Abstract:Drought indices, such as the Standardized Precipitation Index (SPI) are used to 10 quantify drought severity. Due to the SPI probabilistic and standardized nature, a given value of SPI 11 computed in distinct time periods or locations indicates the same relative drought severity but 12 corresponds to different amounts of precipitation. Thus, the present study aims at contributing for a 13 comprehensive analysis of the influence of long-term precipitation variability on drought 14 assessment by the SPI. Long records of monthly precipitation, spanning from 1863 to 2007 in 15 several locations across Portugal, were divided into 30 years sub-periods and the SPI with 12-month 16 time scale (SPI-12) was computed for each sub-period and for the entire period of records. The 17 probability distributions adjusted to precipitation in those different time periods were compared 18 envisaging to detect the SPI sensitivity to the reference period and, therefore, to changes in 19 precipitation. Precipitation thresholds relative to the upper limits of SPI-12 drought categories were 20 obtained and the influence of the time period was investigated. Results have shown that when SPI 21 values derived from the full data record for a recent time period are lower/higher than the SPI 22 values derived from data of the considered time period a recent downward/upward shift of 23 precipitation has occurred. Coherently, a common pattern of drought aggravation from the initial 24 until the more recent period was not detected. However, in southern locations, lower precipitation 25 thresholds of the SPI drought categories were generally found in the more recent period, 26 particularly for more severe drought categories, whereas in the northern locations Porto and 27Montalegre, an increase was detected. The impacts of the reference period on the computed SPI 28 drought severity and frequency are shown, bringing to discussion the need for updating ´normal´ 29 conditions when long term precipitation records are available and precipitation changes are 30 observed. 31