NWC REU 2021
May 24 - July 30



Photo of author

Characterizing Errors in MRMS Quantitative Precipitation Estimates Over Alaska, Hawaii, and Puerto Rico

Emilia Thedens, Robert A. Clark III, Heather Grams, and Steven M. Martinaitis


What is already known:

  • Quantitative precipitation estimation (QPE) has a high spatial-temporal resolution but is subject to errors and biases.
  • MRMS domains outside the United States face additional challenges in regards to remotely sensed QPE due to the mountainous terrain and lack of good radar coverage.
  • MRMS QPE products in the United States tend to underestimate precipitation in mountainous areas with low RQI.

What this study adds:

  • Some areas of Alaska (typically with high radar quality) have error levels comparable to MRMS in the continental United States.
  • Pervasive lack of radar coverage and more intense precipitation lead to higher errors in Alaska.
  • Across all OCONUS domains there is a negative correlation between gauge catch and error; products which use specific attenuation A seem to be less prone to large underestimation errors.
  • Hawaii has consistent underestimation in Q3RAD, Q3DP, and Q3EVAP.
  • Comparison of Q3MS to Q3RAD, Q3DP, and Q3EVAP for 24 hour data over Alaska and Hawaii shows less error in the limited domain for evaluation.
  • Q3MS fills in the radar gaps in Alaska, but its performance in those gaps is still unknown.


Four radar-based quantitative precipitation estimation (QPE) products from the Multi-Radar Multi-Sensor (MRMS) system are evaluated against hourly and daily gauge-based rainfall amounts over Alaska, Hawaii, and Puerto Rico. Three products are 1) Radar-Only QPE (Q3RAD), the initial version of MRMS QPE that is reflectivity-based, 2) Dual-Polarization QPE (Q3DP), a version of Q3RAD which uses dual-polarization variables in its calculations, 3) Dual-Polarization QPE with Evaporation Correction (Q3EVAP), a product which adds an evaporation correction to Q3DP, and 4) Multi-Sensor QPE (Q3MS), which utilizes gauge- based correction and terrain data and is evaluated daily. The evaluation of QPE data takes place from June 2019 to June 2021. Both the Alaska and Puerto Rico domains exhibited spatial error trends corresponding to the quality of radar coverage, with larger errors in regions of poor coverage. The Hawaii domain was characterized by underestimation throughout the region. Puerto Rico was prone to overestimates in the west, but did not have a problem with underestimation. The results of the study are useful to those who wish to learn more about the present limitations and challenges of radar-based QPE in domains outside the continental United States.

Full Paper [PDF]