On forecasting night minimum temperatures over New Delhi

Abstract

Regression equations have been developed to forecast day to day minimum temperatures of New Delhi for the month of January. The minimum temperature has been taken as dependent variable whereas the different weather elements, chosen on physical reasoning, are the independent variables in the equations. Techiques have been evolved to parameterise advection and clouding. The first equation uses only morning data so that forecast can be issued in the forenoon itself. The second and third equations use the evening data as well and arecomparatively better. Both the equations have been tested for independent data of one year and the results have been found very encouraging. Above 80 per cent of the forecast. temperatures for .both the equations lie with in  ±2° C of the actual temperatures. The Root Mean Square Error of the forecasts (independent data) based on the equations is less than 1.5 as compared to 2.1 of persistence alone. The main improvement is seen to be in the cases when temperature changes exceed 3 oC from that of previous day.