원격탐사 영상을 이용한 작물생산량의 공간적 변이 모의를 위한 GRAMI-rice 모형의 적용 Application of GRAMI-rice for Projecting Spatial Variations of Crop Production using Remote Sensing Imageries원문보기
Remote sensing and crop modeling are two separate techniques widely used for monitoring crop conditions. Remote sensing data are sensitive to atmospheric conditions and mostly contain various reflective characteristics of various objects in an image. Processes attenuating the atmospheric effects and...
Remote sensing and crop modeling are two separate techniques widely used for monitoring crop conditions. Remote sensing data are sensitive to atmospheric conditions and mostly contain various reflective characteristics of various objects in an image. Processes attenuating the atmospheric effects and categorizing all pixels in an image must be conducted to improve quality of remotely sensed data. Meanwhile, combining remote sensing and crop modeling techniques may strength an advantage of each other, making up a weakness of each other. GRAMI, a model able to use remote sensing data, is one of such efforts. The main objective of this study was to investigate capabilities of GRAMI-rice of monitoring canopy growth and estimating yield of paddy rice at small and regional scales. Satellite images from RapidEye (BlackBridge, Inc., Germany) were obtained for paddy fields at Chonnam National University (CNU), Gwangju as well as at TaeAn, Chungcheongnam-do, Korea. Atmospheric corrections of these RapidEye images were performed using three different methods of QUick Atmospheric Correction (QUAC), Fast Line of Sight Atmospheric Analysis of Hypercubes (FLAASH), and Atmospheric and Topographic Correction (ATCOR). The corrected RapidEye images using these methods were evaluated in comparison with the unmanned aerial vehicle (UAV) images that were previously verified for the reliability of the accuracy. In addition, classification of the RapidEye satellite images was carried out using the parallelepiped, minimum distance, maximum likelihood methods as well as a NDVI thresholds method. A digitized paddy cover map was used as standard reference data for evaluation of the classification methods. Crop growth was simulated for applications at a small scale using the UAV imageries taken from paddy fields at CNU as well as at a regional scale using the RapidEye satellite imageries taken at TaeAn. The atmospheric correction results of ATCOR for the satellite imageries were well corresponding to those from the UAV imageries. The minimum distance classification method well categorized all pixels into corresponding reference endmembers with an overall accuracy range between 78.62 and 89.54 %. While the three classification methods above mentioned couldn’t categorize same pixels for different time-series imageries, the NDVI thresholds method made it possible to classify same pixels. Simulated leaf area index (LAI) and above ground dry mass (AGDM) values were in statistical agreement with the corresponding measurements under different nitrogen applications. In LAI, root mean square error (RMSE) and model efficiency (ME) values ranged from 0.26 to 0.38 m2 m-2 and from 0.91 to 0.95, respectively. In AGDM, RMSE and ME values ranged from 18.13 to 2.65 kg ha-1 and from 0.89 to 0.99, respectively. Simulated grain yields were also in reasonable agreement with the measured field values with an RMSE of 394.14 kg ha-1 and a ME value of 0.60. Maps of paddy rice growth and yield appear to represent spatial variations of the corresponding field conditions. These results demonstrate the applicability of the GRAMI-rice to monitoring and mapping of rice growth and yield at different spatial scales
Remote sensing and crop modeling are two separate techniques widely used for monitoring crop conditions. Remote sensing data are sensitive to atmospheric conditions and mostly contain various reflective characteristics of various objects in an image. Processes attenuating the atmospheric effects and categorizing all pixels in an image must be conducted to improve quality of remotely sensed data. Meanwhile, combining remote sensing and crop modeling techniques may strength an advantage of each other, making up a weakness of each other. GRAMI, a model able to use remote sensing data, is one of such efforts. The main objective of this study was to investigate capabilities of GRAMI-rice of monitoring canopy growth and estimating yield of paddy rice at small and regional scales. Satellite images from RapidEye (BlackBridge, Inc., Germany) were obtained for paddy fields at Chonnam National University (CNU), Gwangju as well as at TaeAn, Chungcheongnam-do, Korea. Atmospheric corrections of these RapidEye images were performed using three different methods of QUick Atmospheric Correction (QUAC), Fast Line of Sight Atmospheric Analysis of Hypercubes (FLAASH), and Atmospheric and Topographic Correction (ATCOR). The corrected RapidEye images using these methods were evaluated in comparison with the unmanned aerial vehicle (UAV) images that were previously verified for the reliability of the accuracy. In addition, classification of the RapidEye satellite images was carried out using the parallelepiped, minimum distance, maximum likelihood methods as well as a NDVI thresholds method. A digitized paddy cover map was used as standard reference data for evaluation of the classification methods. Crop growth was simulated for applications at a small scale using the UAV imageries taken from paddy fields at CNU as well as at a regional scale using the RapidEye satellite imageries taken at TaeAn. The atmospheric correction results of ATCOR for the satellite imageries were well corresponding to those from the UAV imageries. The minimum distance classification method well categorized all pixels into corresponding reference endmembers with an overall accuracy range between 78.62 and 89.54 %. While the three classification methods above mentioned couldn’t categorize same pixels for different time-series imageries, the NDVI thresholds method made it possible to classify same pixels. Simulated leaf area index (LAI) and above ground dry mass (AGDM) values were in statistical agreement with the corresponding measurements under different nitrogen applications. In LAI, root mean square error (RMSE) and model efficiency (ME) values ranged from 0.26 to 0.38 m2 m-2 and from 0.91 to 0.95, respectively. In AGDM, RMSE and ME values ranged from 18.13 to 2.65 kg ha-1 and from 0.89 to 0.99, respectively. Simulated grain yields were also in reasonable agreement with the measured field values with an RMSE of 394.14 kg ha-1 and a ME value of 0.60. Maps of paddy rice growth and yield appear to represent spatial variations of the corresponding field conditions. These results demonstrate the applicability of the GRAMI-rice to monitoring and mapping of rice growth and yield at different spatial scales
※ AI-Helper는 부적절한 답변을 할 수 있습니다.