Modeling and optimization of experimental designs for soil loss assessment at plot scale

Abstract

Abstract Appropriate investigation of soil erosion triggering/controlling factors is of great importance for natural resources management. Usually, soil erosion assessment is time consuming and impose high costs to researchers and managers. Therefore, design of experiments methods can be good approaches to achieve the purpose to which adequate attention has not been paid yet. This study was aimed to assess two response surface methods of Box-Behnken (BBM) and Face-centered central composite (FCCM) as well as a fractional factorial design of experiment Taguchi method (TM) for modeling and optimization of soil loss process at plot scale. So far, no study\xa0has been conducted to evaluate and compare the ability of BBM, FCCM, and TM in soil erosion modeling, optimization and assessment. Rainfall simulations were performed in the conditions determined by BBM, FCCM experimental designs and TM orthogonal arrays, and in-situ soil loss measurements were performed at small plot scale. Although, results showed that BBM outperformed FCCM and TM, the ANOVA test indicated verified non-significant difference (p=0.921 at 0.01 significance level) in performances of the modeling procedures. It was further shown that TM had an acceptable performance with regard to the very limited number of required measurement tests. Results further showed that there was a good agreement (r=0.97) between the results of TM with just 16 runs and BBM with 56 runs with respective RMSEs of 8.16 and 5.08. Therefore, TM can be suggested for soil erosion studies when there are limitations in time and expenses for simulations. However, if the limitations can be overcome, BBM can be suggested as a better design of experiment in soil erosion studies.