Abstract
The present study evaluates the potential predictability, and prediction skill of surface air temperature (SAT) over the Arabian peninsula (AP) referred to as AP-SAT hereafter, during boreal summer (June–July–August: JJA) from 1982 through 2017. The study was made by considering the single model, and a multimodel ensemble (MME) approach. The seasonal prediction data for JJA SAT initialized at May (Lead-1), and April (Lead-2) observed initial conditions, from six coupled atmosphere–ocean global circulation models included in the North American Multimodel Ensemble, is utilized. The potential predictability (PP) is estimated through the estimation of the signal-to-noise ratio (S/N ratio) and perfect model correlation (PMC), while prediction skill is computed by the temporal anomaly correlation coefficient (TCC). All models show a decrease in potential predictability and prediction skill with an increase in lead time. The CFSv2 and NASA models show higher PP, which indicates a high potential predictive skill for summer AP-SAT in these models. However, both models show quite low values of TCC all over the AP domain, which is an indication of overconfident predictions. The three geophysical fluid dynamics laboratory models show high prediction skill at both leads. An essential finding of the predictive analysis (PMC and TCC) is that the MME does outperform the individual model at both leads for summer AP-SAT. Each model captures the observed relationship between spatially averaged AP-SAT with sea surface temperature (SST) and 200 hPa geopotential height (Z200) during JJA, with varying details. Persistent model biases impact negatively model predictability and skill, and better AP-SAT and SST teleconnection pattern in models lead to higher predictability. Improvements in initial conditions, model physics, and larger ensemble size are necessary elements to enhance the summer AP-SAT potential predictability and skill.