Tropical Geography ›› 2019, Vol. 39 ›› Issue (5): 770-779.doi: 10.13284/j.cnki.rddl.003169

Previous Articles     Next Articles

Comparing and Analyzing Different Spatial Interpolation Methods for Soil-Moisture Estimation in Karst Areas

Sun Zhiyan, Zhou Qiuwen(), Zhang Siqi, Wei Xiaocha, Ma Longsheng   

  1. School of Geography and Environmental Science, Guizhou Normal University, Guiyang 550025, China
  • Received:2019-01-05 Revised:2019-07-12 Online:2019-09-10 Published:2019-11-08
  • Contact: Zhou Qiuwen E-mail:zouqiuwen@163.com

Abstract:

Most precipitation seeps underground and is difficult to capture because of the poor water-holding capacity of soils in karst areas and dual aspects of the hydrogeological structure formed by karstification. Therefore, soil-moisture estimation is critical for karst ecological restoration in karst area and for agricultural development. Using the Yangmei River basin as the research area, environmental factors, such as Normalized Differential Vegetation Index, Wetness Index (WI), altitude, slope, and relative height, were extracted from remote-sensing imagery and Digital-Elevation-Model data. On the basis of correlation analysis, WI was chosen as an auxiliary variable. Spatial soil-moisture interpolation by four methods, namely ordinary kriging, cokriging with WI vector point as auxiliary, cokriging with WI raster, and kriging combined with regression, were applied to soil-moisture sampling and WI raster data. Root Mean Square Error of predicted and measured values and unary linear regression between predicted and measured values were used to estimate the accuracy of different interpolation methods. Results showed the following: 1) From correlation analysis between soil moisture and other environmental factors, we found a significant negative correlation between WI, which is associated with near surface humidity condition, and soil moisture (P=0.01) but no significant correlation between other environmental factors and soil moisture was found. According to the outcome of stepwise regression, WI was chosen and other environmental factors were removed. 2) Kriging combined with regression had the smallest error of estimation in soil moisture in the study area; cokriging had a larger error, and common kriging had the largest error. When interpolating soil-moisture samples by cokriging, using WI vector-point data was more efficient than WI raster data as the auxiliary variable. 3) The soil-moisture surface generated by ordinary kriging was the smoothest, but kriging combined with regression reflected the soil-moisture spatial pattern for the study area to the greatest extent. However, in general, the spatial variation of soil-moisture interpolation results generated by the four methods was all lower than the actual condition as observed. 4) When fewer soil-moisture sample points were used for interpolation, errors of average value and spatial variation results of the four interpolation methods were larger. However, no matter what number of soil-moisture sample points were used, average values of soil-moisture interpolation results from kriging combined with regression were lower than those of ordinary kriging and cokriging, and spatial variations of soil-moisture interpolation results were higher than those of ordinary kriging and cokriging. In conclusion, the accuracy of kriging combined with regression is better than ordinary kriging and cokriging applied in soil-moisture interpolation in karst areas, which were consistent with the result of other research comparing the accuracy of different interpolation methods. To improve the accuracy of the soil-moisture interpolation results, we would need more soil-moisture sample points and higher resolution remote-sensing imagery to extract more-detailed information on environmental factors, particularly those more significantly correlated with soil moisture.

Key words: Kriging interpolation, Kriging combined with Regression, soil moisture, karst

CLC Number: 

  • S152.7+5