Document Type : Research Paper

Authors

• Rohallah Taghizadeh Mehrjardi ¹
• Fereydoon Sarmadian ²
• Mehdi Tazeh ³
• Mahmood Omid ⁴
• Norair Toomanian ⁵
• Mohammad Javad Rousta ⁶
• Mohammad Hassan Rahimian ⁷

¹ Assistant Professor, Faculty of Agriculture and Natural Resources, Ardakan University, Ardakan, Iran

² Professor, Faculty of Agricultural Engineering and Technology, Campus of Agriculture and Resources Tehran University of Natural Sciences

³ Assistant Professor, Faculty of Agriculture and Natural Resources, University of Ardakan, Iran

⁴ Professor, University College of Agriculture and Natural Resources, University of Tehran, Karaj, Iran

⁵ Assistant Professor, Agricultural and Natural Resources Research Center, Isfahan, Iran

⁶ Assistant Professor, National Salinity Center, Iran

⁷ Scientific Board, National Salinity Center, Iran

Abstract

Recently, researchers are increasingly employed Digital Soil Mapping (DSM) techniques to overcome traditional soil mapping difficulties. Apparently, due to the large heterogeneity of soil environments, sampling may be the most important step in digital soil mapping studies. Therefore, in this research, we employed three different sampling strategies including Latin hypercube, Fuzzy-K-Means and random sampling to achieve the best spatial distribution of soil samples in an area around 720 km² located in Ardakan region, Yazd province, Iran. Auxiliary data that used in this study, were including terrain attributes, Landsat 7 ETM⁺ images and a geomorphologic surfaces map. Based on statistical criteria (i.e. mean and standard deviation), results showed that Latin hypercube is the best sampling method. For instance, in the selected points, the mean of wetness index is 18.19 which is the same as the mean of all area. Similarly, the mean of Multi-resolution Valley Bottom Flatness (MrRVF) in the points selected by Latin hypercube strategy is very similar to all area. Moreover, histogram of auxiliary data in selected points (samples) was more similar to histogram of auxiliary data in all area. Also, the results indicated that a good geographical coverage (Fuzzy-K-Means) does not adequately represent the distribution of the variables. Therefore, Latin hypercube is the best strategy to determine sample locations in our study area and hence, it is recommended that researchers apply Latin hypercube method in future digital soil mapping studies.