Spatial and temporal characteristics of rain-spells in New Zealand

The present study analyzes the spatial and temporal distributions of rain-spells and their characteristics in New Zealand. Various rain-spell parameters such as their average number (NRS), average yield (RSY), average intensity (RSI), and average duration (RSD) and the inter-correlations among all variables are analyzed. Intra-annual variability and the rain-spell characteristics during dry and wet years are presented. Daily rainfall totals from 19 stations on both islands for the period 1965–2017 were used. Rain-spell characteristics were defined using a daily rainfall threshold (DRT) of 1.0 mm. Various functions were fitted to represent the relationship between rain-spell characteristics and their duration-RSD. Dry and wet years were defined according to their standardized departures from the long-term mean: Very dry (VD) when z < − 1.0 Dry (D) when − 1.0 ≤ z < − 0.5 Normal (N) when − 0.5 ≤ z ≤ 0.5 Wet (W) when 0.5 < z ≤ 1.0 Very wet (VW) when 1.0 < z Rainfall totals in the different years were subject to cluster analysis (CA) and the various clusters were mapped. Temporal intra-annual uncertainty was estimated in two ways: (1) by calculating the mid-season date (MSD) and the variability from year to year around this date and (2) by calculating the range of percentages accumulated by the MSD. The main results can be summarized as follows:RSY is the main factor that differentiates between dry and wet years, whereas NRS has only a very limited impact on the annual rainfall.The relationship between NRS and RSD is best described by an exponential curve, between the RSY and the RSD by a linear function and by a power function for the relationship between RSI and RSD.The coefficients of the various correlations in all stations serve to prepare charts of iso-lines of equal NRS, RSY, and RSI respectively for various selected RSDs.Most years were clustered into five different clusters according to their spatial distribution and their return period were calculated. Each cluster presents a different spatial distribution. For each cluster, the appropriate synoptic type according to Kidson classification was attributed.