Estimating regional heavy metal concentrations in rice by scaling up a field-scale heavy metal assessment model

The objective of this study was to determine the levels of heavy metals cadmium (Cd) and copper (Cu) in rice by upscaling a field-scale heavy metal assessment (FHMA) model from field to regional scale. The FHMA model was established on the basis of spectral parameters in combination with soil parameters by employing a generalized dynamic fuzzy neural network. The piecewise function and ordinary kriging were developed to suit the upscaled spectral parameters and soil parameters, respectively. In addition, the network structure and fuzzy rules, which had already been developed in the FHMA model, would be subsequently extracted as those of the regional-scale heavy metal assessment (RHMA) model. The results showed that the latter performed well at prediction with a correlation coefficient (R2) and model efficiency (ME) greater than 0.70, and can be applied to other areas, perhaps universally. This study suggests that it is feasible to accurately estimate regional heavy-metal concentrations in rice by scaling up the FHMA if such a strategy is appropriately selected and finds that the piecewise function is well suited to transferring spectral data from a field to a regional scale.