NONPARAMETRIC ANALYSIS OF CHANGES OF HEAVY PRECIPITATION IN TEXAS

Previous estimates of changes in extreme precipitation in the United States have focused on regional-scale changes or changes at individual gauges with mostly complete, long-term records. Here we adopt a composite rain gauge approach and use nonparametric techniques to identify long-term changes in large daily rainfall amounts. Composites are created on a county-by-county basis, with the average period of record for the composite gauges in Texas being 90 years. Then, for a variety of thresholds, a nonparametric peaks-over-threshold analysis is utilized to identify changes in the frequency extreme precipitation events and to infer changes in extreme precipitation amounts. Most gauges show an increase in frequency over the period of record at all thresholds, with average frequency trends larger at more extreme thresholds.

The county-by-county approach permits analysis of county-scale variations in extreme rainfall trends. These county-scale variations are interpreted as random variations in the extreme precipitation records, and are larger in southern and western Texas than in northern and eastern Texas. The large variations indicate that local historical records of floods are not reliable for inferring robust long-term trends. For example, Houston has seen many damaging floods over recent years, but the change in precipitation extremes in Harris County, where Houston is located, is exceptionally large compared to surrounding counties and is unlikely to continue apace.

Extreme daily rainfall events can be damaging, but the runoff from extreme events is often essential for replenishment of surface water resources during times of drought. Thus these trends have both positive and negative implications for Texas society. The observed rates of increase are consistent with the amount expected from an increase in atmospheric temperatures and from climate model projections, suggesting that the larger-scale historical changes are likely to continue.