12.4 An Improved Post-Processing Technique for Automated Precipitation Gauge Time Series

Thursday, 16 January 2020: 9:15 AM
203 (Boston Convention and Exhibition Center)
Amber Ross, EC, Saskatoon, SK, Canada; and C. D. Smith and A. Barr

Precipitation data originating from automated accumulating gauges are derived as differential bucket weight measurements over a defined period. The technological process that determines the precipitation gauge bucket weight and transmits this information to a recording device is often susceptible to mechanical and electrical interference. Depending on the measurement principle, the bucket weight measurement can be confused by diurnal (potentially temperature related) signal oscillations and negative drift from evaporation of the bucket contents. Separating the signal from the noise in order to make accurate precipitation estimates can be challenging, especially when precipitation rates are low. This makes the accurate measurement of solid precipitation even more difficult, especially considering that solid precipitation rates are further reduced by the effects of wind. Environment and Climate Change Canada (ECCC) uses a variety of post-processing techniques in both research and operational monitoring to filter cumulative precipitation time series derived from high-frequency, bucket-weight measurements. Four of these techniques are described and tested here: 1) the near-real-time Operational 15-minute filter (O15), 2) the automated Neutral Aggregating Filter (NAF), 3) the Supervised Neutral Aggregating Filter (NAF-S), and 4) the automated Segmented Neutral Aggregating Filter (NAF-SEG). Inherent biases and errors in the first two post-processing techniques have revealed the need for a robust automated post-processing method capable of removing random noise, diurnal oscillations, and evaporative (negative) drift from the raw data. This evaluation focuses on cold-season (October to April) accumulating-precipitation-gauge data at 1-min resolution from two sources: a control (pre-processed time series) with added levels of synthetic noise and drift; and raw (minimally-processed) data from several WMO Solid Precipitation Inter-Comparison Experiment (SPICE) sites. Evaluation against the control with synthetic noise shows the effectiveness of the fully-automated NAF-SEG technique to recover the control precipitation to within 2% under varying levels of introduced noise. The NAF-SEG filter produced the highest correlation coefficients and lowest RMSE for all synthetic noise levels and had comparable performance to the supervised but manually-intensive NAF-S method. In the minimally-processed (SPICE) test cases, NAF-SEG showed a lower bias in 37 of 44 time series and a similar bias in 5 time series. The results indicate that the NAF-SEG post-processing technique provides substantial improvement over current automated techniques used by ECCC but requires a 24-hour data latency.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner