Arbab, M. (2016) Accuracy Comparison of Geostatistical and Artificial Neural Network Methods to Estimate the Threshold Velocity of Wind Erosion (Case Study: Jazynak Region, Sistan plain). Masters thesis, University of Zabol.
|
Text
Accuracy Comparison of Geostatistical and Artificial.pdf Download (13kB) | Preview |
Abstract
Awareness of spatial variability of wind erosion threshold velocity is very important in focusing on desertification activities and projects. There are several ways to obtain this parameter. Using of tolerated wind tunnel, is more accurate but it requires a lot of time and cost. Therefore, modeling this important parameter, by using the readily available soil properties, seems necessary. The aim of this study is compare the accuracy of geostatistical techniques (kriging and Cokriging) and artificial neural network model for estimation of wind erosion threshold velocity. In this study, wind erosion threshold velocity was measured systematically at a height of 30 cm from the soil surface and using by a device to measuring wind erosion (wind tunnel), on 60 points from part of Sistan Plain (jazinak area). Soil samples were taken from soil surface (0 to 10 cm) at the same locations and soil texture, acidity, salinity, organic matter, mean weight of soilseeds diameter, gravel percentage, percentage of particles larger than 0.84 mm, soil moisture , soil bulk density and soil structure, were measured and in order to modeling, these variables were used as input variables in artificial neural network. Percentage of silt, due to high correlation with wind erosion threshold velocity, was considered as an auxiliary variable in cokriging method. In this study, among the geostatistical methods, cokriging with R2=0.60 and RMSE=0.45 was more accurate than kriging. While the artificial neural network with learning algorithm Levenberg-Marquart with three hidden layer, and three neurons in each layer, with R2=0.98 and RMSE=0.07 has the highest accuracy in estimating wind speed threshold.
| Item Type: | Thesis (Masters) |
|---|---|
| Uncontrolled Keywords: | Sistan plain, Geostatistics, Artificial Neural Network, Wind Velocity Threshold |
| Subjects: | S Agriculture > S Agriculture (General) |
| Depositing User: | admin admin1 admin2 |
| Date Deposited: | 16 Apr 2017 07:03 |
| Last Modified: | 16 Apr 2017 07:03 |
| URI: | http://eprints.uoz.ac.ir/id/eprint/1192 |
Actions (login required)
 |
View Item |
