3. Division of Information Analysis
The mission of the Division of Information Analysis is to develop methods of measuring and analyzing agro-environmental factors, which are used in the domains of remote sensing analysis, isotope technology, statistical methodology, and information processing.
Twenty papers were published in scientific journals, and about 100 short reports were presented or issued at a variety of conferences and symposia. Another 80 articles were published in various media.
Dr. G. Saito, the Associate Division Director for Research, received
the funding, "Cooperative system for supporting priority research"from
the Japan Science and Technology Corporation (JST). This project will continue
for 5 years, and will promote remote sensing research activities at NIAES. One
more project is being funded by JST, 7 projects by the Science and Technology
Agency, 2 by MAFF, and 2 by the Environmental Agency of Japan. The Division
held 11 seminars. Forty-six researchers from overseas visited our Division
this fiscal year.
Topic 1: Determination of the area of a paddy field from
RADARSAT synthetic aperture radar data.
Remote sensing is suitable for detecting land cover and is a suitable method for determining paddy field area. If optical sensor data are collected by remote sensing at a suitable time - just after transplanting - we can easily determine the area of paddy fields. Unfortunately, it is very difficult to acquire optical sensor data during the rice growth season in cloudy weather.
Our study area was the Saga Plain on the island of Kyushu in southeastern Japan. Rice is cultivated in paddy fields that are flooded at planting time.
RADARSAT has C-band (5.3 GHz, 5.6 cm wavelength) synthetic aperture radar (SAR). Most SAR microwave scatter forward on the water surface by mirror reflection. This phenomenon makes the level of backscatter low in water-covered areas. This paddy rice area determination method utilizes the low backscatter at transplanting time and volumetric scatter increases with growth. In this study, we used 2 RADARSAT images from 2 July 2000 and 27 July 2000. Two scenes were acquired in the same S2 mode. The 2 July image captured the end of the rice-transplanting period, and the 27 July image was taken during the rice-growing period, when RADARSAT backscatter had increased.
Mirror reflection makes water-filled areas such as paddy fields appear
dark in SAR images. Using the RADARSAT image of 2 July, dark areas were paddy
field, water bodies, and large flat structure. The water areas determined the
RADARSAT image of 27 July contained water bodies and large flat structures.
First, we used a digital elevation model (DEM) to perform geometrical correction of all data. We then determined the area of paddy field as the area of water in the 2 July image minus that in the 27 July image. We defined the vector boundaries of cities, towns, and villages from a 1:25, 000 map, and we then determined the total area of paddy field in every municipality and compared our estimates and the areas calculated by standard statistical methods. The aggregated correspondence of our estimates was 101.6%, and the correspondence for each municipality varied between 53.8% and 122.0% (standard deviation 13.1%) (
Topic 2: Development of a methodology for estimating paddy
field area by using remote sensing data.
We set up a collaborative project between our institute and the Institute
of Natural Resources and Regional Planning, Chinese Academy of Agricultural
Sciences. The project aimed to develop "An Advanced Method of Predicting
Food Productivity Changes According to Global Warming in Asia Using High Performance
In fact, we developed 2 methods that used remote satellite sensing data and used them to examine the Wuxi area on the mid-coast of China. First, we used a time series of LANDSAT thematic mapper (TM) data that did not include just after transplanting. Second, we developed a method of using SAR data, which could be gathered regardless of weather conditions.
We used LANDSAT TM data from 4 May 1997, 5 June 1997, 11 August 1998,
and 3 February 1999. We calculated a normalized difference vegetation index
(NDVI) for these 4 images, and we were able to distinguish agricultural fields
by detecting large differences in the NDVI values. Band 5 of LANDSAT TM can
detect water and shows water-covered areas as darker than land areas. On 11
August, band 5 detected the paddy fields as water-flooded areas within agricultural
We also used JERS-1 SAR data from 22 January 1998, 3 June 1998, 17 July 1998, and 3 August 1998. JERS-1 SAR uses L-band (23 cm wavelength) microwave. The long wavelength of the JERS-1 SAR allowed it to pass through the rice bodies, and backscatter from the water-filled paddy fields was very low, even when the rice biomass was high.
Use of the JERS-1 SAR data therefore made it easy for us to distinguish the water and land surfaces. We defined paddy fields as area detected to be water surface more than once, and land surface at least once.
shows the results extracted from LANDSAT TM data, and
shows the results from JERS-1 SAR data.
In October 2000, we checked the accuracy of our differentiation of paddy
fields. JERS-1 gave the errors that there are big houses attached paddy fields
the fields were judged as urban area. When the JERS-1 SAR resolution was higher,
the error was lower, because houses and paddy fields could be easily distinguished. (G. Saito)