Patterns of change in high frequency precipitation variability over North America

Patterns of change in high frequency precipitation variability over North America


Play all audios:

Loading...

Precipitation variability encompasses attributes associated with the sequencing and duration of events of the full range of magnitudes. However, climate change studies have largely focused


on extreme events. Using analyses of long-term weather station data, we show that high frequency events, such as fraction of wet days in a year and average duration of wet and dry periods,


are undergoing significant changes across North America. Further, these changes are more prevalent and larger than those associated with extremes. Such trends also exist for events of a


range of magnitudes. Existence of localized clusters with opposing trend to that of broader geographic variation illustrates the role of microclimate and other drivers of trends. Such


hitherto unknown patterns over the entire North American continent have the potential to significantly inform our characterization of the resilience and vulnerability of a broad range of


ecosystems and agricultural and socio-economic systems. They can also set new benchmarks for climate model assessments.


Variability of high frequency precipitation, that is, the variability associated with non-extreme events such as sequencing and persistence of daily precipitation, plays a significant role


in a myriad of terrestrial functions. These include ecosystem and agricultural productivity which are strongly tied to soil-moisture states, biogeochemical processes which are functions of


moisture and temperature states, performance of economic systems which depend on sustained availability of water, etc1. Although recent research has characterized the non-stationarity of


extreme precipitation2,3,4,5,6,7 and its intensification8,9,10 and the change in the mid-range variability such as seasonality11,12,13,14,15, little is known about trends of change in


sequencing of frequent precipitation events arising from climate change and other anthropogenic impacts.


Understanding and accounting for changes in patterns of high frequency precipitation variability has the potential to inform management and design of myriad systems dependent on hydrologic


cycles and to improve predictability of creeping change and associated emergent patterns and risks16. For example, agricultural yields and irrigation requirements are affected by changes in


daily precipitation and its persistence17, 18. If precipitation magnitude is unchanged but precipitation events last longer and are less intense and more frequent, the effect on agricultural


management could be different than if precipitation fell in shorter but more intense and infrequent bouts. Similarly, maintenance and continued efficiency of hydropower plants19 and other


water resources-related systems would be affected by variations in moderate precipitation amounts20 potentially to the point of necessitating the revision of structural design standards or


water management practices21, 22. Aquatic ecosystems are particularly vulnerable to changes in small magnitude events, which can alter natural flow regimes23. Therefore, non-stationarity in


precipitation sequencing, specifically as it pertains to high frequency precipitation, could prove to be a crucial metric to use to strengthen global climate models’ long-term predictions24,


25 and provide much-needed information for a larger network of ecological, social, and economic systems connected by the hydrologic cycle1. Unlike extreme precipitation, the cost associated


with the repercussions from non-extreme precipitation trends is at present unknown. It may be possible to ascertain a monetary cost of variability in non-extreme precipitation associated


with dams and engineering systems, but profound effects on aquatic and terrestrial ecosystems that may not be translatable to direct cost at present could prove to be expensive.


Given the consequential impact of changes in patterns such as the fraction of rainy days in a year and lengths of consecutive wet and dry periods, we investigate the presence and trends of


change in such characteristics using long-term data from raingage measurements over North America. Some studies of precipitation persistence conducted in Europe compare seasonal changes in


wet and dry periods’ duration and occurrence with cyclone activity and temperature trends26 as well as precipitation duration and intensity27, 28. One global study of temporal distribution


of precipitation emphasizes the importance of light and moderate rainfall events29. Our study is the first of its kind to study sequencing pattern changes specifically focusing on the


effects on non-extreme precipitation across all of North America, although some precipitation studies have focused on smaller sections of the continent30,31,32,33 or have focused on extreme


precipitation in the United States34. Studying the variability at such a large scale enables us to incorporate extensive raingage data (over 3,000 stations) and allows for spatio-temporal


analysis on geographic scales ranging from continental to Level III Ecoregions35. Two studies, based in the northeastern United States and central United States, studied the changes in


distribution of intense precipitation by setting a minimum threshold of the station’s mean precipitation value32 and at fixed values corresponding to precipitation ranges defining


“moderately heavy”, “very heavy”, and “extreme” precipitation36. The northeast US study compared the persistence of wet and dry periods of mean and extreme precipitation in order to compare


with changes in total annual precipitation amounts32. The study concludes that non-stationarity in precipitation trends is present, noting that both wet persistence and the 95th percentile


of daily precipitation are increasing. The central United States-based study compares changes in frequency of intense precipitation to several climatological factors, such as tropical


cyclone activity and mean annual temperature. Authors show that “very heavy” and “extreme” rain days have become more frequent but that rain event characteristics such as duration and peak


hourly rain intensity remain unchanged36. Another study based over all of North America analyzes changes in duration of both warm seasons and the dry spells within them at a regional scale


by defining minimum daily precipitation thresholds based on both precipitation amount and the corresponding daily temperature37. This study demonstrates an increase in persistence of dry


periods in eastern and southwestern regions of the United States in the past 40 years.


Drawing on precipitation data from several thousand stations, we choose to translate exceedance or non-exceedance of daily precipitation thresholds to a binary sequence, which allows us to


study sequencing patterns, precipitation persistence, and changing annual fractions of days above chosen precipitation thresholds. This method, therefore, includes the important effects of


the full range of magnitudes of precipitation and does not exclude the effects of non-extreme, or high frequency, precipitation in the analysis of long-term trends. We are then able to


compare independent, long-term trends for both fraction of wet days and persistence of wet and dry periods, and additionally, we can compare those to changes in daily rainfall magnitude.


Through this comparison, trends in changes of sequencing in high frequency precipitation events are examined across a large geographic scale.


We explore the hypothesis that average daily precipitation, fraction of days in a year receiving precipitation, and average length of consecutive wet and dry periods, may have independent


trends at any station, and the clustering of stations with similar behavior reveal spatially coherent trends. Our findings also lead us to investigate the potential role of regional climate


or microclimate as an explanatory variable for changes in local high frequency precipitation patterns.


Daily precipitation data from the Global Historic Climatology Network (GHCN)38, one of the most complete global collections of daily precipitation data39, for 7,194 stations in North America


and Hawaii was made available through Earth Info, Inc. The data set was subjected to numerous, thorough quality control procedures regarding both temperature and precipitation39, 40, and


has been deemed appropriate for studies that analyze trends in light and heavy precipitation41, 42. Data was recorded with a precision of up to a tenth of a millimeter (0.1 mm). In the GHCN


data set, days with trace amounts of precipitation were flagged and assigned a zero value38. Due to the presence of missing or unusable data, daily precipitation records for each station


were filtered based on the criteria in Table 1. Stations meeting these requirements numbered 5,259.


Additionally, a station’s years of coverage were required to not end before the year 2000, which increases the chances stations will overlap in years of coverage so that fair temporal


comparisons can be made and with the hope that trends leading up to present-day may aid in future predictions. It is also assumed that “younger” stations would be better maintained and more


numerous in general. This additional requirement reduced the number of stations passing the quality control criteria (Table 1) to 3,030.


Of the stations that did not pass filtering, 21% were eliminated outright because the total number of available years (usable or not) was less than the minimum requirement of five decades.


Several stations fell into clusters in remote locations, such as the plains areas in southern Saskatchewan, Canada, which suggests the possibility that lack of ease of accessibility


contributed to instrument errors or calibration issues, causing breaks in data recording. In addition, stations in Canada have undergone changes in observational practices independent of


those in the United States and Mexico, and such station inhomogeneity could introduce bias in data records. However, the number of stations that pass the filtering tests in Canada are only a


small fraction of the total number of stations analyzed. The influence of other possible sources of inhomogeneity in station data across all of North America, such as changes in time of


observation, were thoroughly considered and do not diminish the robustness of this study. The locations of passing and failing stations are shown in Fig. 1, where filled circles indicate


stations whose data passed all analysis requirements, and empty circles indicate failing stations. After filtering, the years of coverage for passing stations fell between 1880 and 2010. The


Fig. also indicates the number of stations with different lengths of coverage.


Locations of North American rain gauge stations. Stations passing the quality control criteria listed in Table 1 are color-coded based on the years of coverage as indicated in the inset.


Most stations whose data coverage begins in 1920 or earlier are located in the Midwest, while more recent stations appear to be evenly distributed. The majority (60%) of stations were


established in 1940 or later. Empty circles indicate stations whose data failed to meet quality control criteria and are excluded from the analyses. In the inset, the bar beginning in 1900


represents stations whose data start in 1900 or earlier, as there are a small number of stations passing quality control criteria whose data start between 1880 and 1900. Also, the number


listed to the right of each bar indicates the number of stations with data available for the time period with those particular start and end years. In the analysis, stations with data ending


in 2000 or 2010 are combined based on start year because both sets meet criteria to be representative of that particular period (see Materials and Methods). [Map produced with software


ArcMap v. 10.4.1, http://desktop.arcgis.com/en/arcmap/].


Statistically significant trends in average daily precipitation (ADP(y)) were detected in 1,182 stations. Stations that indicated decreasing ADP(y) (red symbols) numbered 136 and are mostly


located in the Southeast, Pacific Northwest, and several other small pockets. Positive trends in ADP(y) (blue symbols) appeared in 1,046 stations located in the U.S. Midwest, South, and


Northeast. Symbol size is proportional to trend magnitude. Overall, stations east of the 100th meridian west showed the strongest increases in average daily precipitation. [Map produced with


software ArcMap v. 10.4.1, http://desktop.arcgis.com/en/arcmap/].


These observations suggest the possibility that different vegetative covers and topographies could play a more prominent role than previously understood in long-term precipitation sequencing


patterns and emphasize that large-scale regions cannot be generalized in climate predictions given the evident strength of some local climates. Consideration of long-term sequencing trends


on a local or regional scale suggests that long-term trends in precipitation take on a more complex form, one that is likely determined by a combination of interconnected local climate and


anthropogenic factors. The observations presented in this analysis of more nuanced variability in trend highlight the importance of incorporating high frequency precipitation variability as


a performance metrics for climate models. However, further study is necessary to characterize the cause of these local trends.


This work was partially supported by Univ. of Illinois Graduate College Fellowship, a SURGE Fellowship, and NSF Grants CBET 1209402, EAR 1331906, and EAR 1417444.


Department of Civil and Environmental Engineering, University of Illinois at Urbana Champaign, Urbana, 61801, USA


Department of Atmospheric Sciences, University of Illinois at Urbana Champaign, Urbana, 61801, USA


P.K. and S.R. designed the study. S.R. conducted the analysis. Both discussed the findings and contributed to the writing of the manuscript.


Anyone you share the following link with will be able to read this content: