Using synthetic aperture radar (SAR) to measure the area of rice cultivation.

摘要

This study developed a highly accurate method of measuring the area of rice cultivation. The method uses synthetic aperture radar (SAR), which is not influenced by cloud cover. Natural and economic forces greatly affect agriculture (e.g., in terms of global warming and tariff abolition, respectively). It is important that Asian countries secure supplies of rice, a staple food. Therefore, it is essential to accurately determine the area planted in rice. Typically, the planted-rice area is calculated statistically from a sampling survey. However, this method is expensive and time consuming, even in a relatively small country such as Japan. Remote sensing technology has the potential to decrease time and cost constraints. While clouds block optical sensor observations, clouds do not affect SAR. Therefore, a planted-rice measurement method using SAR should be developed. This study first combined RADARSAT data observed by a SAR sensor at two different times with a non-rice field mask. Previous studies have combined SAR data with optical data to calculate the planted area or have treated SAR data from multiple observations as multi-band data. However, clouds interfere with optical observations, and time constraints limit the usefulness of multiple SAR observations. Therefore, we developed a method that does not use optical sensor data and uses SAR data from only two periods: after rice transplantation and at rice maturity. The P-tile method was used to determine a threshold for distinguishing water (a difficulty in past studies). A method for determining an objective threshold considering strong light was also developed. In addition, estimation accuracy was increased by using a non-rice field mask created from a digital 1:25 000-scale map. Finally, a method that combines RADARSAT data from two observations with a non-rice field mask was designed, and the planted-rice area was estimated. The estimated value was +1.5% higher than the statistical estimate and equaled the accuracy of the optical technique. Examining results using the same method in the following year verified the method's accuracy and confirmed its applicability. Next, we developed and verified a planted-rice measurement method that uses a single observation of multi-wavelength, multi-polarimetric SAR data. The launch of ALOS/PALSAR, an L-band SAR, is scheduled for fiscal year 2004. Bragg scattering, which increases measurement error, has been reported when observing paddy in the L band. We solved the problem of Bragg scattering by using multi-polarimetric SAR data. If multi-wavelength observations are available, it should be possible to measure the planted-rice area from a single observation. We thus developed an algorithm that measures the planted-rice area using multi-wavelength, multi-polarimetric SAR data. This algorithm was confirmed in a verification experiment using aircraft multi-wavelength multi-polarimetric SAR data. Verification showed that this method produces a highly accurate estimate of the planted-rice area. This method of using multi-wavelength, multi-polarimetric SAR data will greatly improve the efficiency of rice-area measurements, since the method is not influenced by weather, overcomes the problem of Bragg scattering, and only requires data from only one observation period..