Research On Rice Multiple Cropping Detection In Cloudy And Foggy Area Using Radar Remote Sensing

Rice plays a crucial role as a staple food in many Asian countries,and its planting area and yield are closely related to a nation’s food security and social stability.Rice multiple cropping refers to planting rice multiple times a year,thereby extending the planting area and enhancing the annual yield.Monitoring rice multiple cropping involves identifying rice-crop intensity and detecting key phenological periods.Rice-crop intensity refers to the annual number of rice growth cycles,while the key phenological periods primarily include the transplanting,heading,and maturity stages.Monitoring rice multiple cropping tracks rice growth cycles and assesses biophysical conditions,providing valuable insights for yield estimation and agricultural decision-making annually.Southern China and Southeast Asia are the primary rice-producing regions in Asia.Benefiting from the warm and humid climate,these regions are conducive to rice multiple cropping.However,the climate often leads to cloudy and foggy conditions,which frequently hinder optical remote sensing.Large-scale rice multiple cropping monitoring usually uses medium or low resolution optical images,which are unsuitable for small and fragmented rice fields in complex terrain.Synthetic Aperture Radar(SAR)provides an all-weather monitoring solution with high spatial and temporal resolution.However,detecting rice multiple cropping using radar data on a large scale is challenging due to complex rice backscatter patterns,diverse cultivation practices,inefficient feature extraction,and lack of prior phenological data.In this dissertation,an efficient scheme for time series reconstruction and feature extraction is developed based on Sentinel-1 SAR data.Leveraging temperature information,a novel method for monitoring rice multiple cropping is proposed without relying on prior phenological information.Consequently,rice crop intensity and multi-cycle rice phenology products for 2020 in Southern China and Southeast Asia are produced.The main research contents and achievements are outlined below(1)A method for time-series radar data reconstruction and feature extraction is developed,combining three-order harmonic decomposition and parameter modeling.By exploring rice growth mechanisms and crop rotation patterns,three-order cosine components are used to reconstruct radar time-series curves,enabling accurate representation of complex rice cultivation practices without the need for setting filtering windows or idealized functions.Through mathematical simulation and synthesis analysis of cosine components,the capability of harmonic parameters to indicate attributes such as the number,amplitude,and timing of troughs and crests in time-series curves is thoroughly investigated.Then,a time-series feature extraction method is proposed,comprising profile pattern recognition and trough/crest date calculation.This method addresses the issue of low computational efficiency associated with traditional point-by-point operations.High consistency(R~2>0.96)between the identified trough-crest dates and actual dates is achieved.(2)A method for identifying rice crop intensity is proposed based on backscatter harmonic analysis and temperature criteria.To address issues such as rice trough/crest date displacement and interference from non-rice land processes,common trends in rice VH backscatter under different cropping practices are investigated,and stable trough features indicative of rice growth cycles are identified.Based on the minimum temperature requirements for rice growth,the suitability of trough temperatures is evaluated.Rice multiple cropping schemes and their accumulated temperature requirements are explored,and a suitability mask for different rice crop intensities in Southern China is created.Rice crop intensity is then identified by combining this mask with temperature-suitable troughs,effectively addressing overestimation caused by interference from non-rice features.A rice crop intensity product for Southern China in2020 is produced,achieving an accuracy of 82.13%.(3)A method for identifying key phenological periods in multiple rice growth cycles is developed based on a depolarization index and simplified harmonic analysis.This method addresses challenges such as the ambiguous phenological attributes of VH backscatter troughs,the complexity of harmonic feature extraction rules,and the limitations of single-cycle phenological monitoring.A systematic analysis is conducted on the response mechanism of VH and VV backscatter to rice growth.This analysis led to the development of a depolarization index capable of tracking rice transplanting,heading,and maturity periods.To enhance efficiency and minimize modeling errors resulting from complex rules,a simplified method for extracting time-series features is established.This method involves evaluating dominant periodicity and weighting cosine components.By assessing depolarization index gains and temperature supply during the trough-crest period,phenological features in different rice growth cycles can be identified.Consequently,rice crop intensity and multi-cycle rice phenology products in2020 are produced for Southern China,achieving high accuracy(83.24%)in rice crop intensity identification and high consistency(R~2>0.94)between the identified phenological dates and the actual dates.(4)A method for detecting rice multiple cropping using multiple SAR features is developed.The low proportion of fields with double and triple rice-crop intensities in southern China limits its ability to fully represent global rice cropping patterns.Therefore,Southeast Asia,where rice multiple cropping is more common,is chosen as the study area.A method for evaluating trough-crest gains of the depolarization index and trough intensity of VH backscatter is proposed for detecting rice multiple cropping.This method utilizes the sensitivity of the depolarization index to plant growth and the sensitivity of VH polarization backscatter to field flooding.Thus,rice crop intensity and multi-cycle rice phenology products in 2020 are produced for Southeast Asia,achieving high accuracy(81.82%)in rice crop intensity identification and high consistency(R~2>0.93)between the identified phenological dates and the actual dates.The method proposed in this dissertation offers several advantages,including independence from prior phenological information,adaptability to complex cropping patterns,and high computational efficiency.These qualities make it a valuable tool for enhancing agricultural production management and ecological status assessment,thereby contributing to more sustainable and efficient agricultural practices.