A Comparison of Bangladesh Climate Surfaces from the Geostatistical Point of View

Table 3

Different spatial interpolation methods.

		Thin plate spline	Inverse distance weighting	Ordinary kriging	Universal kriging
	Description of the climate variability	Smoothed spline exactly through the measured spatial points	Straight line exactly through the measured spatial points	Smoothed line that best fits through the measured semivariance considering anisotropy	Smoothed line that best fits through the measured semivariance considering spatial trend
Performance parameters	Performance in spite of the changing spatial point configuration	Performs comparatively worse than the stochastic methods and shows improved performance in cross-validation with the increased number of spatial point observations	Performs comparatively better than other methods when the number of spatial point observations is too little but in general shows improved performance in cross-validation with the increased number of spatial point observations	Performs comparatively better than the deterministic methods and shows improved performance in cross-validation with the increased number of spatial point observations	Performs best of all the methods applied and shows improved performance in cross-validation with increased number of spatial point observations
	Measurement errors’ inclusion	Includes measurement errors partially in results because of the minimal smoothing	Includes measurement errors in results totally	Excludes measurement errors in result by smoothing	Excludes measurement errors in result by smoothing
	Short versus long range variability	Describes mostly long range variability by removing the noise of short range variability	Describes only short range variability	Adjusts between the short and long range variability by incorporating anisotropy and semivariance	Describes mostly long range variability in light of the spatial trend