A case study: Exploratory analysis and proposal for the problem of quality in the meteorological observations of the north of Chile

Abstract

Data quality problems in meteorological variables are a situation that the scientific community constantly faces, mainly because these quality problems materialize as missing data within the time series, which prevents compliance with the established requirements to analyze climate change in a given geographical area. Based on this problem, this article presents an exploratory analysis of the main meteorological variables (Temperature and Precipitation) observed by the meteorological stations distributed in Northern Chile to assess the data quality they present. Data imputation methods are also proposed to address this problem by completing the missing data. In particular, the experiments developed based on the phases of the CRISP-DM methodology are presented in an adapted way considering five different imputation methods of which the residual error closest to zero and the highest positive correction is sought. In the results, CLP, IDC, and RN stand out as the best techniques, which allows us to conclude that these methods can be recommended and proposed as an alternative solution according to the meteorological variable.