Daily Rainfall Data

Open Daily Rainfall Data

Access historical daily rainfall records from the South African Weather Service (SAWS) and WRC station networks. This guide covers the Lynch (2003) quality-controlled database, observed versus patched data flags, mass and residual-mass curve diagnostics, monthly/annual seasonal analysis, MAP interpretation, and completeness heatmaps. The outputs of this tool feed directly into Flood Frequency Analysis via annual maximum series (AMS) or partial-duration series extraction.

Overview

The Daily Rainfall Data tool provides access to historical daily rainfall records from the South African Weather Service (SAWS) station network. The underlying dataset is based on the comprehensive work by Lynch (2003), published as WRC Report No. 1156/1/04: “The Development of a Rationalised Approach to the Evaluation of the Reliability of Daily Rainfall Estimates from the South African Weather Service Database.”

The dataset includes observed and patched daily rainfall records for thousands of stations across South Africa, with quality flags indicating data provenance. The tool lets you search for nearby stations, inspect record completeness, analyse mass curves for trend detection, and assess seasonal rainfall patterns.

Data source

The Lynch (2003) dataset was developed for the Water Research Commission (WRC) and provides a quality-controlled, patched daily rainfall database. It is widely used in South African hydrological practice for design flood estimation, catchment modelling, and water-resource assessment.

Getting started

The workflow starts with selecting a point of interest and ends with a station whose record is suitable for your study.

Select a location

Begin by selecting a point of interest on the map. You can click directly on the map, enter coordinates manually, or search for a place name. The tool uses this location as the centre point to find nearby rainfall stations.

Find nearby stations

Once a location is set, the tool searches for SAWS stations within a configurable search radius. Stations are ranked by distance and displayed on the map with key metadata. Select a station to load its full daily rainfall record and analysis outputs.

Search radius

Start with the default search radius. If few stations are returned, increase the radius to capture more distant stations. In data-sparse regions (e.g. Northern Cape), you may need a larger search radius.

Station data

Each station comes with a set of metadata fields that allow you to judge data quality before loading a full record.

Station metadata

Key metadata displayed for each station:

• Reliable % — Percentage of days with observed (reliable) data. Higher values indicate better original data coverage.
• Patched % — Percentage of days that have been infilled using statistical patching methods. These values are estimates, not direct observations.
• Missing % — Percentage of days with no data at all (neither observed nor patched). High missing percentages indicate unreliable records.
• Record length — Total span of the station record in years, from the first to last recorded day.

Quality assessment

A station with a long record length but a high missing percentage may be less useful than a shorter record with high reliability. Always check the completeness heatmap to understand where the gaps occur within the record.

Record length

The record length indicates the total number of years between the first and last entry in the station file. For design flood estimation, longer records are generally preferred as they capture more hydrological variability. A minimum of 20 years is typically recommended for reliable frequency analysis, with 40+ years being ideal.

Data quality flags

Every daily value in the Lynch (2003) dataset is tagged with a provenance flag that distinguishes direct gauge observations from statistically infilled values.

Observed vs patched data

Each daily value is classified as either observed or patched. Observed data comes directly from the rain-gauge measurement. Patched data has been infilled using statistical methods — typically distance-weighted interpolation from neighbouring stations or regression relationships.

Patched values maintain the statistical properties of the record (mean, variance) but are not direct measurements. They are essential for creating continuous records suitable for frequency analysis, but should be interpreted with caution for event-specific studies.

The “P” flag

In the daily data table, values flagged with “P” indicate patched (infilled) data. Unflagged values are observed measurements. The distinction matters when:

• Performing frequency analysis — consider whether to include or exclude patched values
• Validating extreme events — patched values may underestimate actual extremes
• Assessing data reliability — a high proportion of “P” flags suggests limited original data

Patched data in practice

For most design applications, using the combined observed + patched record is acceptable and recommended, as the patching methodology preserves the statistical characteristics of the rainfall distribution. Always note the proportion of patched data in your engineering reports.

Mass curve analysis

The mass curve (or cumulative rainfall curve) plots the cumulative daily rainfall total over the entire station record. It is a visual summary of the station’s rainfall history and one of the most useful diagnostic tools for assessing data quality.

A well-behaved station will show a relatively smooth, upward-sloping mass curve with a consistent gradient. The gradient of the curve at any point represents the rainfall intensity during that period.

How to interpret:

• Consistent slope — Indicates a stable rainfall regime with no significant data issues.
• Steeper segments — Periods of above-average rainfall (wet years).
• Flatter segments — Periods of below-average rainfall (dry years) or potential data gaps.
• Sudden jumps — May indicate station relocation, gauge changes, or data errors.
• Flat sections — Extended periods of missing or zero data requiring investigation.

Quick quality check

If the mass curve shows abrupt changes in slope that cannot be explained by known climatic events (e.g. droughts), investigate the station history. The station may have been moved, the gauge type changed, or the observation practices altered.

Residual mass curve

The residual mass curve (RMC) is a more sensitive diagnostic than the standard mass curve. It plots the cumulative departure from the mean annual rainfall over time, removing the dominant linear trend and making subtle changes, trends, and anomalies much easier to detect.

For each year, the residual is calculated as:

$$R(t) \;=\; \sum_{i=1}^{t} P_i \;-\; \bar{P}\cdot t$$

Residual mass curve

where $P_i$ is the annual rainfall in year $i$, $\bar{P}$ is the mean annual rainfall, and $t$ is the number of years from the start of the record.

Detecting patterns:

• Rising segments — Periods of above-average rainfall (wetter than the long-term mean).
• Falling segments — Periods of below-average rainfall (drier than the long-term mean).
• Change points — Abrupt changes in slope direction may indicate climatic shifts, station moves, or data-quality issues.
• Overall trend — A persistent upward or downward drift suggests a long-term change in rainfall regime.
• Oscillations — Regular cycles may reflect multi-decadal climate variability (e.g. ENSO influence).

Change-point detection

A sudden, sharp change in RMC slope that does not correspond to a known climatic event is a strong indicator of a non-climatic discontinuity. Common causes include station relocation, gauge replacement, change in observer, or a shift from manual to automatic recording. Such change points should be noted and may warrant splitting the record for frequency analysis.

Monthly & annual analysis

Seasonality and long-term averages are essential context when deciding whether a station is representative of your study catchment.

Seasonal patterns

The monthly analysis displays mean monthly rainfall totals as a bar chart, revealing the seasonal distribution of rainfall at the station. South Africa has diverse rainfall regimes:

• Summer rainfall — Most of the interior, with peaks in December–February (e.g. Gauteng, Free State, Limpopo).
• Winter rainfall — Western Cape, with peaks in June–August.
• Year-round rainfall — South coast and parts of KwaZulu-Natal, with bimodal or relatively even distribution.
• Late summer — Eastern regions may show peaks extending into March.

Understanding the seasonal pattern is critical for selecting appropriate design-storm methods and for judging whether the station is representative of your study catchment.

MAP interpretation

The Mean Annual Precipitation (MAP) is the long-term average of the total annual rainfall at the station. It is one of the most commonly used parameters in South African hydrology and is required as input for many design methods (e.g. Rational, SCS, SDF).

MAP comparison

Compare the station MAP with the MAP from design rainfall grids (e.g. WR2012 or K5_1060 via the Design Rainfall tool). Significant discrepancies may indicate that the station is not representative of the surrounding area, possibly due to local topographic effects, rain-shadow, or urban-heat-island influences.

Completeness assessment

The completeness heatmap provides a year-by-month grid showing data availability for the entire station record. Each cell represents one month and is colour-coded to indicate the proportion of days with data:

• Dark green — Complete or near-complete month (90–100% of days have data).
• Light green — Mostly complete (70–90% of days).
• Yellow/orange — Partial data (30–70% of days).
• Red — Mostly missing (less than 30% of days).
• Grey/empty — No data at all for that month.

The heatmap quickly reveals temporal patterns of data availability. Look for:

• Continuous blocks of missing data (station was inactive).
• Seasonal gaps (e.g. winter months consistently missing in a summer-rainfall region).
• Degradation over time (data quality declining in recent years, possibly indicating station closure).
• Early record gaps (station was established but recording was intermittent).

Frequency analysis implications

For annual maximum series extraction, a year should ideally have at least 80% data completeness, particularly during the wet season. Years with significant gaps during the rainfall season should be excluded from frequency analysis, as the true annual maximum may have been missed. See Flood Frequency Analysis for how AMS records flow into fitting a GEV or Log-Pearson Type III distribution.

Advanced analysis

The following analyses are planned or under development for the Daily Rainfall Data tool and will be released as they become available.

Frequency analysis. Fit statistical distributions (GEV, Log-Pearson III, etc.) to the annual maximum series extracted from the daily record. Estimate design rainfall depths for various return periods directly from station data, providing a site-specific alternative to regionalised estimates. (Coming soon.)

Disaggregation. Temporal disaggregation of daily rainfall totals into sub-daily intervals using established disaggregation methods. This enables the derivation of design-storm hyetographs from daily data where sub-daily records are unavailable. (Coming soon.)

Stochastic simulation. Generate synthetic daily rainfall sequences that preserve the statistical properties (mean, variance, autocorrelation, seasonal patterns) of the observed record. Useful for continuous simulation studies and reservoir yield analysis. (Coming soon.)

Data sources

Station names in the Lynch (2003) dataset carry a single-character suffix indicating the original data source. Understanding this suffix is important for assessing provenance and potential biases.

Suffix	Source	Description
W	SAWS	South African Weather Service — the primary national meteorological service. These are the most common and generally most reliable stations.
A	ARC	Agricultural Research Council — stations operated for agricultural purposes, often in farming regions. May have different siting standards.
S	SASRI	South African Sugarcane Research Institute — stations concentrated in KwaZulu-Natal and Mpumalanga sugarcane-growing regions.
P	Private	Privately operated stations (e.g. mines, estates, municipalities). Data quality may vary significantly depending on the operator.

Source considerations

SAWS (W) stations generally follow the strictest observation standards and should be preferred where available. ARC (A) and SASRI (S) stations provide valuable additional coverage in agricultural areas. Private (P) stations should be used with caution and cross-checked against nearby SAWS stations where possible.

References

• Lynch, S.D. (2003). Development of a Rationalised Approach to the Evaluation of the Reliability of Daily Rainfall Estimates from the South African Weather Service Database. WRC Report No. 1156/1/04. Water Research Commission, Pretoria.
• Smithers, J.C. & Schulze, R.E. (2003). Design Rainfall and Flood Estimation in South Africa. WRC Report No. 1060/1/03. Water Research Commission, Pretoria.
• Schulze, R.E. (1997). South African Atlas of Agrohydrology and Climatology. WRC Report TT82/96. Water Research Commission, Pretoria.
• Kunz, R.P. (2004). Daily Rainfall Data Extraction Utility. School of Bioresources Engineering and Environmental Hydrology, University of KwaZulu-Natal.

---

Open Daily Rainfall Data