Fitting of Probability Distributions for Bias Minimized Modeled Temperature Data

摘要

The study aims at reducing the bias in the temperature data obtained from the satellite models by comparing with that obtained from the observatories. Data obtained from the observatories are always more accurate than those from the satellite model. Bias correction methods such as difference method (DM) and modified difference method (MDM) were attempted to minimize bias of the satellite modeled temperature data compared to the data recorded at observatories. Best bias correction method was identified based on the coefficient of variation. MDM recorded lower CV in the corrected modeled data based on daily and weekly maximum and minimum temperatures. This indicated its use in smoothening of the data. Difference Method (DM) recorded lower CV in the corrected modeled data of weekly minimum temperature and monthly data of maximum and minimum temperatures. Generally, studies of temperatures are more related to months rather than week/years. For the bias corrected modeled (satellite data) monthly data, distributions functions were fitted and their goodness of fit was verified using Chi-square test. No generalized single model was found to be the best fit for both the climatic parameters. Cauchy, Gamma and chi-squared distributions were resulted as the best fit more than once in the analysis for Maximum temperature. For the Minimum temperature, Log-logistic distribution was more in number compared to other distributions.