Your Cart

Your cart is empty

Home  >  Volume 35

Dimension Reduction and Clustering of Micrometeorological Variables Using P mode Principal Component and Hierarchical Cluster Analyses by Mojisola Oluwayemisi Adeniyi (pages 175-184)
Sale price: $5.00


Variations in meteorological variables greatly influence socioeconomic activities, especially in developing nations such as Nigeria. However, routine meteorological data are limited to a few variables. Though, unavailablevariables can be estimated from the available ones, the transfer functions connecting the available routine parameters and the unavailable ones do not exist. This paper provides the transfer functions by grouping 36 observed micrometeorological parameters from Nigerian Micrometeorological Experiment 1 (NIMEX_1) for subsequent estimation of parameters in each group from a member of the group. The NIMEX_1 was carried out at Ile-Ife, Nigeria in 2004 between Day of years 56 and 68. Principal component analysis is used to reduce the dimensionality in the observed data while cluster analysis is used to group them into variables with similar variances. Thereafter, polynomial regression is used to estimate some variables in a cluster from a member of the cluster.The cross-validated coefficient of determination between observed and estimated parameters ranges between 0.64 and 0.99.Moreover, the cross-validated root mean square errors are mostly less than the standard deviations in the observed data. Therefore, the obtained transfer functionsare applicable in estimatingmeteorologicalvariables in areas with similar weather conditionsto the prevailing weather during NIMEX_1 experiment.

Keywords:P mode PCA, cluster analysis, micrometeorological variables, polynomial regression