A quantitative evaluation of the high resolution HARMONIE model for critical weather phenomena

E V Van Der Plas,B Wichers Schreur,K Kok

doi:10.5194/asr-8-149-2012

E V Van Der Plas, B Wichers Schreur + Show 1 more

Open Access

https://doi.org/10.5194/asr-8-149-2012

Copy DOI

Abstract

Abstract. The high resolution non-hydrostatic Harmonie model (Seity et al., 2012) seems capable of delivering high quality precipitation forecasts. The quality with respect to the European radar composite is assessed using the Model Evaluation Tool, as distributed by the NCAR DTC (Developmental Testbed Center, 2012), and compared to that of the reference run of Hirlam (Unden et al., 2002), the current operational NWP model at KNMI. Both neighbourhood and object-based verification methods are compared for a week with several high intensity precipitation events in July 2010. It is found that Hirlam scores very well in most metrics, and that in spite of the higher resolution the added value of the Harmonie model is sometimes hard to quantify. However, higher precipitation intensities are better represented in the Harmonie model with its higher resolution. Object-based methods do not yet yield a sharp distinction between the different models, as it proves difficult to construct a meaningful and distinguishing metric with a solid physical basis for the many settings that can be varied.

Highlights

The interest in high resolution numerical weather prediction is mainly driven by the presumed ability to skillfully predict extremes in critical weather situations
When the timing or location of a shower is only a few minutes or kilometers off, a pixel-per-pixel comparison will see this as a double mismatch: a shower is forecasted where it is not observed, giving rise to a false alarm, and the observed shower is not forecasted, counting as a “miss”
The IMPACT project aims to evaluate the high resolution non-hydrostatic model HARMONIE (Seity et al, 2012), being developed by the Hirlam/Aladin consortium, for a series of cases in which the weather was critical to the operations of Schiphol Airport (Amsterdam, The Netherlands)

Summary

Introduction

The interest in high resolution numerical weather prediction is mainly driven by the presumed ability to skillfully predict extremes in critical weather situations. When the timing or location of a shower is only a few minutes or kilometers off, a pixel-per-pixel comparison will see this as a double mismatch: a shower is forecasted where it is not observed, giving rise to a false alarm, and the observed shower is not forecasted, counting as a “miss”. These may be very useful forecasts, giving relevant information to most of the end-users. In this paper a period with several events with convective precipitation will be studied using two verification approaches that try to circumvent the double penalty problem

Methods

Neighbourhood methods

Object-based verification

Observations