Practical kriging models with divide and conquer algorithms for spatial heights forecast

Abstract

In environmental management, fast decisions and judgments can minimize the severity of the damage. Computer programming techniques and interpolation methods can be utilized to save time and improve data accuracy. This paper presents two algorithms (DVC and SDVC). The first algorithm employs the divide and conquer strategy, whereas the second approach enhances the first by introducing a slope threshold parameter. The kriging models are used to predict spatial heights at unknown locations. The DVC algorithm offers six different kriging models to choose from. The experimental findings demonstrated that our four exponential-based models outperform the Gaussian and Spherical models in terms of accuracy. The two best performances of exponential-based models are utilized to improve the SDVC algorithm. Regarding processing time, the DVC algorithm significantly reduces the waiting time from 7705 to 33 min. The SDVC algorithm decreases the waiting time further down to 4 min but with a trade-off in model accuracy. Finally, we recommend using the SDVC algorithm with a slope threshold of 8%. This research leads to a significant reduction in decision-making time and a practical increase in computation speed when generating 3D surface plots for environmental management applications such as flood protection, transport planning, and farming management.