5:15 PM - 6:30 PM
[ACG37-P05] Validation of Two Vegetation Indices: NDVI and PRI Derived from GCOM-C/SGLI
Keywords:Remote Sensing, Forest, GCOM-C/SGLI, NDVI, PRI
Forests play a significant role in terrestrial ecosystems and greatly influence the material cycle in terrestrial ecosystems. Therefore, an accurate understanding of forest conditions is fundamental and indispensable for understanding terrestrial ecosystems. In order to accurately understand the state of forests, which cover a large area, local observations are often insufficient. An example of such a method that can cover a large area is satellite remote sensing.
One example of satellite that is used for observation of terrestrial ecosystems is GCOM-C (Global Change Observation Mission-Climate), which was recently launched by JAXA (Japan Aerospace Exploration Agency). GCOM-C is equipped with a sensor called the SGLI (Second-generation Global Imager). This sensor is capable of observing a larger number of spectral bands than the sensors on other earth observation satellites. This means that GCOM-C/SGLI has a higher wavelength resolution. In addition, GCOM-C/SGLI has a wide observation range, which means that the frequency of observations is high, and thus the temporal resolution is high. Therefore, it is expected that GCOM-C/SGLI can be used for more detailed and frequent observation of forest conditions.
Usually, various indices are calculated from satellites' data to understand the condition of vegetation, such as forests. From the GCOM-C/SGLI data, several indices can be calculated, and typical indices derived from the GCOM-C/SGLI data are NDVI (Normalized Difference Vegetation Index) and PRI (Photochemical Reflectance Index).
The NDVI is widely used in remote sensing of vegetation and can be calculated from many earth observation satellites, but the NDVI derived from GCOM-C/SGLI has a feature that multiple NDVI can be calculated.
That is, it is possible to calculate four NDVIs: two in the red region and two in the near-infrared region since GCOM-C/SGLI has two observation bands each. This is a major feature of GCOM-C/SGLI, and the use of these four NDVIs will enable more detailed remote sensing of forests.
PRI is an indicator that reflects the chemical changes of xanthophylls in chloroplasts (xanthophyll cycle). Xanthophylls are closely related to the heat dissipation of excess energy in photosynthesis, and by quantitatively evaluating the xanthophyll cycle and combining it with other vegetation indices, it is thought that the gross primary production of forests can be estimated. Therefore, PRI, which reflects the xanthophyll cycle, is one of the most important indices in remote sensing of forests. However, the PRI is calculated from the reflectance of very close wavelengths, 531 nm and 570 nm, which requires high wavelength resolution. Therefore, it is not easy to calculate PRI from satellite data. Nevertheless, the GCOM-C/SGLI is able to observe these wavelengths independently. So observing the PRI is a great advantage of the GCOM-C/SGLI.
As described above, although the GCOM-C/SGLI data can be used to calculate the NDVI (four types) and PRI, which are significant vegetation indices, the detailed validation of the accuracy of the data has not been done so far. In this study, we focused on two vegetation indices: NDVI and PRI, which can be calculated from GCOM-C/SGLI observation data, and aimed to validate their accuracy.
For the accuracy verification, we used the ground validation data for three years from 2018 to 2020 at several sites in Japan (Teshio, Tomakomai, Takayama, Mase, Fuji-yoshida, and Fuji-hokuroku). These ground validation data were acquired by spectroradiometers installed at each site. In addition, the phenology change of the forest was detected by the photographs of the forest taken by the fisheye camera installed with the spectroradiometer. The relationship between the phenological changes and the changes in each index was investigated.
The results of ground-based verification suggested that NDVI and PRI can be obtained from GCOM-C/SGLI. However, the difference between the ground-based measurements and satellite data was sometimes large, especially for PRI. It was also found that each vegetation index had a unique relationship with changes in forest phenology.
In the future, it will be necessary to develop a method to detect and remove the outliers of NDVI and PRI obtained from GCOM-C/SGLI.
One example of satellite that is used for observation of terrestrial ecosystems is GCOM-C (Global Change Observation Mission-Climate), which was recently launched by JAXA (Japan Aerospace Exploration Agency). GCOM-C is equipped with a sensor called the SGLI (Second-generation Global Imager). This sensor is capable of observing a larger number of spectral bands than the sensors on other earth observation satellites. This means that GCOM-C/SGLI has a higher wavelength resolution. In addition, GCOM-C/SGLI has a wide observation range, which means that the frequency of observations is high, and thus the temporal resolution is high. Therefore, it is expected that GCOM-C/SGLI can be used for more detailed and frequent observation of forest conditions.
Usually, various indices are calculated from satellites' data to understand the condition of vegetation, such as forests. From the GCOM-C/SGLI data, several indices can be calculated, and typical indices derived from the GCOM-C/SGLI data are NDVI (Normalized Difference Vegetation Index) and PRI (Photochemical Reflectance Index).
The NDVI is widely used in remote sensing of vegetation and can be calculated from many earth observation satellites, but the NDVI derived from GCOM-C/SGLI has a feature that multiple NDVI can be calculated.
That is, it is possible to calculate four NDVIs: two in the red region and two in the near-infrared region since GCOM-C/SGLI has two observation bands each. This is a major feature of GCOM-C/SGLI, and the use of these four NDVIs will enable more detailed remote sensing of forests.
PRI is an indicator that reflects the chemical changes of xanthophylls in chloroplasts (xanthophyll cycle). Xanthophylls are closely related to the heat dissipation of excess energy in photosynthesis, and by quantitatively evaluating the xanthophyll cycle and combining it with other vegetation indices, it is thought that the gross primary production of forests can be estimated. Therefore, PRI, which reflects the xanthophyll cycle, is one of the most important indices in remote sensing of forests. However, the PRI is calculated from the reflectance of very close wavelengths, 531 nm and 570 nm, which requires high wavelength resolution. Therefore, it is not easy to calculate PRI from satellite data. Nevertheless, the GCOM-C/SGLI is able to observe these wavelengths independently. So observing the PRI is a great advantage of the GCOM-C/SGLI.
As described above, although the GCOM-C/SGLI data can be used to calculate the NDVI (four types) and PRI, which are significant vegetation indices, the detailed validation of the accuracy of the data has not been done so far. In this study, we focused on two vegetation indices: NDVI and PRI, which can be calculated from GCOM-C/SGLI observation data, and aimed to validate their accuracy.
For the accuracy verification, we used the ground validation data for three years from 2018 to 2020 at several sites in Japan (Teshio, Tomakomai, Takayama, Mase, Fuji-yoshida, and Fuji-hokuroku). These ground validation data were acquired by spectroradiometers installed at each site. In addition, the phenology change of the forest was detected by the photographs of the forest taken by the fisheye camera installed with the spectroradiometer. The relationship between the phenological changes and the changes in each index was investigated.
The results of ground-based verification suggested that NDVI and PRI can be obtained from GCOM-C/SGLI. However, the difference between the ground-based measurements and satellite data was sometimes large, especially for PRI. It was also found that each vegetation index had a unique relationship with changes in forest phenology.
In the future, it will be necessary to develop a method to detect and remove the outliers of NDVI and PRI obtained from GCOM-C/SGLI.