ANISOTROPIKC SEMIVARIOGRAM MODELING WITH OUTLIER MODIFICATION USING THE YEO-JOHNSON TRANSFORMATION METHOD (CASE STUDY: GROUNDWATER LEVEL DATA IN KALIMANTAN)
Indonesia possesses the largest tropical peatland area in the world, approximately 13.43 million hectares, with 33.5% of it located in Kalimantan. Peatlands, rich in organic matter, play a significant role in the global carbon cycle but are vulnerable to forest and land fires due to the lowering of...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Project |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/81546 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Indonesia possesses the largest tropical peatland area in the world, approximately 13.43 million hectares, with 33.5% of it located in Kalimantan. Peatlands, rich in organic matter, play a significant role in the global carbon cycle but are vulnerable to forest and land fires due to the lowering of the groundwater table (GWL), which increases the risk of drought. This research aims to understand the impact of outlier modification on anisotropic semivariogram modeling in mapping the distribution of groundwater table height in Kalimantan's peatlands. Semivariogram modeling is a geostatistical method that can be used to analyze the spatial pattern of GWL variable. Anisotropic semivariograms account for distance and direction between locations. The Ordinary Least Square (OLS) method, supplemented with the Trust-Region (TR) numerical method for estimating theoretical semivariogram parameters, resulted in the Gauss model being the most suitable semivariogram model. Outlier modification was performed by detecting outliers using Moran’s I statistical test and transforming the data with the Yeo-Johnson transformation. TMAT values at unobserved locations were predicted using Ordinary Kriging (OK) interpolation, leveraging the obtained theoretical semivariogram model.
The research findings indicate that the Yeo-Johnson transformation effectively reduces the variability of the experimental semivariogram and produces a more stable and accurate model. The best semivariogram model was obtained through the Matheron and Cressie-Hawkins approaches after the Yeo-Johnson transformation, with the lowest Root Mean Square Error (RMSE) and GWL predictions closest to the observed data. The predicted GWL distribution map at unobserved locations indicates an anisotropic spatial influence in the 45° direction or the Northeast-Southwest sector.spatial anisotropikc influence in the direction of 45° or the Northeast-Southwest sector. |
|---|