Drought Monitor for the Jonkershoek Valley, Western Cape, South Africa



This analysis was prepared by the Fynbos Node of the South African Environmental Observation Network (SAEON).

Here we present the record of stream flow rates for the Langrivier catchment and rainfall from the Dwarsberg weather station in the Jonkershoek Valley for the period January 1961 to the end of March 2018. The Dwarsberg weather station is at 1214 metres above sea level on the boundary of the catchments of the Eerste, Berg and Sonderend rivers and is a good indicator of rainfall feeding the Berg and Theewaterskloof dams.

We plan to update this page monthly. For live weather data and the record over the past month please access the Dwarsberg weather station directly. You may also be interested in our Constantiaberg weather station on the Cape Peninsula or our Engelsmanskloof weather station in the Cederberg.


PLEASE NOTE: We cannot guarantee that these data or analyses are error free, particularly with regards to the older historical data and the comparability of the older and newer instrumentation. Gaps in the data were filled using linear interpolation from the most recent data for up to a 30 day period. This filled all gaps other than the period August 2008 to September 2011 - these data exist in hardcopy, but are still being digitized. There are much better gap-filling techniques, but the appropriate method depends on the question being asked.


Streamflow record



Here are the recent monthly streamflow values for Langrivier relative to averages across the entire time period (1961 to current).

Black error bars around the mean (grey column) indicate 95% confidence intervals. Monthly values lower than the lower bar are among the lowest 5% of recorded flows for that month. January, February, March and May 2017 are the lowest streamflows ever recorded for their respective months. Note that missing months (e.g. October-November 2015) are those that overlap a data gap of over 30 days.


We can also view this cumulatively:



Or compare across all years on record:



Note that “Summer” denotes the Austral summer (here defined as October - March), overlapping 2 years. In this figure we have lumped summers by the preceding year, e.g. summer 2016 represents the period October 2016 to March 2017, which incidentally has been the driest summer on record in terms of streamflow.

*Data for October - December from the summer of 2015/2016 are missing due to vandalism of the weir, but January - March 2016 was among the lowest 10% of flows for that period, suggesting that 2015 was likely the driest year on record. Of those with complete data, 2003 was the driest year, followed by 1974.


The following dynamic figures illustrate the mean monthly stream flow. You can hover over a piece of the time series to get the date and streamflow value, and adjust and zoom the window of interest by sliding the toggles in the bar at the bottom of each figure. Note the historical data (pre September 2011) has been infilled previously, but we still need to ascertain where the gaps were and what method was used.




Rainfall Record


Here are the recent monthly rainfall values relative to averages across the entire time period (1945 to current):

Black error bars around the mean (grey column) indicate 95% confidence intervals. Monthly values lower than the lower bar are among the lowest 5% of recorded flows for that month.

*Indicates that these months were missing data in 2015 due to power supply problems.


And if we view this cumulatively:



Or compare across all years on record:



So 2015 and 2016 have been the lowest rainfall years on record, while 2017 and 2014 were not far behind… Note that data for the period 1992-2013 were collected by volunteers and were not necessarily collected on the first of the month or every month, creating issues in the split between years and between summer vs winter data. This appears to be more of an issue in the 2000s. Nevertheless, the low rainfall reported for 2000-2004 does coincide with a period of low streamflow.


Or dynamically:




Note the frequent “zero-rainfall” months in the period 2000 to 2013, likely because the volunteers who collected the data were not able to visit the rain gauge in those months. The gauge can collect >1000mm before overflowing, so this is unlikely to have been a problem for rainfall totals.



Streamflow ~ Rainfall


There have been various calls for some sort of indication as to how streamflow responds to rainfall. We all want to know how much it needs to rain to fill the dams. While we don’t have a fully coupled atmosphere-hydrological model on hand, and we’re still reading up on the most appropriate time-series modelling approach (especially considering the data gaps…), here’re some quick and dirty indicators based on the periods of overlapping streamflow and rainfall data (1961 - 1991, 2013 - 2016).

Firstly, cumulative monthly streamflow as a function of monthly rainfall:


While it looks messy, one clear pattern is that a rainfall event in a winter month typically results in greater streamflow than a similar sized event in a summer month. This is especially evident if you compare large rainfall events in late summer/autumn (March, April, May) with those in winter (June, July, August). This is not surprising and likely a result of accumulated soil moisture (or lack thereof) throughout the season, while lower evapotranspiration in winter will also be coming into play.

What does this look like if we aggregate months by season (following the Summer = October to March, Winter = April to September approach from above).


A clear role for accumulated soil moisture throughout the season. Note that the lowest blue point was the winter of 2015. Interestingly the winter of 2016 had the next lowest rainfall, but higher streamflow than a few winters with higher rainfall. This could just be spatial variability in rainfall in the catchment, or the censoring created by our summer/winter divide. Ideally we’d look at different accumulation lags - 30, 60, 90 days etc, but we’ll need to do some reading about the best way to do this. It seems like a good job for Wavelet analysis.

We’ll need to think a bit more about all this, and especially how to look at evapotranspiration for inclusion in later updates. Feel free to post comments/suggestions below!


If you spot any issues, have any queries, or would like to use or cite these data please contact the SAEON Fynbos Node or jasper or abri, both @saeon.ac.za.




comments powered by Disqus