Technical documentation: Regional Deterministic Precipitation Analysis

This is the technical documentation for the Regional Deterministic Precipitation Analysis. The documentation provides a general overview of the dataset, a description of how the dataset was created, potential applications and any limitations of the dataset. For a general overview of climate information concepts, explore Climate information basics.

Overview

The Canadian Precipitation Analysis System (CaPA) in its deterministic configuration (RDPA) produces an optimal estimate of the amount of precipitation over 6 and 24 hour periods. This objective estimate incorporates readings from in situ precipitation gauges, weather radars and forecasts produced by the Regional Deterministic Prediction System (RDPS). CaPA-RDPA produces four analyses per day on 6 hour precipitation amounts valid at synoptic hours (00, 06, 12 and 18 UTC) and two analyses on 24 hour amounts valid at 06 and 12 UTC. The analysis is produced on a grid at 10 km resolution and covers essentially the domain of North America (Canada, United States, Mexico).

Table 1. Main characteristics
Variables

PR : Precipitation amount in mm over 6h (24h)

CFIA : Confidence Index of the analysis

Geographic domain Covers essentially the domain of ​​North America (Canada, United States, Mexico).
Spatial resolution

15 km : From 2011 to October 2012

10 km : From October 2012 to now

Temporal resolution Accumulations over periods of 6 and 24 hours
Time series length Available since April 6th 2011
Last update September 18th  2018

Variables and formats

CaPA-RDPA produces two variables on grids :

  • Precipitation amounts (PR variable) in mm during the 6 hour (24h) interval prior to the valid time of the analysis. For example, for the 24 hour precipitation analysis valid on October 10th, 2018 at 12 UTC, the precipitation amount would cover the period going from October 9th, 2018 at 12 UTC to October 10th , 2018 at 12 UTC.
  • As a by-product, the system generates the Confidence Index of the Analysis (CFIA variable). This index, without unit, serves to inform on the weight of observations in the value of the analysis. Values for this index range from 0 to 1. A value of 0 at a given grid point means that the analysis at this location comes only from the trial field (short-term precipitation forecast from RDPS) while a value approaching 1 means that the contribution of neighboring observations is dominant.

Each analysis contains these two variables, PR and CFIA.

Applications

Quantitative estimates of near-real-time precipitation are needed for many applications, including weather forecasts, flood forecasts, crop management, forest fire prevention, hydropower production, and dam safety. Precipitation event summaries are regularly prepared from CaPA to meet internal and external demands. The precipitation analysis also allows Emergency Measures Organizations (EMOs) to monitor precipitation events causing damages to the territory, even in areas where weather stations are absent. The spatial representation provided by the gridded precipitation analysis offers a great advantage over the networks of meteorological stations whose density decreases significantly from south to north of the country. More over, the analysis offers a complete and seamless coverage of the entire domain.

Methodology

The analysis system is based on optimal interpolation (OI), also known as residual kriging, to combine all sources of precipitation information: surface stations, radar QPEs and a trial field from the RDPS. Specifically, to obtain an analysis value at a grid point, the trial field is corrected using a weighted linear combination of the sum of the innovations (difference between the observations and the trial field) at the station locations near the grid point. The weights of the linear combination are calculated in a way to minimize the estimate of the analysis error. A variogram is required to calculate the covariance matrices of the observation and trial field errors that will then be used to estimate the weights of the linear combination. This variogram, based on innovations, is updated with each analysis production run in order to take into account for seasonal changes in the atmospheric regime.

In the CaPA system, upstream of the IO, there is a lot of processing that is performed to validate the quality of observations before assimilating them. Knowing that the CaPA system is fully automated and that the trial field is of reasonable quality, the strategy is to reject more observations than necessary with the aim of letting a fraction of erroneous data through. To apply this relatively strict strategy, several quality control modules are activated during the entire processing of surface observations.
Before assimilating the Quantitative Precipitation Estimates (QPEs) from weather radars, a GOES-based cloud mask is applied to further decontaminate them. Also, using masks built on statistics derived from radar QPEs, shadow areas and quasi-permanent echoes are removed. For each radar, the bias is corrected by calculating the so-called Mean Field Bias (MFB) from recent and past precipitation data. And finally, we limit the range of radar to 120 km and only liquid precipitation is considered.

Limitations

To meet the constraints of optimal interpolation and produce the best precipitation analysis possible in terms of skill, a cubic root transformation is applied to the precipitation data. At the end of the processing, the inverse transformation is performed to generate the analysis in the real space. This back-transformation causes a bias in the analysis. A correction method has been implemented in CaPA to minimize the impact of this type of bias. The other source of bias is of course caused by the trial field itself which contains a bias. CaPA's goal is also to correct the trial field bias but it is virtually impossible to eliminate it completely. Clearly, if the intention is to sum numerous analyses over a long period of time, we must keep in mind the presence of these sources of bias.

In winter, several precipitation gauges are affected by a problem of precipitation under-catch during snow events in windy weather. To reduce the impact of this negative bias in observations, all those affected by this problem are eliminated. This relatively strict strategy helps maintain the quality of the analysis. In return, there are a lot less stations assimilated in winter.

Contact information

Contact the Climate Services Support Desk

References

Fortin, V., G. Roy, T. Stadnyk, K.Koenig, N. Gasset and A. Mahidjiba (2018). Ten Years of Science Based on the Canadian Precipitation Analysis: A CaPA System Overview and Literature Review, Atmosphere – Ocean, Volume 56, 2018 – Issue 2

Fortin, V., G. Roy, N. Donaldson and A. Mahidjiba (2015). Assimilation of radar quantitative precipitation estimations in the Canadian Precipitation Analysis (CaPA). Journal of Hydrology, 531(2): 296-307.

Lespinas, F., V. Fortin, G. Roy, P.F. Rasmussen,T. Stadnyk, (2015). Performance evaluation of the Canadian Precipitation Analysis (CaPA). J. Hydrometeorol., 16: 2045-2064

Page details

Date modified: