Predictability of Heavy Precipitation in the Waikato River Basin of New Zealand

Abstract

This paper investigates the predictability of heavy precipitation in the economically important Waikato River basin of New Zealand. A 2-yr archive of Global Forecast System (GFS) model data to 1180 h for the period August 2005-August 2007 forms the basis of the study. GFS model predictions of precipitation are compared to surface measurements from 22 stations in and around the river basin. Categorical hit rate and bias, threat score, false-alarm ratio, probability of detection, RMSE, skill score, and mean error are all plotted as a function of model lead time for the 2-yr period.The general synoptic structures of heavy precipitation events are often identified at long lead times. An example is shown for January 2006 when heavy rainfall is identified by GFS at 1150 h, accurate location and timing is given at 148 h. Precipitation amounts are systematically underpredicted by GFS and the spatial distribution of rainfall is limited by model resolution. The value of GFS lies in its ability to provide early warning of potential heavy precipitation.Three heavy rainfall events were selected for higher-resolution Weather Research and Forecasting (WRF) model simulation. The WRF model produced more realistic geographical distributions of precipitation but cannot compensate for errors in the global model. WRF underpredicts precipitation for all three cases; the best simulation generates 92% of observed precipitation.Unobserved spurious convective rainfall can be generated by the WRF model following the passage of a frontal weather system. A possible mechanism is suggested that links this to underpredicted frontal rainfall.

Influence of bulk microphysics schemes upon Weather Research and Forecasting (WRF) version 3.6.1 nor'easter simulations

“…6 and 7 show WRF to favor higher precipitation amounts and is consistent with the positive bias scores in Table 3. Previous modeling studies of strong convection by Ridout et al (2005) and Dravitzki and McGregor (2011) found that both GFS and the Coupled Ocean-Atmosphere Mesoscale Prediction System produced too much light precipitation and too much heavy precipitation, which contrast with the above results. Unlike these two studies, nor'easters track too far offshore to be fully sampled by rain gauge data and S-band weather radars.…”

Sensitivity Study on High-Resolution WRF Precipitation Forecast for a Heavy Rainfall Event

“…Wang et al [9] examined the sensitivity of heavy precipitation to horizontal resolution, domain size, and rain rate assimilation for case studies using a convection-permitting model. Dravitzki and McGregor [10] investigated heavy rainfall events over the Waikato River Basin of New Zealand generated with higher-resolution WRF, and Goswami et al [11,12] showed that domain size is as important as grid spacing and initial conditions for heavy rainfall events. Additionally, Li et al [13] analyzed the influence of horizontal resolution, domain size, and physical parameterization schemes to evaluate an optimized WRF precipitation forecast over a region of complex topography during the flood season.…”

Developing machine learning tools for long-lead heavy precipitation prediction with multi-sensor data

“…A Global Forecast System model was proposed to predict the flood in in Waikato River basin of New Zealand. However, the amount of precipitation was significantly underestimated [4]. A European Flood Forecasting System was developed to determine the skill for flood forecast [9].…”