Precipitation in the Waikato River Catchment

Observational data and numerical models were used to investigate precipitation in and around the Waikato River catchment. This economically important catchment relies on a dependable precipitation supply for agriculture and hydroelectric generation, with stations generally receiving 2,000 +/- 300 mm of precipitation annually. Long-term and inter-annual variability of total and extreme precipitation were examined using up to 100 years of observational data. Precipitation volumes within the catchment were represented by a five-day smoothed, area-averaged time series, and extreme events were defined as exceeding the 95th percentile. Atmospheric circulation oscillations correlated with the frequency of light precipitation but not with the probability of occurrence or with the magnitude of heavy precipitation events. Also no significant linear variations in precipitation (either annual totals or extreme precipitation characteristics) were found over this period, although temperature increased by 1.15+/-0.45'. A total of 63 heavy precipitation events were identified between 1996 and 2001. An analysis of the prevailing synoptic conditions reveal that heavy precipitation was associated with the passage of cold fronts of cyclones with minima at both 500 and 1000 mb heights. Extended periods of enhanced baroclinicity (succession of cyclones) or blocking anticyclones east of New Zealand have led to flooding in the Waikato catchment. Storm tracking showed that 10% of cyclones originating in the Tasman Sea result in heavy precipitation in the catchment. The accuracy and value of the GFS global precipitation forecasts <= 180 hours were investigated. Depending on forecast lag, the global models correctly predicted the presence of precipitation in 70-80% of forecasts, but the magnitude and distribution were often inaccurate. The probability of receiving precipitation is increased when more members of a lagged ensemble predict it. Forecasts with lags shorter than approximately 96 hours were appropriate to use as boundary constraints for mesoscale modelling. The ability and limitations of mesoscale models to simulate the spatial distribution of precipitation were examined through high-resolution WRF simulations of three heavy precipitation events, and ten different model settings were compared for the January 2006 event. The model consistently under-predicted precipitation. The timing and location of convective precipitation, which accounted for 50% of the precipitation during two events, was physically unconstrained but regional totals were comparable to observations. A continuous two-year numerical simulation was run to provide a precipitation climatology for data-sparse areas. The simulation gave good spatial representation of precipitation and other meteorological variables but tended to under estimate the magnitude of heavy precipitation and over-estimate light precipitation.