A comprehensive yield evaluation indicator based on an improved fuzzy comprehensive evaluation method and hyperspectral data

Abstract The accurate and timely estimation of winter-wheat yield at the field and regional scales is critical to developing agricultural management strategies and reducing the effect of changes in environmental conditions on crop yield. Growth status and trend (GST) monitoring has been widely applied to estimate agronomic parameters using remote sensing methods. Many studies have employed GST monitoring, however, most of them were based on a single agronomic parameter and can therefore only represent one-sided or local GST information. Additionally, each agronomic parameter is interactive. Meanwhile, little studies have systemically combined multiple agronomic parameters into one comprehensive indicator to estimate crop yield using remote sensing data. Thus, the objectives of the current research were to build a comprehensive yield evaluation indicator (CYEI) using the improved fuzzy comprehensive evaluation (FCE) method and evaluate the performance of CYEI to monitor GST and estimate yield. The results showed that the CYEI can fully reflect the information of the leaf area index, leaf biomass, leaf water content, and leaf nitrogen content. Compared with various agronomic parameters, the CYEI based on the improved FCE method was more closely correlated with the yield (the R2 values of the validation set were 0.63, 0.69, and 0.63 at the booting stage, anthesis stage, and milk development stage.). The CYEI was estimated using a linear model constructed using the optimal VIs, and the results for the three growth stages achieved a higher precision (R2 = 0.74, 0.74, and 0.68 for the booting, anthesis, and milk development stages, respectively) than the traditional single agronomic parameter. The CYEI and Bayesian information criterion were then used to select VIs and then build a partial least squares regression model to estimate the yield. The estimation accuracy was found to be satisfactory, with R2 values of 0.55, 0.64, and 0.66 at the booting, anthesis, and milk development stages, respectively. Finally, a more intuitive image-scale yield monitoring method was obtained based on unmanned aerial vehicle remote sensing hyperspectral imagery. In the future, the proposed method can be used to obtain wheat growth information and provide a new prediction indicator to better estimate yield in precision agriculture.