=0.60, p<0.1).  Although the long-term average showed a near even split of rainfall between seasons, 67% of the Kromme’s streamflow came in winter months.

The Mann-Kendall trend test (Mann 1945, Kendall 1975) was applied to both annual and seasonal rainfall, PET and runoff to determine if there is any statistical evidence of progressive change over time, over and above the interannual variability. Results indicate a statistically detectable declining trend in both annual rainfall (MK test statistic z =-1.37, p=0.08) and winter rainfall (z = -1.41, p=0.08) in the Kromme from 1960–2018, with no indication of a trend over time in summer rainfall (z= -1.0, p=0.3).

PET showed clearer evidence of an increasing trend in annual, winter and summer values (z values 3-5, all p-values <0.01). Annual and winter runoff totals showed detectable declining trends (z = -1.9 and -2, p= 0.03, and 0.04), in keeping with the rainfall, as did the annual and winter runoff ratios (z = -1.6 and -1.9, p= 0.06, and 0.08), in keeping with the increase in PET overall. No trends were detectable for summer runoff (all p-values > 0.01).

These observed trends are in general agreement with climate model predictions for future decades for the region.

Nelson Mandela Bay’s second-biggest supply dam, the Impofu Dam, has reached critical levels. (Image: NMBM)

More to do, for everyone

Water managers and users are well aware of the problem given the Algoa Water Supply System has had supply shortfalls and needing use restrictions in 2009­–2010, 2017–2018 and again in 2020. The current dam levels in the system are similar to the extreme low values observed in mid-2018 when a very large rainfall event in September 2018 helped the system rebound.

We can only hope that happens again soon for some short-term relief; however, the longer-term analyses and modelled climate predictions do suggest a continuing drier future for this region on average. This should be accounted for in planning going forward.

The sensitivity of the catchment streamflow response to evaporative demand further supports the need for removal of invasive alien plants which heighten the evaporative losses. SAEON researchers and students are continuing to do hydrology research in the Kromme catchment with the aim of improving hydrological modelling for change prediction.

Further reading  

  • DWS (Department of Water and Sanitation) (2020). Algoa Water Supply System Operational Analysis 2020–2012: August 2020 Monthly Monitoring Report.
  • DWS (2011). Water Reconciliation Strategy Study for the Algoa Water Supply Area (Pretoria, South Africa: Aurecon for Department of Water and Sanitation).
  • Hargreaves, G. and Samani, Z. (1985). Reference Crop Evapotranspiration from Temperature. Applied Engineering in Agriculture 1, 96–99.
  • Kendall, M.G. (1975).Rank Correlation Methods, 4th edition, Charles Griffin, London.
  • Mann, H.B. (1945).Non-parametric tests against trend, Econometrica 13:163–171.
  • Lynch, S.D. (2003). Development of a Raster Database of Annual, Monthly and Daily Rainfall for Southern Africa (Pretoria, South Africa: Water Research Commission (WRC)).
  • NMBM (Nelson Mandela Bay Municipality) 2020a. Nelson Mandela Bay Municipality Dam Levels. https://www.nelsonmandelabay.gov.za/damlevels
  • NMBM (Nelson Mandela Bay Municipality) 2020b. Nelson Mandela Bay Municipality Notice: Restrictions on the Use of Water 87 – 09 September 2020. https://nelsonmandelabay.gov.za/DataRepository/Documents/water-restrictions-9-september-2020_y66k4.pdf
  • Schulze, R.E. and Maharaj, M. (2004). Development of Database of Gridded Daily Temperature for Southern Africa (Pretoria, South Africa: Water Research Commission (WRC)).

Figure 1. Location of the Kromme catchment area in the Eastern Cape (above), the upper Kromme catchment area showing active rainfall gauges and an interpolated surface for mean annual rainfall (MAP) produced by Lynch 2003 (below).

Figure 2. Estimated average rainfall, potential evapotranspiration (PET) and runoff by month for the Kromme catchment based on data for 1960–2018 (error bars indicate standard error of the mean), shown for a water year from April–March. This alternative water year was selected over the typical October–September water year to account for the winter peak in runoff, keeping the recession period following it within the same water year.

Figure 3. Annual rainfall (bottom), runoff (top) and runoff ratio (middle) for the Kromme catchment indicating seasonal winter-summer split and long-term averages. NB: Displayed by water year, rainfall 1960-2018, runoff and runoff ratio 1960–2016.

Figure 4. Monthly rainfall, PET (expressed as a negative value) and the rain-PET deficit shown with estimated monthly streamflow from the Kromme catchment (Kromriver Dam reservoir inflow).