Design rainfall
Design Rainfall Estimation in South Africa: Software, TR102 and RLMA&SI
Last reviewed by Stephan Dreyer · Reviewed by Robert Fortuin
Every design flood estimate in South Africa starts from the same question: how much rain should the design assume? Whether the method is the Rational Method, SCS-SA or a full rainfall-runoff model, a “1:100-year flood” is built on a “1:100-year rainfall”, and that number has to come from somewhere. In South Africa it usually comes from the Design Rainfall Estimation in South Africa software by JC Smithers and RE Schulze. This article covers where to download it, how it works, how it replaced TR102, and what to check when you see a design rainfall quoted.
Download the Design Rainfall Estimation software
The current version is Design Rainfall in South Africa, Version 3 (July 2012). JC Smithers distributes it from a public Google Drive folder, “Design Rainfall Estimation in South Africa (DRE-SA)”. The download page given in the original user manual, on the old UKZN hydrorisk site, no longer responds.
The folder holds one file, Design Rainfall Estimation in SA.zip (about 40 MB). Inside it:
| File | What it is |
|---|---|
Rainfall3.jar | The program, a Java application |
sagrid.zip | The national 1′ × 1′ grid database; unzips to SAgrid.dbf (about 76 MB) |
Installation.pdf | Installation and user instructions (Appendix A of the WRC report) |
Final Report K5_1060 Design Rainfall and Flood Estimation in SA.pdf | WRC Report 1060/1/03, the method behind the software |
j2re-1_4_0-win-i.exe | The Java 1.4 runtime installer the software originally shipped with |
The report is also in the DWS National Flood Studies Programme library as Design Rainfall and Flood Estimation in South Africa (Smithers and Schulze, 2003).
Installing and running it
- Unzip the package into one folder. The manual uses
C:\design_rainfall\. - Unzip
sagrid.zipinto the same folder to createSAgrid.dbf. - Install a Java runtime. The bundled Java 1.4 installer is the one the manual describes; if it fails on a current version of Windows, install a current Java runtime instead.
- Double-click
Rainfall3.jar. If nothing opens, start a command prompt in the folder and runjava -jar Rainfall3.jar.
The manual’s minimum requirements are modest (32 MB RAM, 140 MB of disk, Windows 98 or later). For other operating systems, such as Linux, the manual says to install the Java runtime for that system.
What you enter
- Location: latitude and longitude (decimal degrees, degrees and minutes, or minutes), or a SAWS daily rainfall station by number (for example 0239482) or by name (for example Cedara).
- Durations: any of 23, from 5 minutes to 7 days: 5, 10, 15, 30 and 45 minutes; 1, 1.5, 2, 4, 6, 8, 10, 12, 16, 20 and 24 hours; and 1 to 7 days.
- Return periods: 2, 5, 10, 20, 50, 100 and 200 years.
- Block size: a radius, in minutes of arc, around the site. The program estimates design rainfall at every 1′ × 1′ grid point in the block, which shows how design rainfall varies across a catchment. A 2′ block returns the 24 grid points around the site.
What you get
The results go to a plain-text file that you name. It echoes your selections, then lists:
- the 1- to 7-day design depths at the selected station and at the five closest daily rainfall stations, with each station’s distance, record length, MAP and altitude;
- the design depths at the site and at every grid point in the block, for each duration and return period;
- a 90% lower and upper bound for every value.
What “design rainfall” means
Design rainfall is the rainfall depth (mm) expected at a location for a given duration (say 15 minutes, 1 hour or 1 day) and return period (say 1:50 or 1:100 years; see return periods explained). Divide depth by duration and you get the average intensity (mm/h) used in peak-flow formulas. The full set of depth-duration-frequency values for a site is often drawn as IDF (intensity-duration-frequency) curves.
Design rainfall is a statistical estimate extracted from long rainfall records. The quality, length and density of the rain-gauge network therefore set the ceiling on how good any flood estimate can be.
TR102: the long-serving reference
For decades the standard reference was TR102 (Southern African Storm Rainfall, Adamson, 1981), published by the then Department of Environment Affairs. TR102 gives 1, 2, 3 and 7-day design depths for return periods of 2 to 200 years at about 1 946 rainfall stations, from a censored log-normal distribution fitted to each station’s partial duration series. Shorter durations were derived from the daily values with fixed ratios: TR102 puts the 1-hour depth at 0.60 of the 24-hour depth in summer rainfall regions and 0.41 in winter rainfall regions (DWS guide, Table 3-1).
TR102’s limitation is its age. It reflects records available in the late 1970s, and four more decades of data have changed the estimates at many stations. The WRC study behind the current method compared the two at 2 184 daily rainfall stations: below the 1:50-year return period the differences were generally under 20%, and at longer return periods the TR102 values were generally higher.
DWS no longer uses TR102 storm rainfall in its own flood studies (A Guide to Flood Frequency Estimation Methods as applied in the DWS, July 2025, section 3.8). TR102 survives inside the Design Flood Estimation Tool because the Alternative Rational Method and the Standard Design Flood procedures were built on it.
RLMA&SI: the method behind the software
The software implements the RLMA&SI method, the Regional L-Moment Algorithm with Scale Invariance, developed by Smithers and Schulze for Water Research Commission project K5/1060 and published as WRC Report 1060/1/03 (2003). It joined two earlier studies:
- Short durations (24 hours and less), from digitised rainfall records at 172 stations with at least 10 years of data (Smithers and Schulze, 2000a).
- Long durations (1 to 7 days), from daily records at 1 789 stations with at least 40 years of data, grouped into 78 relatively homogeneous regions and fitted with the GEV distribution using L-moments (Smithers and Schulze, 2000b).
Two ideas drive the method:
- Regionalisation. Fitting a distribution to one gauge is noisy for rare events, because even an 80-year record holds few genuinely extreme storms. Pooling data across a homogeneous region with L-moment statistics gives more stable estimates of rare events.
- Scale invariance. The daily records are far longer and more reliable than the short-duration ones, so the 1-day growth curves are used for every duration. The index value (the mean annual maximum) is estimated for 1 day from MAP, latitude and altitude, then scaled to longer durations directly and to shorter ones through the 24-hour value, with pivot durations at 15 minutes, 2 hours, 1 day and 3 days.
The result is design rainfall at a 1′ × 1′ spatial resolution anywhere in South Africa, for durations from 5 minutes to 7 days and return periods from 1:2 to 1:200 years, each with 90% bounds. You will see the same dataset called RLMA&SI, K5/1060 or DRE-SA.
Using the output in a flood calculation
A typical chain for a small-catchment design flood:
- Run the software for the site, with a block size that covers the catchment.
- Determine the critical storm duration, for peak-flow methods the catchment’s time of concentration.
- Read the design depth for that duration and return period. For a catchment, average the grid points that fall inside it, which is how the DFET manual uses the grid output.
- Reduce the point depth to an areal depth with an areal reduction factor where the catchment is large enough to need one.
- Convert rainfall to runoff with the chosen method (Rational, SCS-SA, or a hydraulic model’s inflow hydrograph).
An error at step 3 scales everything after it. Two studies of the same site can differ simply because one used TR102 values and the other RLMA&SI, so check the rainfall source before concluding that the hydrology disagrees.
DFET accepts the software’s output through its “1′ × 1′ Grid RLMA&SI” option. Paste the depths for a run of consecutive durations that brackets the time of concentration, and always include the 24-hour depth (the SCS method needs it) and the 1, 2, 3 and 7-day depths (the Alternative Rational Method needs them). DWS uses the same software for its Smithers-Schulze regional rainfall approach, averaging the grid values over a defined catchment area.
Design rainfall in Delta Hydro Studio
Delta Hydro Studio’s Design Rainfall tool returns the K5/1060 depth-duration-frequency table for the nearest grid point anywhere in South Africa, with lower and upper bounds, and lists the nearest Design Rainfall Estimation stations for comparison. It fits IDF curves to the table (Sherman by default) so intensities between the published durations come from one equation. The same tool reads NOAA Atlas 14 in the United States and the Bureau of Meteorology IFD in Australia. Studio’s Design Flood Estimation tool pulls the design rainfall for a delineated catchment automatically.
What to look for when reviewing a report
- Source stated. The report should name its design rainfall source (RLMA&SI from the Design Rainfall Estimation software, TR102, or a drainage-manual IDF relationship) beside its rainfall table.
- Appropriate durations. Urban drainage design needs sub-hourly design rainfall. Quoting only 1-day depths for a 20-minute catchment is a red flag.
- Point or areal. A depth read at one grid point is a point value. A large catchment needs the grid averaged over its area and an areal reduction factor.
- Consistency across methods. Where several flood methods are compared, they should share the same design rainfall. Otherwise the comparison mixes method differences with input differences.
- Climate context. Standard design rainfall datasets describe the historical record. For long-lived infrastructure it is increasingly common, and in some jurisdictions expected, to test sensitivity to rainfall intensification.
Frequently asked questions
Is the Design Rainfall Estimation software free?
Yes. The Google Drive folder is public, and the package downloads without registration or payment.
What is the difference between RLMA&SI, K5/1060 and DRE-SA?
They refer to the same work. RLMA&SI is the method, K5/1060 is the Water Research Commission project number (published as WRC Report 1060/1/03), and DRE-SA is the name of the software’s download folder.
Can I still use TR102?
It is still built into DFET for the Alternative Rational Method and the Standard Design Flood, so it appears in many reports. DWS no longer uses TR102 storm rainfall, and at long return periods TR102 generally gives higher depths than RLMA&SI. A report that relies on TR102 should say why.
Does the software give intensities or IDF curves?
It gives depths. Divide a depth by its duration for the average intensity. Fitting IDF curves to the table is a separate step, which Studio’s Design Rainfall tool does.
Design rainfall extraction, duration analysis and method-consistent application are part of every hydrological study and stormwater management plan we produce. For how the design rainfall becomes a design flood, see design flood estimation in South Africa.
Downloads and documents referenced
Links go to the publisher's own copy: the Department of Water and Sanitation, the Water Research Commission or the software's developers. Check there for newer versions.
- Design Rainfall Estimation in South Africa (DRE-SA): software download folderSoftware · Version 3, July 2012 · University of KwaZulu-Natal
- Design Rainfall and Flood Estimation in South Africa (Smithers and Schulze, 2003)Report · WRC Report 1060/1/03 · Water Research Commission
- A Guide to Flood Frequency Estimation Methods as applied in the DWS (July 2025)Guideline · Eleventh edition, version 2 · Department of Water and Sanitation
Try it yourself
- Rational Method Peak Flow Calculator →
Compute peak discharge Q = C·i·A for small catchments, with hectare and square-kilometre inputs.
- Time of Concentration Calculator →
Estimate Tc with the Kirpich and SANRAL Drainage Manual formulas from watercourse length and average slope.
Start a conversation
Tell us about your project
Need this applied to a real site? Tell us about the project and we'll reply within one business day.
Reply within one business day
