Improved Automated Surface Temperature Guidance

Abstract

Currently, two sets of automated numerical-statistical forecasts of maximum/minimum (max/min) temperatures for calendar day periods are produced in the day-to-day operations of the National Weather Service. The ?early? guidance forecasts are based on output from the Limited-area Fine Mesh (LFM) model, while the ?final? guidance relies primarily on predictions from the hemispheric Primitive Equation (PE) model. This paper describes recent improvements to the early guidance surface temperature prediction system. The Techniques Development Laboratory recently developed new early guidance equations to forecast calendar day max/min temperatures for projections out to approximately 60 h and hourly temperatures at 3 h intervals out to 51 h for approximately 230 stations in the conterminous United States. A combination of LFM model output, surface weather observations and climatic factors were used in this development. We derived three sets of temperature prediction equations for both the 0000 and 1200 GMT forecast cycle as follows: 1) max/min equations for the first (24 h) period and 3 h equations for projections of 6 to 27 h;2) max/min equations for the second (36 h) period and 3 h equations for projections to 27 to 39 h; and 3) max/min equations for the third (48 h) period and 3 h equations for projections of 39 to 51 h. To enhance consistency among the various max (or min) and 3 h forecasts, all the equations within each set are comprised of the same 10 predictors. We also derived a separate set of 60 h max/min equations. Comparative verification indicates that, in sharp contrast to past results, max/min forecasts from the new early guidance system are now better than those from the final guidance system. In addition, the automated 3 h temperature predictions are superior to persistence forecasts based on 3 to 6 h old temperature observations, as well as to persistence forecasts based on reports taken 24 h earlier.