1. An autocorrelated conditioned Latin hypercube method for temporal or spatial sampling and predictions.
- Author
-
Le, Van Huong and Vargas, Rodrigo
- Subjects
- *
DISTRIBUTION (Probability theory) , *LATIN hypercube sampling , *SOIL respiration , *TIME series analysis , *CARBON cycle , *SOIL sampling - Abstract
A data-driven method is presented for improving sampling designs from times series (1D approach) or spatial arrays (2D approach) of digital information. We present the autocorrelated conditioned Latin Hypercube Sampling (acLHS). This method combines a conditioned Latin Hypercube (cLHS) to obtain a representative sample of the joint probability distribution function and an autocorrelation model to reproduce the spatial or temporal dependency function (i.e., temporal or spatial variability). The acLHS method was tested with two case studies using data of soil CO 2 efflux (i.e., the CO 2 flux from soils to the atmosphere known as soil respiration) that are useful for carbon cycle science. First, we used data representing a time series (1D approach), and then spatial data (2D approach) across the conterminous United States (CONUS). Results show that acLHS was more efficient than other sampling methods (i.e., fixed sampling, cLHS) as it better reproduced the joint probability distribution and the temporal or spatial variability of the variable of interest. Finally, we use a Bernstein copula-based stochastic co-simulation method (BCSCS) and demonstrated that the acLHS reduces modeling prediction errors compared with other methods. The acLHS is a flexible method that can be applied to any variable of interest as a time series (1D approach) or as a spatial format (2D approach). • A geostatistical sampling method to improve representativeness of samples. • Implementation works for time series (1D) and spatial information (2D). • The method improves accuracy and precision for prediction purposes. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF