Paddy Rice Mapping Based On Phenological And Temporal Features

Paddy rice agriculture plays an important role in several environmental issues including greenhouse gas emission,food security,water use and disease transmission.However,regional paddy rice maps are often inaccurate and surprisingly scarce in spite of continuous works and advancement of paddy rice mapping algorithms and applications.This thesis aims to advance the analysis of satellite sensors for paddy rice mapping.This is accomplished by developing efficient methods and approach suitable for large amount of data,and verified in the regions containing ideal paddy rice fields.First,the effect of phenology in paddy rice mapping were investigated under the object-based image analysis(OBIA)paradigm.Crop phenology was combined with the high spatial resolution multi-spectral data to accurately classify the rice crop.Phenology was used to capture the seasonal dynamics of the crops,while multispectral data provided the spatial variation patterns.Phenology was extracted from MODIS NDVI time series,and the distribution of rice was mapped from China's Environmental Satellite(HJ-1A/B)data.The results indicated improved classification accuracy when using the phenology with the spectral data.Second,temporal features extracted from coarse resolution data were evaluated in paddy rice classification of fine resolution data.A total of 33 temporal features were analyzed to find out their contribution in paddy rice mapping.The results indicated that the use of temporal feature improved the overall classification accuracy of a single date multi-spectral image by 18.75% from 65.62% to 84.37%.The comparison showed that the mapped paddy area was analogous to the agricultural statistics.The study highlighted the importance of feature selection to achieve higher classification accuracy.Overall,obtained results demonstrate the potential of the combined use of temporal and spectral features for accurate paddy rice classification.Third,four classification strategies(i.e.,single-date image,phenological bands,phenological metrics and textural feature)were evaluated to characterize the intra-annual phenological dynamics and effects of textural features on paddy rice detection.The objective of the study was to develop new phenology and textural-based strategies to detect paddy rice from HJ-1A and MODIS imagery.Results indicated that the phenology-based strategies could hold the phenological variation within paddy rice and significantly improved the detection accuracy.Seasonal amplitude,GLCM(Grey-Level Co-occurrence Matrix)entropy and spectral features of the heading stage were proved to be important in identifying paddy rice at 30 m spatial resolution.This study demonstrated the integration of 250 m phenolgical feature with the 30 m spectral data of HJ-1A images.It is concluded that combined HJ-1A and MODIS time series imagery are promising in facilitating the rapid mapping of paddy rice at a regional scale.Finally,a method was presented to determine the timing of flooding and transplanting of rice over a year using EVI(Enhanced Vegetation Index)time series to improve the existing phenology-based approach.The paddy rice areas were mapped at a fine resolution(16 m)utilizing time series MODIS and GF-1 images.The EVI from the time series MODIS images were used to determine the time window of flooding and rice transplanting phase in order to select the fine resolution GF-1 images.The paddy rice areas were identified based on the unique physical features of rice plant during the transplanting and flooding phase by capturing the seasonal dynamics using the Normalized Difference Vegetation Index(NDVI)and the Normalized Difference Water Index(NDWI).The mapping was performed on the parcel level.This study demonstrated a robust,simple and automated approach to identify and map paddy rice fields in the tropical zones.Overall,this thesis demonstrates the usefulness of phenological and temporal features in paddy rice detection.