Search

It is well known that as a consequence of climate change, higher temperatures are causing extreme precipitation events to intensify, leading to greater flooding. However, the relationship between temperature and the temporal distribution of precipitation within storms is not well understood, with limited research focus on precipitation event loading or where the bulk of the precipitation occurs within the storm duration. Here, we investigate the relationship between temperature and the temporal pattern of precipitation, with a focus on event loading. Historical trend analyses based on station observations reveal that precipitation events have become increasingly front loaded (i.e., a greater percentage of precipitation falling earlier in the storm) across Australia over the past six decades. This increased frontal loading of precipitation events coincides with increasing trends in representative storm temperatures, with higher temperatures associated with a greater proportion of short-duration convective events. Linking these precipitation events with the representative storm temperatures shows that precipitation events become more front loaded with increasing temperature across nearly all event durations and intensities, with the emphasis on shorter duration (<6 h) events in the tropics. There is a clear systematic shift toward more front-loaded temporal patterns of precipitation with increasing temperature, coupled with intensification of embedded bursts. These results have implications for potentially increased flooding, with hydrological applications needing to account for nonstationarity in the temporal pattern of precipitation. Significance Statement: To date, there is very little understanding of how temporal patterns of precipitation events change with increasing temperatures. Here, we investigate the relationship between temperature and the temporal pattern of precipitation events with a focus on the timing of when the bulk of precipitation occurs (termed event loading). Our results indicate a clear systematic shift toward more front-loaded temporal patterns of precipitation with increasing temperature, coupled with intensification of embedded bursts. Greater shifts in temporal patterns of precipitation are observed for shorter-duration precipitation events, particularly in the tropics. The impact of changing temporal patterns of precipitation on flood estimation will require careful examination due to the risk of increased flood peaks. [ABSTRACT FROM AUTHOR]

• Event-based rainfall statistics are very sensitive to instrumentation changes. • Rainfall data should be aggregated to match the coarsest instrument resolution. • Small rainfall amounts should be removed to ensure homogeneity. Sub-daily rainfall is used in modelling of small catchments, design of urban drainage, calculating soil erosion, understanding the water balance of vegetated catchments, and assessing the impact of climate change on short duration storms. Hence it is critical to ensure that sub-daily rainfall records are homogeneous, but with a global shift to automated measurement of rainfall, there has been a change in the rainfall recording instrumentation. Although changes in instrumentation should be documented, in many cases this is not the case and station metadata are missing. As result, studies of rainfall patterns, particularly those investigating trends, often cannot report their treatment of possible inhomogeneities and must assume the data is suitable for further analysis. Here, we argue that standard quality assurance methods for checking inhomogeneities in rainfall may not identify a change in the rainfall record due to a change in instrumentation. Through testing (1) an aggregation of rainfall to the coarsest instrument resolution, (2) removal of rainfall below a minimum rainfall depth, (3) removal of rainfall below a minimum accumulation (event) depth, and (4) a combination of the above adjustments, we present recommendations for ensuring sub-daily rainfall is homogenous with minimum alteration to the rainfall record. We assess the proposed methods using the case study of Australia's sub-daily rainfall monitoring network, where, in 1996 the pluviograph network was switched to tipping buckets, creating a large-sample test bed with a known systematic instrumentation change. Our results show that event-based statistics are considerably more sensitive to instrument change than annually aggregated rainfall statistics and hence standard methods for quality assurance may not identify possible inhomogeneities. We suggest an aggregation of the data to the coarsest instrument resolution, with the removal of small rainfall accumulations, can alleviate inhomogeneities for event-based statistics, whereas, for annual statistics, aggregation to the coarsest instrument resolution is sufficient. In the absence of a known instrumentation change, removing rainfall at or below the instrument resolution is also a viable technique for improving rainfall record homogeneity for event-based statistics. These findings will aid future sub-daily rainfall studies by justifying the use of a minimum rainfall or event depth for subsequent analysis. [ABSTRACT FROM AUTHOR]