One of the applications of remote sensing in agriculture is to obtain crop status for estimation and management of variable rate of inputs in the crop production. In order to select the appropriate wavelengths related for crop characteristics, the field experiments were conducted in June-July in the farming area of Hokkaido University (43° 4' N 141° 20' E), Japan in two years. To make difference in growth conditions, the field was divided into 8 areas and four level of fertilizer (Ammonium Nitrate) 0, 30, 60, and 90 kg haι with two repetitions were applied at the reviving stage (GS 26). Hyper spectral reflectance data using a portable field spectroradiometer (351 to 2500 nm) were recorded from 10 am to 2 pm, under cloudless conditions at four different growth stages of winter wheat in first year and at eight different growth stages in second year. The partial least squares (PLS) regression was used to determine important wavelengths. Some wavelengths in green, amber, red, red edge, near infrared and short wave bands were identified by PLS as significant wavelengths for measured crop reference data. All two-band combinations of the several vegetation indices were subsequently calculated in an algorithm for determining linear regression analysis against SPAD value, protein content, and grain yield. R square matrix used in ArcMap to make contour plot. Using overlaying in analysis tools between first and second year results, a number of grouped wavebands with high correlation were revealed. The results from the calibration models built by PLS showed strong relationships between predicted and measured SPAD value, protein concentration and Yield.