Multi-Sensor QPE - Warning Decision Training Division (WDTD)

Navigation Links

Products Guide

Multi-Sensor QPE

Short Description

Multi-Sensor QPE supplements radar data with combination of gauges and NWP QPF (HRRR/RAP blend for CONUS) to gap-fill in poor radar coverage areas. The Multi-Sensor QPE is available in two products: Pass 1 and Pass 2. Pass 1 is available first, but with less gauges included. Pass 2 has a higher latency, but with most gauges included.

Subproducts

1-, 3-, 6-, 12-, 24-, 48-, and 72-hour accumulations

Multi-Sensor QPE Pass 1 -- less latency, but less gauges
Multi-Sensor QPE Pass 2 -- more latency, but more gauges

Primary Users

NWS: WFO, RFC

Input Sources

Radar QPE w/ Gauge Bias Correction, Mountain Mapper QPE, NWP model QPF (varies based on domain), HADS and MADIS gauge networks

Resolution

Spatial resolution: 1km x 1km

Temporal resolution: Updated hourly at the top of the hour. All products end at the indicated time.

Multi-Sensor QPE Pass 1 -- 60-min latency, 60-65% of gauges included
Multi-Sensor QPE Pass 2 -- 120-min latency, 99% of gauges included

Product Creation

The Multi-Sensor QPE is based on a decision tree that determines which input product shall be assigned at each grid cell: Radar QPE w/ Gauge Bias Correction, Mountain Mapper QPE, or NWP model QPF. These decisions are based on radar coverage (via RQI), the precip type environment, and topography. The preferred input is the Radar QPE w/ Gauge Bias Correction, with the other two inputs supplementing in areas where radar coverage is poor.

Technical Details

Latest Update: MRMS Version 12

Accessible on:

Operational AWIPS (via LDM, if set-up)
MRMS Development site

Description of flow chart below

There are significant radar coverage gaps across the western CONUS and in other MRMS domains (e.g., Alaska). To supplement the radar data, the new Multi-Sensor QPE utilizes a combination of gauges (via Mountain Mapper QPE) and NWP (HRRR/RAP blend for CONUS) to gap-fill in poor radar coverage areas.

The logic uses the Radar QPE w/ Gauge Bias Correction as the primary precipitation source with its weight in the scheme defined by the RQI product.
Beyond the areas using radar coverage, the combination of Mountain Mapper and NWP are dependent on a number of variables.
1. NWP (via the 1-h forecast QPF) is utilized in snow environments and also in stratiform rain environments over areas of non-complex terrain.
2. Mountain Mapper QPE is utilized in areas of stratiform rain over areas of complex terrain.
Each source is seamlessly blended to avoid any discontinuities in the data.

NWP Model QPF Details

Below are the NWP models used as input to the Multi-Sensor QPE products, as of Spring 2020:

Domain	NWP Model QPF
CONUS	HRRR_CONUS
Alaska	HRRR_AK exper.
Hawaii	HRRR_HI exper. --> coming soon!
Caribbean	No Multi-Sensor QPE products
Guam	No Multi-Sensor QPE products

Domain Considerations

It’s worth pointing out that these Multi-Sensor products are implemented differently depending on the domain. This table quickly summarizes this information.

References

Strengths

The Multi-Sensor QPE product uses a variety of inputs (radar, gauge, model QPF) to have the best QPE coverage across the CONUS.
- The Radar QPE w/ Gauge Bias Correction leverages the spatial coverage of radar data, adjusted with direct rainfall observations. It has proven to outperform the Radar-Only product.
- The Mountain Mapper removes beam blockage artifacts and accounts for flow regimes.
- Mountain Mapper is not reliant on radar coverage.
- NWP model QPFs fill in the rest of the radar gaps that Mountain Mapper doesn’t which include the snow and convective rain environments, as well as stratiform rain in non-complex terrain.

Limitations

The Multi-Sensor QPE is most hampered by its latencies, for both Pass 1 (60 min) and Pass 2 (120 min).

It is subject to the limitations of its inputs:

Given the limitations of rain gauge measurements (e.g. calibration), radar-estimated QPE (e.g. Z-R relationships), and the Gauge QC algorithm, errors can still arise in the final product.
When the real-time precip gradient strongly differs from the PRISM climatology, large errors can occur in Mountain Mapper.
When the gauge network is sparse, the interpolation is far-reaching and can cause errors.
There are limitations in the HRRR when it comes to biases and precipitation placement/generation.

Quality Control

Gauge Quality Control

Applications

Due to its latencies, the Multi-Sensor QPE products are not as useful during flash flood warning operations. However, they can be used to:

Update rainfall totals in Flash Flood Statements
Compare with previous hours' Radar Only QPE to get a sense or real-time biases in the Radar-Only product
To validate rainfall totals
For input to hydrological models
For flood forecasting

Example Images

Fig. 1: Multi-Sensor QPE product for the 24-hour period ending 1600 UTC 23 February 2019 over northern California and Oregon. The left panel shows the initial MRMS legacy radar QPE and the right panel shows Multi-Sensor QPE. The middle section of the figure shows the statistical measures and scatter-plots of the two data sets.