Typhoon 19 of 2019 hit Koriyama City in Fukushima Prefecture, Japan, on October 13, 2019. The outflow of the rivers caused flood damage to built-up areas in the city center, and rice production was also damaged because rice paddies were flooded just before harvest time. This study applied a learning-based method to detect flooding in both built-up and rice paddy areas by using changes in backscattering coefficients before and during the flood based on Sentinel-1 synthetic aperture radar (SAR) data. Both built-up areas and rice paddies damaged by Typhoon 19 were used for training and test data. We used changes in these SAR data for training and used a support vector machine (SVM) as a classifier to detect flood damaged areas. The combination of changes in backscattering coefficients and texture (entropy) information improved the accuracy of flood detection by a kappa coefficient of 0.15, compared with backscattering-only input. In addition, a comparison of F values in each category in the results of dual and single polarization demonstrated that VV polarization improved the accuracy of extracting data on flooded built-up areas, while VH polarization improved data extraction for flooded rice paddy areas.

JAXA currently operates a total of six Earth-observation satellites: the Advanced Land Observing Satellite-2 “DAICHI-2” (ALOS-2), the Global Change Observation Mission – Climate “SHIKISAI” (GCOM-C), the Global Change Observation Mission – Water “SHIZUKU” (GCOM-W), the Greenhouse gases Observing SATellite “IBUKI/IBUKI-2” (GOSAT/GOSAT-2), and the Global Precipitation Measurement/Dual-frequency Precipitation Radar (GPM/DPR). The data acquired by these earth observation satellites are distributed through data delivery systems such as G-Portal.

However, this data distribution is limited to scene-by-scene units, and when long-term data or data from multiple sensors are used in a composite manner, large amounts of data must be downloaded locally; moreover, further pre-processing such as geometric conversion and radiation correction is required. This is a technical hurdle for new users who are not familiar with satellite data.

In order to improve the usability of the data distributed directly by JAXA, we should follow the FAIR (“Findable,” “Accessible,” “Interoperable,” and “Reusable) principles, which are generally referred to as the appropriate release method for open data. In order to properly publish open data based on the FAIR principle, it is important to distribute data as much as possible from the same site, in the same format, and without waste, in order to expand the number of users who use the data across the board.

We have thus developed and released the “JAXA Earth API,” a service that enables easy access and use of JAXA’s various satellite data (open data) through an application programming interface (API). This API enables users who are not aware of satellite data-naming conventions, resolution, format, map projection methods, etc. to immediately acquire, visualize, and use data. Herein we introduce the JAXA Earth API system specifications, the web applications, and the response after its release.

Machine learning has recently come into widespread use for the highly accurate classification of cultivated land and land cover using satellite data. Accurate classification requires a sufficient amount and quality of training data, but the collection of data for training is very costly. Therefore, to evaluate the relationship between the required amount and quality of training data and classification accuracy, this study evaluated paddy rice discrimination in California, US, using ALOS-2 PALSAR-2 data with random forest in a case study.

The US Department of Agriculture (USDA) Cropland Data Layer (CDL) made considerable training data available on land cover distribution in 2021. The amount of training data was evaluated after the data volume increased from 100 to 100,000 samples. The quality of the training data was determined by randomly replacing a certain percentage of paddy/non-paddy labels in the training data with incorrect labels. This case study then evaluated the correlation between the amount of training data and the accuracy (ACC) of classification. We found that at least 1,000 training samples are necessary to achieve 0.95 ACC stably under the condition of this study. Next, the study evaluated the correlation between the quality of training data and classification accuracy and found that ACC can be maintained above 0.95 for an error ratio of up to 20 % if there are more than 1,000 samples.