Aiming at the pricing of primary agricultural products for the large-scale suppliers and the wholesalers in agri-supply chain management, an approach for the large-scale supplier pricing is presented based on the threshold cointegration method of wholesale prices online including the GBand-TAR modified Band-TAR model. Our empirical work shows that it is more appropriate for a large-scale supplier pricing with his wholesalers based on the threshold cointegration method than the conventional linear cointegratiun method in spatially separate markets in an agri-supply chain of soybean in China in short time. Firstly, the three pairs of prices in spatially separate markets are of long-run equilibrium and threshold cointcgration. The forecast lest shows that the thresh-old cointegration approach is superior to the conventional linear cointegration approach in short time. Secondly, there are two thresholds of GBand-TAR in which the threshuht parameters represent relative transaction costs. Larger thresholds or wider neutral band corresponds to the greater distance between markets. Thirdly, the esti-mation of M-TAR shows that the large-scale supplier is more sensitive to incrcase of wholesaler prices than de-crease of wholesaler prices. The supplier can price on the forecast of market price by the threshold ECM inclu-ding the GBand-TAR if the equilibrium error of threshold lag is not in the interval of thresholds in which there is not profitable trading opportunities for the supplier.