| 08 Oct 2026
Polarimetric QPE with an X-band radar on the northwest coast of Peru: a ray-coherence diagnostic and a reliability-weighted specific-attenuation estimator
Abstract. The northwest coast of Peru is an arid region punctuated by extreme rainfall events, making accurate quantitative precipitation estimation (QPE) critical for hydro-meteorological risk management. This study develops, validates, and selects locally calibrated polarimetric QPE algorithms for a new dual-polarization X-band radar in Piura, Peru, using raindrop size distribution (DSD) data from the 2024 rainy season applied to radar data from the February–April 2025 season and evaluated against a network of rain gauges.
The manufacturer's default algorithm severely overestimated total rainfall (percentage bias, PBIAS, of +25.6 %), an error removed by the local calibration. Among the locally calibrated candidates, four variants of a specific-attenuation (AH) algorithm were formulated around different data-selection thresholds for the total differential-phase span, ΔΦDP, that constrains the ZPHI retrieval. At the gauges these variants proved nearly indistinguishable, yet noise in the span estimate perturbs each ray by a nearly multiplicative factor, imprinting radial streaks that daily point accumulations cannot register. To measure this artifact directly from the radar data, we introduce a gauge-independent, scan-level diagnostic based on the azimuthal anomaly of ln R, whose primary metric is the ray-coherent fraction of anomaly variance (coh). Applied to more than 26 000 scans, it reveals contamination in all precipitation regimes and under every threshold configuration, whereas a span-independent R(ZH, KDP) control sits at the noise floor. This motivates a reliability-weighted estimator that blends R(AH) with the local composite through a per-ray inverse-variance weight driven by the estimated noise of the ΔΦDP constraint itself. The weighted estimator matches the best threshold variant at the gauges while reducing the ray-coherent contamination by factors of 2.4–2.7 in the moderate and heavy regimes, approaching the control floor without blurring the field.
Line 87ff "First, non-meteorological echoes were filtered using a threshold on the co-polar correlation coefficient (ρhv < 0.8)."
Please mention how range gates with ρhv < 0.8 are treated. Are they completely discarded (no data)?
Line 88ff " Next, a custom algorithm was applied to remove remaining non-meteorological radial spikes. This filter
identifies and removes any contiguous echo region with an azimuthal extent of less than 10◦ that begins within 10 km of the
radar. The filter is tuned to the interference artifacts present in this dataset. "
Please provide more details about this filter, maybe a citation providing a description. How was it tuned to the RF interference? How complete was the removal?
Figure 6
In particular the strong "spokes" look like RF interference. Please explain why you are sure that these spokes are not residuals from your RF interference filtering.