Comparative Assessment of MRMS Gauge-Corrected QPE and GPM IMERG Version 07 Final Run Precipitation During the May 2019 Extreme Flood Event in the Neosho River Watershed, Oklahoma

The Neosho River watershed in northeastern Oklahoma has experienced repeated flooding for decades, and the May 2019 event was among the most severe, with peak discharge near Commerce approaching a 100-year return level. Accurate precipitation data are essential for reconstructing such events and for supporting hydrologic modeling and flood-risk assessment. This study compares ground-based Multi-Radar Multi-Sensor (MRMS) Gauge-Corrected Quantitative Precipitation Estimation (QPE) with satellite-based GPM IMERG Version 07 Final Run precipitation for the full month of May 2019, treating MRMS as the near-surface reference on a common 0.1° hourly grid.

IMERG tracked the event's daily timing well, producing a daily correlation of approximately 0.97. At the hourly scale, however, the correlation dropped to about 0.79, and IMERG consistently underestimated rainfall intensity, with an hourly bias of -0.20 mm and a cumulative event deficit of roughly 98 mm relative to MRMS. Intensity-based evaluation showed that IMERG performed poorly on light-rainfall days (correlation near 0.08) while overestimating rainfall on moderate and heavy days, with positive biases of approximately 9.5 mm and 17.8 mm respectively.

Spatial structure analysis using daily variograms revealed that both products agreed on the event-scale storm footprint (range of 24.86 km), but MRMS retained far stronger spatial contrasts (sill of ~916 mm²) compared to IMERG (~138 mm²), indicating that IMERG smooths out fine-scale rainfall gradients. Hourly variogram comparisons showed near-zero correlations for range (0.05) and nugget (-0.03), confirming that IMERG cannot resolve the sharp spatial patterns that MRMS captures at convective scales. Storm movement analysis found similar total centroid path lengths (MRMS: 242.8 km; IMERG: 247.1 km), but IMERG showed a daily centroid bias of approximately 90.9 km, likely driven by infrared morphing during microwave data gaps.

Overall, IMERG Version 07 captures general storm evolution but requires bias correction and spatial downscaling before it can reliably support hydrologic modeling or flood-risk assessment in the Neosho River watershed.