Temperature and Precipitation in the US States: Long Memory, Persistence and Time Trend

This paper analyzes the time series properties of the anomalies of temperature and precipitation the US states by using techniques based on fractional integration. In doing so, we are able to capture time trend components along with properties such as long memory and persistence. Our results show that for the aggregated data, long memory is presence in both temperature and precipitation anomalies since the order of integration is significantly positive in the two cases. The time trend is also positive, being higher for the temperature anomaly. Looking at the data disaggregated by states, for the temperature anomaly, there are only seven states where the time trend is found to be insignificant, with most of them located in South East areas, while for the rest of the cases, the time trend is significantly positive. Long memory is present in all cases, though the differencing parameter substantially change from one state to another ranging from 0.09 in Nebraska and Kansas to 0.18 in Florida and Michigan. For the precipitation anomaly, the time trend is found to be insignificant in a large number of cases, and the degree of integration is smaller than for the temperature. In fact, short memory cannot be rejected in fourteen states, and the highest orders of differencing are obtained at Arizona (0.11) and Texas (0.12). In general, we highlight that one cannot draw conclusions about persistence and trends in these two climate-related variables based on aggregate information of the overall US, given widespread heterogeneity across the states. Moreover, the degree of persistence across the states seem to be negatively correlated with their level of preparedness in terms of handling climate change.