Projection of paddy rice productivity using remote sensing data and a grid-based crop model = 원격탐사 자료와 격자기반 작물 모형을 이용한 벼 생산성 추정

다국어 초록 (Multilingual Abstract)

The principal issue for food security continues to be of interest because the global population is steadily increasing and crop productivity is expected to decrease. Accordingly, it is necessary to understand the current circumstances for stable food production and to take countermeasures to satisfy the food demand in the near future through accurate and continuous monitoring of crop productivity. Crop productivity has been traditionally monitored based on statistical data of a crop obtained from a ground survey method. However, limitations of this technique are that it is time-consuming and labor-intensive as well as it cannot consider a spatial variation of fields of interest or an inaccessible area. An image-based remote sensing technique based on aerial or satellite data has been proven to be an excellent approach to complement the limitations of the ground survey method as it enables timely, efficient, and convenient monitoring of crop productivity. Among the various approaches to monitoring crop productivity using image-based remote sensing data, a combination with a crop model is a better promising approach due to being able to reflect crop physiological characteristics using biophysical parameters.
This study aims to examine whether (1) remotely sensed images from an unmanned aerial system (UAS) can accurately monitor rice growth conditions based on spectral data, (2) a crop model combined with the UAS-based images can accurately simulate rice growth conditions and yield in a field scale, and (3) the model combined with satellite data can reproduce rice yield and production from regional to national scales.
For monitoring of rice growth conditions based on spectral data, a UAS was constructed, which includes an unmanned aerial vehicle (UAV), a multispectral camera, a real-time monitor, and handcrafted portable calibration boards for post radiometric correction using an image processing technique. UAS-based monitoring and field experiments were performed in paddy fields at Chonnam National University (CNU), Gwangju, South Korea in 2013. Spectral reflectance images from the UAS were in statistically acceptable agreement with measured paddy data with a Nash-Sutcliffe efficiency (NSE) range from -37.75 to 0.99 and a root mean square error (RMSE) range from 0.01 to 0.11, respectively. Also, UAS-based normalized difference vegetation indices (NDVI) well represented canopy growth conditions of paddy in fields treated with three different nitrogen regimes. The GRAMI-rice crop model combined with the UAS-based images was used to simulate rice productivity. The GRAMI-rice model was designed to spatiotemporally simulate above ground dry mass (AGDM), leaf area index (LAI), net primary production (NPP), and yield of paddy rice. The model was calibrated using data obtained at the CNU paddy fields in 2013 and applied to simulate paddy productivity grown with conventional farm management practices at the Gimje plain in South Korea in 2014. In model evaluation results, NSE values for all the variables of interest for rice growth and productivity ranged from 0.113 to 0.955. RMSE values between simulated and observed grain yields ranged from -247 to 456 kg ha-1. Also, a study to simulate rice yield and production on a national scale was performed using the GRMAI-rice model combined with satellite images. Data used as input parameters for the model are as follows: (1) vegetation indices (VIs) and solar insolation data estimated from the geostationary ocean color imager (GOCI) and the meteorological imager (MI) of the communication ocean and meteorological satellite (COMS), (2) air temperatures from the Korea local analysis and prediction system (KLAPS), and (3) distribution maps for paddy fields and transplanting dates estimated from the moderate resolution imaging spectroradiometer (MODIS). Estimated paddy fields were in good agreement with the Korea land cover map. The model was calibrated to simulate rice yields using data obtained from 11 counties and applied to 62 counties with an area of more than 5,000 ha in South Korea for four years from 2011 to 2014. Simulated rice yields were in statistically acceptable agreement with the observed data with an NSE range from -0.208 to 0.553 and an RMSE range from 0.326 to 0.441 ton ha-1, respectively. Also, rice productions in 73 counties including the calibration sites were reproduced with an NSE range from 0.668 to 0.698 and an RMSE range from 25.22 to 33.00 kt ha-1, respectively. In conclusion, the present research demonstrated that reliable projection of rice productivity could be achieved from fields to national scales using a crop model combined with image-based remote sensing data.

목차 (Table of Contents)

• ABSTRACT XII
• 1. Preface 1
• 2. Research problems and objectives 4
• 3. Research contents and results
• Chapter I. Construction of an Unmanned Aerial Vehicle Remote Sensing System for Crop Monitoring
• Abstract 6
• 1. Introduction 7
• 2. Materials and methods
• 2.1. Construction of a simple UAV remote sensing system
• 2.1.1. UAV and associated equipment 10
• 2.1.2. Multi spectral camera 10
• 2.1.3. Reference boards 12
• 2.1.4. Multispectral radiometer 12
• 2.2. Image acquisition and processing
• 2.2.1. Image acquisition 13
• 2.2.2. Image processing 14
• 3. Results
• 3.1. Evaluation of UAV-based reflectance images 16
• 3.2. Monitoring crop growth using UAV-based NDVI maps 17
• 4. Discussion 19
• 5. Conclusion 22
• References 23
• Chapter II. Application of an Unmanned Aerial System for Monitoring Paddy Productivity using the GRAMI-rice Model
• Abstract 39
• 1. Introduction 41
• 2. Study sites 44
• 3. Materials and methods
• 3.1. UAS platform 45
• 3.2. Grid-based GRAMI-rice model 45
• 3.3. Dataset
• 3.3.1. Field data 47
• 3.3.2. UAS images 48
• 3.4. Statistical analysis 50
• 4. Results
• 4.1. Evaluation of parameterized model 51
• 4.2. Practical application 52
• 5. Discussion
• 5.1. Within-season calibration of the crop model 53
• 5.2. Simulation performance 53
• 5.3. UAS remote sensing 55
• 6. Conclusion 57
• Reference 58
• Chapter III. Simulation of rice yield and production using the GRAMI-rice model and GOCI imageries for South Korea
• Abstract 80
• 1. Introduction 82
• 2. Study area and data
• 2.1. Study area 86
• 2.2. Data collection and manipulation
• 2.2.1. KME LC and STRM DEM 87
• 2.2.2. MODIS surface reflectance and land cover products 87
• 2.2.3. KLAPS air temperatures 88
• 2.2.4. COMS reflectance and solar insolation 89
• 3. Methods
• 3.1. Detection of paddy fields and transplanting dates 91
• 3.2. Simulation of rice yield and production 93
• 3.3. Statistical analysis 95
• 4. Results
• 4.1. Estimated spatial distribution of paddies and transplanting dates 97
• 4.2. Evaluation of simulated rice yields 98
• 4.3. Evaluation of simulated rice production 99
• 5. Discussion 100
• 6. Conclusion 105
• Reference 106
• ABSTRACT (in Korean) 133