This study was motivated by two main concerns including (a) prediction of the Persian Gulf Sea surface temperature (PGSST) anomalies using an autoregressive integrated moving average (ARIMA) model and (b) detection of the climate change signatures in the considered SST data. An ARIMA model was, therefore, developed to predict the SST anomalies having lead times from 1 to 3 months. While the SST time series for the period of 1950-2006 used to fit the model, corresponding records for January 2007 to June 2011 were applied as the test data. The developed model had a minimum value of Akaike information criterion, and its parameters were significantly different from zero. The correlation coefficients between the observed and simulated data for the lead times of 1, 2, and 3 months were found to be significant and equal to 0.72, 0.69, and 0.65, respectively. The corresponding hit rates were estimated as 79, 75, and 72 %, indicating a reasonable forecasting capability of the model. The Heidke's forecast scores were 0.59, 0.52, and 0.48 for the prediction schemes having 1, 2, and 3 months of lead time, respectively. It is shown that the Persian Gulf skin temperatures have warmed up about 0.57 A degrees C during the 732 successive months of the period 1950-2010 noted as an upward significant trend. Although a significant trend was not evident for the 1950-1969 and 1970-1989 period, the PGSST has abruptly increased during the recent two decades. Almost all of the observed warming in the PGSST data is related to this period.