Improve the prediction of soil bulk density by cokriging with predicted soil water content as auxiliary variable

Abstract

Purpose Soil bulk density (SBD) is a key soil physical property affecting the transport of water and solutes, which is essential to estimating soil carbon and nutrient reserves. However, it is considered to be time consuming, labor intensive, and expensive to obtain in situ SBD data. Therefore, it is important to estimate SBD at an acceptable level of accuracy using auxiliary variables with limited fieldmeasured data. Materials and methods The Heqing Dam agriculture area in Yunnan Province, Southwest China, was selected as the study area. Sampling points of 114 were obtained based on a uniform 1 km×1 km grid methodology, which were broken into training data and test data using ArcGIS software. The training data (91 soil samples) were used for prediction and the test data (23 soil samples) were used for validation. Predicted soil water content (SWC) was estimated using kriging with high accuracy. The predicted SWC was used as auxiliary data to improve the prediction of SBD by cokriging. The correlation coefficient (r) between field-measured and predicted values, mean error (MAE), and root mean square error (RMSE) were used to validate the performance of the geostatistics. Results and discussion The SBD was highly correlated with measured (r=−0.73, p<0.01) and predicted SWC (r=−0.73, p<0.01). The two parameters MAE and RMSE showed an improvement after introducing the predicted SWC. The MAE and RMSE decreased from 0.111 and 0.032 g cm −3 by kriging to 0.070 and 0.018 g cm −3 by cokriging with predicted SWC, respectively. The r value increased from 0.449 by kriging to 0.852 by cokriging. Compared with kriging, the application of cokriging with predicted SWC resulted in a relative improvement of 45.5 %. Conclusions This study demonstrates that predicted SWC used as auxiliary data can improve the prediction of SBD. Therefore, when the predicted data have been demonstrated to be of high accuracy and are highly correlated with the dependent variable, they have the potential to be a good source of auxiliary data for cokriging.

Spatial variability in the amount of forest litter at the local scale in northeastern China: Kriging and cokriging approaches to interpolation

“…Our results agree with Chang et al (1998) and Yang et al (2016), who also found that cokriging was superior to kriging. Both COK P and COK PCA showed an improvement in the prediction of litter amount in comparison to OK.…”

“…Basaran et al (2011) concluded that cokriging was superior to kriging in predicting soil hydraulic conductivity with limited available data on a 12.8 km 2 area. Yang et al (2016) found that cokriging improved the result by 45.5% compared with kriging, with soil bulk density as a primary variable and soil water content as an auxiliary variable. Yang et al (2016) found that cokriging improved the result by 45.5% compared with kriging, with soil bulk density as a primary variable and soil water content as an auxiliary variable.…”

“…The initial comparison between kriging and cokriging was based on theoretical analysis that indicated the superiority of cokriging and that it ensured consistency between an estimation of a sum and the separate estimation of each of its terms (Wackernagel, 1994). Recently, the enhancement of cokriging for spatial interpolation has been investigated in practical application (Adhikary, Muttil, & Yilmaz, 2017;Basaran et al, 2011;Kuntz & Helbich, 2014;Yang, Luo, Jiang, Li, & Yuan, 2016). Basaran et al (2011) concluded that cokriging was superior to kriging in predicting soil hydraulic conductivity with limited available data on a 12.8 km 2 area.…”

Spatial distribution of soil chemical properties in an organic farm in Croatia

“…These analyses are carried out using hybrid methods, such as co-kriging (Wen et al, 2015). The use of auxiliary variables normally increases the accuracy of the spatial predictions as observed in previous works (Stein and Corsten, 1991;Zhang et al, 1992;Yang et al, 2016b;Ceddia et al, 2015;Chen et al, 2016). Nevertheless, several works observed that the use of co-variates did not improve the accuracy of the interpolation (Martínez-Cob, 1996;Castrignano et al, 2011;Ceddia et al, 2015).…”

Spatial variability of soil carbon to nitrogen ratio and its driving factors in Ili River valley, Xinjiang, Northwest China

“…Typical methods of measuring soil C/N ratio require data collection of SOC and TN, which can be considered a time-consuming and labor-intensive work that is not cost effective (Yang et al, 2016). Cokriging interpolation has been gradually applied to aspects of soil property prediction because it provides a higher accuracy of estimation, especially in the case of low sampling density, than the other method (Yao et al, 2014).…”