Dealing with spatial outliers and mapping uncertainty for evaluating the effects of urbanization on soil: A case study of soil pH and particle fractions in Hong Kong

Abstract

The effects of urbanization on soil have been accelerating around the globe and there is a need for these effects to be evaluated precisely. In order to do this, spatial outliers and mapping uncertainty should be handled properly. The present study aimed to deal with these two problems using a case study of soil pH and particle fractions (i.e., sand, silt and clay) in Hong Kong. Based on 133 topsoil samples collected in an urban and peri-urban mixed area of Hong Kong, robust estimators for spatial variogram were first adopted with a procedure for identifying spatial outliers to remove spatial outliers. Then, 1000 models were simulated for each soil property using the methods of maximum likelihood and Monte Carlo Markov Chain, in order to characterize parameter uncertainty. Finally, 100 of the 1000 simulated models were randomly selected to construct soil maps using kriging, and the interpolation uncertainty was characterized using Sequential Gaussian Simulation, generating 1000 simulations for each model. Based on the total 100,000 simulated values at each location for each soil property, temporal variation of the soil property at the location across the past half century was derived against the soil series map established in 1960s for this area. Probability that soil property value changed after urbanization was also computed, in order to show if soil property has changed with a significant confidence. Results showed that the situation of spatial outliers in the soil data of the present study was not serious and mapping uncertainties were large. Temporal variations of soil pH and particle fractions obtained through comparison between soil maps of pre-urbanization and post-urbanization were not completely attributed to the effects of urbanization, but partially to mapping uncertainties. Urbanization in Hong Kong during the past half century influenced soil pH slightly and influenced soil particle fractions relatively strongly.