What the radar thinks it saw
- View
QPE, flash flood guidance, recurrence intervals and FLASH. What they measure, what they do not, and why the heaviest rainfall is the kind the radar estimates worst. Which is biased low by design.
In this lesson you will
- Identify the MRMS QPE and rainfall rate products used to corroborate flood reports.
- Define flash flood guidance, the QPE to FFG ratio, and the average recurrence interval, and interpret a ratio of one or more.
- Describe the FLASH system and its unit streamflow product.
- State the radar QPE limitations that matter most for flood-producing rainfall, and the WPC excessive rainfall outlook categories.
QPE is an estimate and the E is load-bearing
MRMS merges every radar in the network onto a common grid and produces rainfall rate estimates every two minutes, plus accumulated QPE in two flavours: radar-only, and multi-sensor, which is the radar field adjusted toward the rain gauges that reported inside it. The second is better and it is also later, because it has to wait for gauges.
Neither is a measurement of rainfall. The radar measures reflectivity and a relationship converts it to a rate, and that relationship is a statistical fit to a drop size distribution that varies with the storm. Dual-pol has improved this considerably by letting the radar infer something about drop shape and size. It has not made it a measurement.
| Error | Cause | Direction |
|---|---|---|
| Warm rain underestimate | Many small drops give less reflectivity per inch of rain than the Z-R fit assumes | Too low |
| Beam overshoot | At long range the beam is above the rain-producing layer | Too low |
| Bright band | Melting snow is wet, large and highly reflective just below the freezing level | Too high |
| Beam blockage | Terrain or structures in the near field | Too low |
The first row is the one that matters here. It is flagged because it is not an occasional artifact, it is the normal condition during the rainfall that causes most of the flooding.
MRMS publishes Probability of Warm Rain, and the stated reason is that "warm rain enhancement of rain rates is underestimated by current Z-R relationships". Read that as an admission in the system's own documentation. Wherever warm rain dominates, a QPE field that looks unremarkable is entirely consistent with a road under a foot of water.
Flash flood guidance, and the ratio
Rainfall on its own does not tell a forecaster whether anything will flood, because the same three inches behaves differently on saturated ground and on dry. Flash flood guidance, issued by the river forecast centers, closes that gap: it is the amount of rain over a given area and duration needed to bring small streams up to bankfull, given what the ground is currently like. It is recalculated as conditions change, so it falls through a wet week.
- QPE to FFG ratio
- Observed rainfall divided by the guidance for the same duration, computed for 1, 3 and 6 hour windows. A ratio of 1 or more means the rain that fell has met or beaten what it takes to flood small streams there. It is the single most direct corroboration a flood report can get.
- ARI
- Average recurrence interval: how unusual the observed rainfall is at that place, expressed as a return period. A 100 year ARI says the rainfall is rare, which is useful context and is not the same statement as "this will flood".
- FLASH
- Flooded Locations and Simulated Hydrographs. An NSSL system that runs hydrologic models on the MRMS rainfall field over a grid of roughly one kilometer, turning a rainfall estimate into a streamflow estimate.
- CREST
- The distributed hydrologic model inside FLASH. Its headline output is maximum unit streamflow, the flow per unit of drainage area, which is comparable between a large basin and a small one.
For CREST unit streamflow, the reference figures in the literature are around 2 cubic meters per second per square kilometer as a warning-guidance reference, and around 8 as indicative of a potentially significant flash flood. Treat both as order-of-magnitude signposts rather than thresholds; they are model output, and the model inherits every QPE error in the table above.
Flash flood guidance is calculated for small streams. An urban catchment where the water runs off concrete into a storm drain does not behave like a small stream, and documented FLASH cases show urban flooding at ratios well under 1. A low ratio is weak evidence against a flood report and no evidence at all against an urban one.
Forecasting the chance of exceeding guidance
WPC issues the product that everything else hangs off, the excessive rainfall outlook, and its definition is worth knowing precisely because it is not what most people assume. It forecasts the probability that rainfall will exceed flash flood guidance within 40 kilometers, about 25 miles, of a point. It is a probability about a neighborhood, not about your address.
| Category | Probability of exceeding FFG within 40 km |
|---|---|
| Marginal | at least 5% |
| Slight | at least 15% |
| Moderate | at least 40% |
| High | at least 70% |
High risk is issued on days 1 to 3 only, and it is rare. The 40 km radius is the same neighborhood scale that lesson 7 borrows for report cross-validation, and that is not a coincidence.
The practical reading for a spotter: a Marginal is not "nothing will happen". It is a one-in-twenty chance of guidance being exceeded somewhere within a 25 mile circle of any given point, across a large area, which over a season is a great deal of flooding. And a Moderate or High over your county is the day to have the gauge emptied and the low-water crossings in mind before dark.
Check yourself
A spotter reports a road under a foot of fast-moving water. The radar-only QPE over that point shows 1.2 inches in the last hour, and the QPE to FFG ratio is 0.7. What is the correct handling?
- Accept it as plausible. Warm rain QPE runs low here and an urban catchment can flood below a ratio of 1
- Reject it. The ratio is below 1, so the rainfall was insufficient to flood small streams
- Reject it. 1.2 inches in an hour is above the 1 inch per hour flood criterion, so the report double-counts
- Hold it until the multi-sensor QPE arrives, and reject it if that is also below the guidance
Hint
Two separate things in this lesson point the same way about a warm-rain rainfall estimate and about what flash flood guidance is calculated for.
Answer
Accept it as plausible. Warm rain QPE runs low here and an urban catchment can flood below a ratio of 1
Both corroborating numbers are biased against this report in ways this lesson documented. Radar QPE underestimates warm rain, which is the normal case in deep tropical moisture and the reason the Probability of Warm Rain product exists, and flash flood guidance is computed for small streams rather than storm drains. The one inch per hour figure in the third option is not a criterion for anything.
Sources
- Gourley et al. 2017, Bull. Amer. Meteor. Soc. 98(2), 361–372: the FLASH project, improving the tools for flash flood monitoring and prediction
- WPC Excessive Rainfall Outlook
- NSSL Severe Weather 101, flood basics
- NSSL FLASH project
- NOAA MRMS product guide
- Martinaitis et al. 2020, J. Hydrometeor. 21(7), 1441–1464: the HMT multi-radar multi-sensor hydro experiment
- Wang et al. 2011, Hydrol. Sci. J. 56(1), 84–98: the CREST distributed hydrological model