| Natural rubber(Hevea brasiliensis)is an important production material in the fields of industry,national defense,medical and health care.As the only source of natural rubber production,rubber native to Brazil has the growth habits of thermophilic but weak cold resistance.Although the successful movement planting of rubber trees to the northward,the latitude of rubber planting area in China higher than the place of origin caused great differences of planting environment.Chilling injury has become one of the main disasters in rubber production and the main factor restricting the development of rubber in China.In view of the shortcomings of remote sensing monitoring of rubber chilling injury in China and the urgent demand for chilling injury monitoring in rubber production,taking Yunnan as the research area,where is the origin of rubber planting in China and a typical representative area of rubber plantations planting in the northern edge of the tropics,this paper aims to overcome the current research deficiencies that the methods of remote sensing monitoring of rubber chilling injury based on spatial interpolation of meteorological data mostly,the low spatial resolution of remote sensing images used,and the lack of time series data caused by cloudy and foggy in Yunnan.By filling in the missing pixels of the observation data and spatio-temporal fusion technology obtained the high spatial-temporal resolution remote sensing data,then the high-precision identification of the planting distribution and phenological characteristics of rubber plantations in Yunnan and the remote sensing monitoring of chilling injury are realized based on the data set,which provides method and data support for rubber production and disaster emergency,and is of great significance to ensure rubber safety production and continue to expand rubber planting area.The main research contents and conclusions(research findings)are as follows:(1)A robust remote sensing image invalid pixels in tropical repair method based on a time series images analysis method are proposed,namely,the enhanced triple repairing method.It achieved excellent repair results for the Landsat TM/OLI,Sentinel-2 and MOIDS images(the R~2 value of each band was above 0.8).The repair accuracy of the proposed method is more stable than that of current methods,because it uses certain judgment criteria when automatically retrieving for high-quality reference images in time series data,less affected by specific reference images.The proposed method solves the problems,such as a poor repair accuracy due to cloud and cloud shadow recognition accuracy and judgment criteria on high-quality reference images in the current method,and enabled to use the available valid observation data to achieve the maximal repair of missing information.Thereby,the usable amount of optical remote sensing image data is increased.Based on this method and combined the all-weather surface air temperature products with 1 km spatial resolution,the time series data sets of vegetation index and surface temperature of rubber plantations are reconstructed in falling leaf stage under the complex terrain of tropical mountainous area.After using this method to repair the annual winter remote sensing data in Yunnan rubber planting area,the average loss of MODIS data is only approximately 1%,which meets the requirements of establishing a complete winter remote sensing data set in Yunnan rubber planting area.The reconstructed high spatio-temporal remote sensing surface temperature data is highly correlated with the site measured data.The lowest average temperature inversion accuracy R~2 is 0.88,the average R~2 is 0.91,the average RMSE(Root Mean Square Error)is 1.33?,and the highest deviation is 1.97?;the minimum temperature inversion accuracy R~2 is 0.86,the average R~2 is 0.89,the average RMSE is 1.71?,and the maximum deviation is 2.19?,which all reaching the significance level of 0.01.The fused surface temperature data can keep the same trend with the measured data of the station.The temperature estimation deviation is small,and can accurately capture the complete temperature change process during the chilling injury of rubber plantations.(2)A method for rubber plantations identification in cloudy and foggy areas was proposed on the basis of vegetation phenological characteristics in the paper.The method first analyzed the optical remote sensing window within the study area and then used spatio-temporal data fusion technology to construct a vegetation index data set with high temporal resolution within the remote sensing window.The phenological features of vegetation were captured to map rubber plantations using fused time series indices to realize the high-precision distribution mapping of the planting area of rubber plantations.The vegetation phenological characteristics were determined by analysis of the annual cloud coverage time series curve of the study area and the vegetation index time series of the fusion products.It is found that the period from December to March of the next year is the best period to extract rubber characteristics with long-term remote sensing data.The main land categories in Yunnan rubber planting area can be effectively identified in the period.Among the five commonly used time series indices:TCA(Tasseled Cap Angle),NDVI(Normalized Difference Vegetation Index),EVI(Enhanced Vegetation Index),NDMI(Normalized Difference Moisture Index),NBR(Normalized Burn Ratio),the EVI time series had a better ability to capture the different growing features of vegetation in this study;The highest classification accuracies were achieved for the best fusion results of EVI,which indicated that the EVI was the best choice among the five time series indices for time series features-used rubber classification,and fused optical remote sensing data with high spatio-temporal resolution could map the rubber distribution with high accuracy,overall accuracy of89.51%,kappa of 0.86.The planting area edge of rubber plantations in Yunnan Province approaches 25°to the north,located in Dehong Prefecture.The lowest altitude of Rubber plantations planting is is located in Hekou County,while the highest is over 1500 m.a.s.l.in Pu'er City,far exceeding the traditional rubber planting altitude at 600?1200 m.(3)Based on remote sensing invalid pixel filling and spatio-temporal fusion technology,a method for chilling injury monitoring of rubber is proposed,which integrates remote sensing surface temperature data and vegetation index disturbance characteristics.This method adopts the vegetation index and surface temperature remote sensing data of Landsat and MODIS to realize the remote sensing monitoring and disaster assessment of rubber plantations,and construct the remote sensing monitoring index of rubber forest,including?):vegetation index disturbance characteristics,?):chilling injury disaster causing factors extracted from remote sensing surface temperature data.This method makes up for the deficiency of remote sensing monitoring of rubber chilling injury only based on vegetation index,and the accurate monitoring of rubber chilling injury realized.The previous rubber chilling injury were all identified by the proposed method for nearly 20 years in the study area(in the winter of 2004,2007,2010 and 2015,in the spring of 2005,2008,2011 and 2016),finally,the assessment distribution map of chilling injury prone areas in rubber planting area of Yunnan Province was summarized.The results show that the Hekou area is the area with the highest frequency of chilling injury in the rubber planting area of Yunnan,and Xishuangbanna is the area with the lowest frequency.Most previous chilling injury occurred in higher altitude planting areas,while the traditional rubber planting areas near cities,towns and lower altitude were less affected by chilling injury.The previous new rubber planting areas are mostly expanded to high-altitude mountains with frequent chilling injury,which indicates that rubber planting is constantly expanding into areas unsuitable for rubber planting. |
PDF Full Text Request