Katerina Bojkovska, University “St. Kliment Ohridski”, Republic of Macedonia

This study focuses on spatial interpolation, specifically discussing data preprocessing, parameter optimization for interpolation methods, and the selection of variograms. We propose a normalization process for data during preprocessing. To achieve the best results, we investigate the parameter settings for interpolation methods and the choice of variogram models, taking into account the characteristics and inherent patterns of the data. We establish appropriate parameters through a comprehensive adjustment of lag sizes and intervals, selecting the variogram model that best fits by comparing the fitting curves of different models. Using the long-term average temperature data of Henan province as an example, exploratory data analysis reveals that the data, after logarithmic transformation, better meets the requirements for interpolation. Subsequently, we simulate the spatial distribution of temperature using the Inverse Distance Squared (IDS), Ordinary Kriging (OK), and Co-Kriging (CK) methods, incorporating the identified parameters and variogram models. An evaluation based on mean error and root mean square error indicates that while the predicted maps produced by the three interpolation methods show overall similarities, the CK method provides a more accurate reflection of temperature distribution. The results suggest that the proposed methods and procedures can serve as effective tools for temperature prediction and can be readily applied to similar situations.

APA:
Bojkovska, K. (2023). The Study of Temperature Variability Prediction in Regions with Limited Samples Is Crucial for Accurate Climate Understanding. Journal of Science and Engineering Management, 4(2), 55-66. https://doi.org/10.33832/jsem.2023.4.2.05