Journal of Rangeland

Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:

---

Insight into the rangeland ecosystems, its components and their relationships, soil and vegetation for instance, is among the keys of proper range management. On the other hand, the use of machine learning methods to discover the relationship between these factors can help in Cost reduction of Soil sampling and testing. In this study therefore, the effect of some soil characteristics on distribution of Artemisia sieberi is tested in Nodoushan rangelands, Yazd province. Coverage information was measured in 320 plots put alongside of 40 transects. The soil profiles were dug at the beginning and end of each transect to take soil samples from two depths of 0 to 10 and 10 to 30 cm. samples were taken to the lab to be analyzed. Coverage percent of Artemisia sieberi was determined by using a machine learning method. Soil parameters and six algorithms were used for the task. Then, with weighting the factors, their effect on the prediction of the coverage was investigated. The results of the model showed that “Gaussian process model” with (RMSE= 1.385) and (R= 0.998) in the training dataset and (RMSE = 0.960) and (R = 0.9999) in the test dataset has a higher accuracy on the prediction of  the coverage percentage than other models. Weighting results also showed, that among the soil parameters, sodium at depths of 0 to 10 and 10 to 30 cm, has the most effect on vegetation estimation. The results generally showed that high soil parameters and machine learning methods are good way of  coverage prediction in the area.

Keywords: Soil parameters, Coverage percentage, Data Mining, Artemisia sieberi

Full-Text [PDF 711 kb]   (1462 Downloads)    

Add your comments about this article : Your username or Email:

---

---