Uncertainty analysis for the evaluation of agricultural soil quality based on digital soil maps

Spatial soil quality information is needed for the agricultural development in Hong Kong. This study produced digital maps of more than 20 soil quality indicators of a study area in Hong Kong, using digital soil mapping techniques. These maps were then employed to evaluate the soil quality of this area via the application of scoring functions and integrated quality index (IQI) methods. The accuracy and uncertainty of the evaluated spatial soil quality information were assessed based on a probability sample and sequential Gaussian simulation (SGS), respectively. The sources of uncertainty were analyzed using stochastic sensitivity analysis. The results showed that mapping accuracy varied dramatically among soil indicators, with soil quality index (SQI) in the area ranging from 0.43 to 0.87. The obtained spatial soil quality information appeared to be moderately accurate with high uncertainty, which suggests that it cannot be fully relied on. The soil quality could have been overestimated with a probability of more than 0.95 for nearly half of the study area while being underestimated for 0.2% of the study area. Hence, this study shows that it is vitally important to derive uncertainty for soil quality information evaluated based on digital soil maps. Generally, heavily weighted soil quality indicators in the soil quality evaluation model contributed the most uncertainty, such as available phosphorus (A-P), total phosphorus (TP), and bulk density in this study.