HARE Irrigation Project · Hydrological reference

Rainfall Intensity–Duration–Frequency

Explore HARE watershed-average rainfall depths and intensities estimated from NASA GPM IMERG V07 half-hourly observations, 2000–2025. Seven durations and six return periods, with transparent uncertainty and study limitations.

26 annual maxima30 min – 24 hours162.035 km² watershedGumbel estimates
Preliminary research estimates — not an official design standardThis dataset represents precipitation averaged across the HARE watershed, not rainfall at an individual bridge, culvert or rain gauge. Gauge validation, satellite-bias assessment and independent review remain necessary before adoption for safety-critical engineering design. In particular, 50- and 100-year estimates extrapolate beyond the 26-year record.

HARE IDF calculator

Choose a storm duration and annual return period.

Method of moments · Gumbel Type I
Design rainfall depth—mm of rainfall over selected duration
Average rainfall intensity—mm per hour, averaged over duration
Bootstrap 95% depth interval—mm · sampling variability only

Intensity–duration–frequency curves

All six fitted return periods. Select a curve to emphasize it on the chart.

Duration axis is logarithmic to show short and long storms clearly. Curve values are watershed averages, not point-rainfall estimates.

Numerical rainfall-frequency tables

The highlighted column matches the selected return period.

Estimated rainfall depths (mm)

Average rainfall intensities (mm/hour)

Scientific verification and limitations

Input-data integrity: 26 annual records (2000–2025); seven durations; all years report 12 complete months, matching expected image counts and zero missing rainfall observations. The 2024 annual maxima were also independently reproduced from the raw 30-minute rainfall CSV exports. The other 25 years have not been independently recalculated from raw observations.

2025 provisional-data sensitivity

Fifty-year depth estimates with and without 2025. The recorded 2025 dataset includes 4,512 provisional images (count includes extraction overlaps).

Duration2000–2025 (mm)2000–2024 (mm)Absolute change (mm)

The 50-year 24-hour estimate is 75.62 mm using 2000–2025 and 76.20 mm using 2000–2024. This modest difference does not establish that the IMERG rainfall product is unbiased or that its extreme-value extrapolations are reliable.

Earlier rainfall analysis

The earlier HARE 50-year 24-hour estimate of 58.983 mm is not reproduced by this consistently reprocessed seven-duration series (75.622 mm). The earlier extraction method has not been fully reconstructed, so the difference is unresolved. Do not interchange or combine these estimates in hydrological models.

Key cautions

  • Native IMERG resolution is roughly 0.1° (~11 km), with 30-minute temporal sampling; short localized convective bursts can be underestimated. No direct 5-, 10- or 15-minute IDF is provided.
  • Local recording-gauge comparison and possible bias correction are outstanding.
  • The 26-year observational period is relatively short for extrapolating 50- and 100-year rainfall extremes.
  • Confidence limits reflect resampling uncertainty only; satellite bias, stationarity assumptions and spatial averaging are not included.
  • GEV fits were evaluated exploratorily but were not adopted because of unstable 12-hour fitted shape estimates.
  • The series represents whole-watershed-average rainfall. Do not apply the values as a point IDF without separate validation and point-to-area analysis.

Methods and reproducibility

How rainfall maxima were constructed

Half-hourly IMERG calibrated rainfall rates (mm/hour) were spatially averaged over the imported HARE watershed polygon. Each rate was multiplied by 0.5 hours to calculate its 30-minute rainfall depth. Moving sums covering 1, 2, 4, 6, 12, 24 and 48 consecutive half-hour records provided annual maxima at 30 minutes, 1, 2, 3, 6, 12 and 24 hours. A one-day overlap protects storms crossing month boundaries. The data are annual-maxima rather than partial-duration series.

P30min = IMERG_rate_mm_per_hour * 0.5
Pduration(t) = sum(P30min for consecutive observations in the duration window)
AnnualMax(duration,year) = max(Pduration(t))
Gumbel frequency model and intensity calculation

A separate Gumbel distribution was fitted by the method of moments to each duration's 26 annual maxima. Parameters are expressed in mm. The estimated rainfall depth is converted into average intensity by dividing by the duration in hours.

s = sqrt(sum((x - xbar)^2) / (n - 1))
beta = s * sqrt(6) / pi
mu = xbar - 0.5772156649 * beta
yT = -ln[-ln(1 - 1/T)]
P(duration,T) = mu(duration) + beta(duration) * yT
I(duration,T) = P(duration,T) / (duration_minutes / 60)

Reported depth uncertainty limits are percentile-bootstrap 95% confidence intervals from 4,000 resampled annual-maxima series. They should not be interpreted as 95% prediction intervals for individual future storms.

Using the depths in HEC-HMS Frequency Storm

The downloadable HEC-HMS table provides depth-duration values for all six return periods. Choose one return period and enter the seven depth values for that annual exceedance probability. Since these rainfall depths were already estimated from watershed averages, do not apply another point-to-area reduction by default. HEC-HMS Frequency Storm documentation discusses the consistent-frequency interpretation and the area-reduction setting. Before engineering use, ensure the applicable storm area and local validity are evaluated.

Sources and references

Data source: NASA GPM IMERG V07 — Google Earth Engine data catalog. Method documentation: USACE HEC-HMS Frequency Storm. Statistical foundation: Gumbel, E. J. (1958), Statistics of Extremes. Model-input data: HARE watershed-average IMERG V07 annual maxima, 2000–2025, compiled 11 October 2026. Data-version status: preliminary 0.1.