New procedure in developing adjustment algorithm for harmonizing historical climate data sets

Summary

Description

Existing historical weather data in most developing countries have gaps as a result of stolen or old equipment and shortage of trained observers. This confounds analysis of climate change trends, extreme events and climate risks. In order to grapple with the problem, automatic weather stations and weather generating software have been routinely used as alternatives to fill data gaps. This study therefore seeks to analyze the statistical association between the various datasets as a way of developing adjustment algorithms in order to generate a single, fit-for-purpose climate data set. Four weather stations from Uganda’s four major agro-ecological zones were purposively selected for the study; Characteristic daily weather data (1991–2013) were then obtained from the UNMA archives. Adcon telemetry automatic weather data (2010-2013) were acquired from the NARO database. Software generated datasets were attained from Weatherman and MarkSim programs. These data sets were re-arranged into suitable formats using RClimDex. Pearson's product moment correlation (r) and Simple linear regression (R-squared) were used to measure strength of linear relationship for rainfall series; Paired Samples T-test was used to make pair-wise comparisons for temperature data (at 5% significance level). There was a strong, positive, statistically significant relationship between the observed and simulated/automatic rainfall data (r>0.7, p<0.05) with about 60% of the variation explained by the fitted model. There was no significant difference in mean temperature records between generated/automated weather stations and manually observed ones (p>0.05). It is therefore recommended to use weather generators and automatic stations in filling out weather data gaps and harmonizing climate data.

Related Content