Difficulties commonly encountered in precipitation measurement by radar include errors from radar reflections from the surface, errors in extrapolating from measurements aloft, and errors through inadequately sampling a fluctuating signal. These error sources are discussed, along with the skill of the solutions that have been implemented. Copyright © 2005 Royal Meteorological Society

Techniques are assessed for analysing the skill with which weather radar data can be extrapolated to provide short-term rainfall forecasts. In addition to visual inspection, forecast skill is assessed using catchment-averaged statistics - comparing analyses with rain gauge averages, forecasts with analyses and forecast river flow with measured flow. Copyright (C) 2005 Royal Meteorological Society

The high temporal variability of rainfall requires that measurements be taken at a high frequency for that variability to be recorded well, but most conventional gauges do not have this capability. The gauge described has a resolution estimated to be about 6 s and its measurements do capture the variability. Furthermore, the large datasets associated with high-frequency measurements can be avoided and the essential information retained by storing the data as breakpoints. These are a series of data pairs, each of which consists of the rain rate itself and the time when that rate commenced.A gauge in which the collected rain is formed into a series of drips all of approximately the same known size was a practical choice. The design and calibration of this gauge, in which particular attention was paid to producing a robust instrument incorporating standard components wherever possible, is described. Long-term comparison in the field with a collocated tipping bucket gauge was used in a calibration scheme in which equality between the gauges was sought for both the long-term accumulation and the short-term rain rates. Although the gauge depends on the formation of equisized drips, it was found that drip size increased slowly with rain rate, and so two calibration parameters were required to convert the time interval between drips into the mean rain rate between the drips. After an initial aging period with inconsistent drip formation, the calibration was stable and the onset of the lowest rain rates (0.1 mm h(-1)) could be determined to within 1 or 2 min and streaming (i.e., the series of drips merging into a continuous stream) did not occur for rain rates less than 100 mm h(-1). Some sample applications for the gauges are described: the extraction of breakpoints, the estimation of 1-min rain rates from Dine's tilting siphon data, and their use in a field experiment.

A new approach is presented to the problem of correcting radar estimates of surface rainfall for errors resulting from the changes in reflectivity between the surface and the height at which the radar measures reflectivity. In this approach, a set of five typical shape functions (TSF) of a sample of 89000 vertical profiles was found. These vertical profiles were formed from volumes scan information, interpolated onto 10 levels, at 250 m spacing, with level 6 being set to the height of the top of the bright band. The TSF class for each vertical profile was determined from a limited set of observations of reflectivity aloft using Bayesian discriminant functions. Regression equations were developed for each TSF class to estimate the lowest interpolated level reflectivity (level 1) from these observations aloft. Application of the discriminant functions shows that the 62% of the dependent sample could be correctly classified into its TSF class for data at heights greater than 500 m below the bright band. In four of the five TSF classes, there was significant skill in estimating the surface reflectivity, explaining 80% of the variation of the level 1 reflectivity. The fifth class explained 49% of the level 1 variation. On average, using this approach of determining the TSF for each vertical profile rather than using a single profile for each volume scan increased the variation of the surface reflectivity explained by observations aloft from 59.5% to 77.4%. Information on the skill of estimating the surface reflectivity was also produced, using the posterior probability and the explained variance of the regression equations.

This paper reports on the first application of a multispectral textural Bayesian cloud classification algorithm ("SRTex'') to the general problem of the determination of high-spatial resolution cloud-amount and cloud-type climatological distributions. One year of NOAA-14 daylight passes over a region of complex topography (the South Island of New Zealand and adjacent ocean areas) is analyzed, and exploratory cloud-amount and -type climatological distributions are developed. When validated against a set of surface observations, the cloud-amount distributions have no significant bias at seasonal and yearly timescales, and explain between 70% (seasonal) and 90% (annual) of the spatial variance in the surface observations.The cloud-amount distributions show strong land/sea contrasts. Lowest cloud frequencies are found in the lee of the major alpine feature in the analysis domain (the Southern Alps) and over mountain-sheltered valleys and adjacent sea areas. Over the oceans, cloud frequencies are highest over sub-Antarctic water masses, and range from 90% to 95%. However, over the sea adjacent to the coast on the western side of the Southern Alps, there is a distinct minimum in cloud amount that appears to be related to the orography.The cloud-type climatological distributions are analyzed in terms of both simple frequency of occurrence and conditional frequency of occurrence, which is the frequency of occurrence as a fraction of the total number of times that the cloud type could have been observed. These distributions reveal the presence of preferred locations for some cloud types. There is strong evidence that uplift over major mountain ranges is a source of transmissive cirrus (enhancing occurrence by a factor of 2) and that the resulting cirrus coverage is most extensive and frequent in spring. Over the ocean areas, SST-related effects may determine the spatial distributions of stratocumulus, with higher frequencies observed over sub-Antarctic waters than over subtropical waters. Also, there is a positive correlation between mean cloud-top height and SST, but no similar relationship is found for other cloud types.

The melting of snow as it falls through the 0 ° C level is a significant meteorological process that is important for its impact as the bright band of enhanced reflectivity in radar observations. Thus, it is necessary to understand the variability of the phenomena and to determine the factors upon which it depends. This paper reports on preliminary investigations into the observations of the bright band over the UK using vertically pointing radar. These results are compared with output from a simple model of the melting of snowflakes and with other observations from Canada and the Netherlands. The vertical depth of the bright band was determined from the vertical pointing radar data for four cases of widespread frontal rainfall. An increase in the depth of the bright band was seen with increasing background reflectivities. Depths of 100–150 m at 10 dBZ increased to 200–400 m at 25 dBZ. Results from a simple model of the melting of snowflakes were compared with the vertical pointing radar observations. Similar trends were seen in the model output, but in general the model produced deeper but less intense bright bands. Notable in the model results was the lack of strong dependence of the depth on vertical air motions. Indeed, the bright band depth only increased by approximately 30 m in a downdraft of 1 m s −1 . Comparisons of the bright band characteristics with other observations from elsewhere show that the bright band depth was similar to that observed by Klaasen (1988) in the Netherlands, but shallower than those observed by Fabry & Zawadski (1995) in Canada. Copyright © 2001 Royal Meteorological Society

The breakpoint format for rainfall data, by recording the rainrate and the times when the rainrate changes, consists of variable length periods during which rain of constant intensity fell. It is compact and retains a level of detail equivalent to that found in high temporal resolution data. However, breakpoint data are available only at the relatively few locations where automatic raingauges produce pluviographs. On the other hand, radar observations of rainfall provide precipitation data at a high spatial density but at poor time resolution. Advantage can be taken of this high spatial density to give more locations where breakpoints are available through interpolation between the radar images to provide precipitation data at sufficient temporal resolution for the extraction of breakpoints. The interpolation scheme, based on the PARAPLUIE method, detects the movement and evolution of precipitating echoes to provide additional interpolated radar images. Breakpoints, extracted at pixels of the images that contain raingauges, showed similar overall accumulations and characteristics when compared with those of the collocated raingauges. The similarity is particularly so for gauges within a 50 km radius of the radar where the spatial resolution of the radar and the temporal resolution of the raingauges are equivalent. Copyright © 2001 Royal Meteorological Society

The Otaki Precipitation Estimation by Radar (OPERA) programme was designed to investigate the processes that lead to enhancement of rainfall over the Tararua ranges of New Zealand. These ranges rise to 1500 m above the coastal plain and enhancement of rainfall by windflow over these hills leads to annual hill‐top rainfall of over four times that upwind. The OPERA experimental campaigns aimed to characterise the enhancement processes by analysing data collected from a transect of high‐resolution rain gauges and a locally deployed, high‐resolution radar, supported by scanning radar and satellite observations. Measurements made during these experiments showed that orographic enhancement led to hill‐top accumulations often twice that upwind, and up to as much as a factor of seven in one case. The data suggest that the most frequent occurring enhancement mechanism was triggered convection. This mechanism leads to an increase in rainfall over the hills of around a factor of two, primarily through an increase in the duration of rain. Seeder/feeder‐type enhancement occurs less frequently but leads to larger enhancements. Copyright © 2000 Royal Meteorological Society

This paper identifies relationships between air mass properties and mesoscale rainfall when moist air blows over New Zealand's Southern Alps from the Tasman Sea. Around 50% of the variance in six-hourly rain volumes summed across three separate cross-mountain raingauge transects and in six-hourly rain volume spilling across the alpine divide are statistically explained by the following properties of the approaching air mass: relative humidity, wind velocity normal to the mountains, air mass stability and synoptically induced upward motion. These factors also explain about 25% (r≈0.5) of the variance in the downwind distance reached by the spillover rainfall. For the highest 10% of six-hourly rainfalls, spillover distance and magnitude are negatively correlated with the 700 or 500 hPa temperature. Multiple linear regression equations suitable for predicting rainfall intensity and spillover are developed. A progression is described in the magnitude and depth of vertical motion and resulting condensation rates over the mountains as the properties of the incoming air mass evolve through a storm. These changes, together with greater downwind advection of ice particles compared to raindrops, explain the observed statistical relationships between the air mass properties and mountain rainfall.

Bayesian methods are used to develop a cloud mask classification algorithm for use in an operational sea surface temperature (SST) retrieval processing system for Advanced Very High Resolution Radiometer (AVHRR) area coverage (LAC) resolution data. Both radiative and spatial features are incorporated in the resulting discriminant functions, which are determined from a large training sample of cloudy and clear observations. This approach obviates the need to specify the arbitrary thresholds used by hierarchical cloud-clearing methods, provides an estimate of the probability that an instantaneous field of view is cloudy (clear), and allows the skill of different cloud discriminant models to be objectively analyzed.Results show that spatial information is of particular importance in reducing the false alarm rate of the cloudy class. However, while the use of complex textural measures such as gray-level difference statistics-as opposed to simple statistics such as the standard deviation-improves the skill of nighttime cloud-masking algorithms, they are of little advantage during daytime hours.Cloud mask discriminant models having similar high Kuipers' performance index scores (i.e., 0.935) are developed for both day and night satellite data from the Southern Hemisphere midlatitudes. Applied to LAC orbital (i.e., operational) data, the characteristics of the cloud masks appear to be similar to those derived from analysis of the training sample data. However, in this case. to enhance processing performance, a hybrid algorithm is employed-obviously cloudy instantaneous fields of view (IFOVs) are first removed via a gross Ihreshold check and the Bayesian method applied only to the remaining IFOVs. This same (hybrid) algorithm is also applied to an ensemble of 30 days of AVHRR LAC data from the New Zealand region. Analysis of the resulting time-composited SST data (means and standard deviations) shows there is little evidence of a day-night bias in the performance of the Bayesian cloud-masking algorithm and that the resulting SST data may be used to determine the variability of oceanographic features.Although this paper uses AVHRR data to demonstrate the principles of the Bayesian cloud-masking algorithm, there is no reason why the approach could not be used with other instruments.

A rain gauge is described that quantizes rainwater collected by a funnel into equal-sized drops. Using a funnel of 150-mm diameter, the quantization corresponds to 1/160 mm of rainfall, enabling the measurement of low rainfall rates and the attainment of a fine temporal resolution on the order of 15 s without unduly large sampling errors. Two drop-producing units are compared and an operational rain gauge design is presented. Field comparisons with conventional rain gauges are made, showing excellent correlations for daily rain totals, and intercomparisons between clusters of dropper gauges are also given. Examples of highly resolved rainfall events are shown demonstrating the ability to measure low rainfall accumulations and also coherent high intensity events of short duration, which are not detectable with conventional rain gauges.

Much of New Zealand's precipitation is modulated by windflow over its orography. Rain-gauge networks are normally sparse, and because measurement programmes are relatively expensive, there are considerable benefits in modelling and mapping spatial patterns of precipitation in order to provide information at sites where no data exists. In this paper a simple, numerically robust diagnostic model (VDEL) for estimating orographic precipitation is used to estimate annual precipitation normals for a mountainous region of the southern North Island, New Zealand. The VDEL model, developed initially to analyse storm precipitation patterns, is based on the estimation of orographically forced vertical motion as V-s.del Z(s) where del Z(s) is the height of the terrain and V-s the low-level horizontal windflow upstream of the mountain range. Precipitation patterns were diagnosed by VDEL for four main windflow directions and calibrated against annual precipitation normals by means of a multivariate regression mapping function. Areal average precipitation, estimated from catchment runoffs, was reproduced with errors in the 5 to 10 per cent range. (C) 1997 by the Royal Meteorological Society.

A simple potential flow model is developed for airflow over low hills. The model readily allows incorporation of simple rain microphysics so that drop trajectories can be traced in the region of a hill complex. Rainfall intensity variations are found in the horizontal and in the vertical, and a straightforward modification to the model allows for time dependence to be studied.The model is shown to serve as a useful diagnostic tool to describe rain drop redistribution due to perturbed airflow. In particular, increased rainfall is predicted in the lee of hill peaks and depleted rainfall is predicted on ridges. The redistribution effects dominate at low levels, in contrast to predictions based on seeder-feeder enhancement.Comparison with field measurements shows good qualitative agreement over low hills and, in some cases, close quantitative agreement. Redistribution of rain appears to be a significant process that should be considered in rainfall interpolation and prediction schemes involving hilly terrain. (C) 1997 Elsevier Science B.V.

Rain gauge, radar, and atmospheric observations during a prolonged northwesterly storm in November 1994 have been used to study factors influencing the distribution of precipitation across the Southern Alps, Despite the persistent northwesterly flow, the location and intensity of precipitation varied markedly during this storm, providing an excellent dataset for these investigations. Data from 36 recording gauges in the northern half of the Alps were supplemented by data from 57 daily gauges, which were partitioned into 6-h values. These data were grouped according to distance from the alpine divide, and best-fit transect curves, normalized for rainfall intensity, were established every 6 h. The fraction of the total transect precipitation falling in leeside catchments varied between 0.11 and 0.70, while a ''spillover distance'' index varied between 6 and 29 km. Comparison with atmospheric profiles of temperature and wind from Hokitika on the west coast of New Zealand and with European Centre for Medium-Range Weather Forecasts analyses revealed that precipitation was confined upwind of the divide during a period of blocked flow near the start of the storm, and only extended into leeside catchments with the onset of stronger flow and reduced static stability. Regression equations involving these factors explained up to 93% of the spillover variations. It is suggested that ascent and precipitation maxima an shifted upstream during blocked flow, while spillover is enhanced during stronger and/or unstable flow as the upstream influence lessens and snow and ice particles drift farther downwind before falling below the freezing level. Further case and modeling studies are needed to demonstrate the wider applicability of these findings.

Twelve months of Southern Hemisphere (maritime) midlatitudes Advanced Very High Resolution Radiometer local area coverage data at full radiometric and spatial resolution have been collocated with rain-rate data from three Doppler weather radars.Using an interactive computing environment, large independent samples of cloudy-altocumulus, cumulonimbus, cirrostratus, cumulus, nimbostratus, stratocumulus, stratus-and cloud-free scenes have been identified (labeled) in the collocated data. Accurate labeling was ensured by providing a supervising-analyst access to appropriate diagnostics, including difference and ratio channels, 3.7-mu m reflected and emissive components, spectral histograms, Coakley-Bretherton spatial coherence plots, mean, standard deviation, and gray-level difference (GLD) statistics. This analysis yielded 4323 cloud and no-cloud samples at a spatial resolution of 8 x 8 instantaneous fields of view (IFOV), from 257 NOAA-11 and NOAA-12 orbits.Bayesian cloud discriminant functions calculated from the labeled samples and utilizing feature vectors including radiometric and GLD spatial characteristics successfully classified scenes into one of the seven cloud and no-cloud classes with significant skill (Kuipers' performance index 0.63). Utilizing the posterior probability of the classified samples enabled some clouds that were classified erroneously to be identified (and discarded), improving the skill of the discriminant functions by an additional 10% or so. Removing the GLD statistics from the feature vector reduced the skill of the cloud discrimination by about 20% (relative to the nondiscarding discriminant function), while increasing the misclassification of midlevel clouds. However, some cloud classes can only be discriminated from their multispectral signatures. Day and night discriminant functions show similar skill.Within raining cloud classes, rain rate has been related to the spatial and radiometric characteristics of the cloud. The skill of the rain-rate estimates is dependent on the cloud type. For nimbostratus and altocumulus classes 20%-25% of the rain-rate variation can be explained by predictors that measure the temperature, spatial texture, and degree of isotropy in the sampled clouds. Raining and nonraining Samples of altocumulus, cumulus, cirrostratus, and nimbostratus can be delineated with at least 60% accuracy.This approach, whereby cloud classes are identified then rain rates estimated as a function of cloud type, would seem to resolve some of the usual problems associated with rain-rate analyses from midlatitudes infrared and visible satellite data. It also extends rain-rate diagnosis to nonconvective (frontal) cloud systems.

A preliminary investigation of rainfall distribution over a raingauge array situated in the Waitakere Ranges of West Auckland, North Island, New Zealand, during the passage of four cold fronts, has revealed a systematic oro- graphic influence on two occasions. The rainfall on these occasions increased substantially with distance from the coast, whether the gauge array was, by virtue of the low-level wind direction, on the lee side or windward side of the hills. The present studies suggest strongly that rainfall rates at sites with different enhancements are linearly related. Catch ratios of up to 3:1 were measured for offshore winds. These studies are distinguished by the dense network of gauges used.

It is found that substantial errors can arise when rainfall intensities are estimated from a single digitisation pass over a raingauge chart. The errors can be reduced to acceptable levels either by low‐pass filtering or coherent averaging. The advantage of the former method is that only a single digitisation pass is required, whereas the latter method has the advantage that time resolution is not lost. Tests on real rainfall chart data show that coherent averaging over three passes is equivalent to low‐pass filtering having an order of magnitude reduction in response time. It is therefore recommended that coherent averaging be adopted as a method of improving rainfall data quality.