PRECISION OF PREDICTING SOIL-SALINITY BASED ON VEGETATION CATEGORIES OF ABANDONED LANDS

Citation
T. Toth et al., PRECISION OF PREDICTING SOIL-SALINITY BASED ON VEGETATION CATEGORIES OF ABANDONED LANDS, Soil science, 160(3), 1995, pp. 218-231
Citations number
19
Categorie Soggetti
Agriculture Soil Science
Journal title
ISSN journal
0038075X
Volume
160
Issue
3
Year of publication
1995
Pages
218 - 231
Database
ISI
SICI code
0038-075X(1995)160:3<218:POPSBO>2.0.ZU;2-1
Abstract
Abandoned saline lands on the Huang-Huai-Hai Plain of China are occasi onally used for crops. Our objective was to determine how precisely th e values of soil variables, such as ion concentrations, pH, and penetr ation resistance, could be predicted to fall into a particular range o f values based on the presence of specific categories within the semi- natural vegetation. The correlation between sets of soil (mostly chemi cal) and plant (cover percentage) variables in a 100 x 220-m plot was examined by calculating canonical variates. The correlation between th e first canonical variates of the two sets was 0.88, and indicated the presence of a strong relationship between soil and vegetation. The se mi-natural vegetation of the Bite was classified either as dominated b y Phragmites communis or by Imperata cylindrica plant species, In this plot, these two categories represent overlapping ranges of several so il variables. On average, the PHRAG-MITES category occurs at a higher elevation and has a greater salt concentration and smaller penetration resistance and pH than does the IMPERATA category. Based on our obser vations, leaching does not affect the concentration of bicarbonate ion s in the surface soil layers. However, as the concentration of other i ons decreases and the ratio of bicarbonate to other ions increases, pH and penetration resistance increase in some patches. The degree of se parability of different soil variables by the vegetation categories wa s inferred from (i) the significance of the difference of their means using the Mann-Whitney test and (ii) from the correlation between the original variables and the linear discriminant function of the two veg etation categories based on the soil variables. The precision of predi cting ranges of individual soil variables from vegetation categories w as determined by one-dimensional thresholding and ranged from 67 to 84 %. Misclassification by vegetation categories into the range assigned to the other vegetation category occurred at least twice as often at t he borderlines between the vegetation categories than inside each cate gory. Vegetation categories are considered an easy and economic way of determining the likely success of cropping in saline areas.