The performances of these methods for estimating the growing period are compared in Fig. 5. The maximum estimation error with the Fourier transform (51 days) is much more than that with the wavelet transform (− 25 days), owing to the localized characteristics of the wavelet transform. The wavelet transform can detect localized changes separately in the time domain. In contrast, the trigonometric function of the Fourier transform is not localized in the time domain. Hence, the Fourier transform causes errors that occur in separate positions to interfere. Since the planting and harvesting dates estimated by using the wavelet transform did not include these errors, the growing period estimated by using the wavelet transform was also superior to that estimated by using the Fourier transform.