Study On Remote Sensing Monitoring And Assessing Oilseed Rape Freezing Injury

Oilseed rape(Brassica napus L.)ranks first among the oil crops in China,and China has the second largest cultivated area and yield of oilseed rape in the world,in which about 90%of the acreage is contributed by winter varieties.The wintering period is necessary for winter oilseed rape,as a result,it suffers the risk of freezing injury during the wintertime.Traditional field campaign usually measures the damage directly in the fields and get the freezing injury index,but it is time-consuming,expensive,and in poor representation.Remote sensing is an efficient technical means of acquiring large-scale surface information in time,and it has distinct advantage in crop damage monitoring and assessment.This research selected Anhui Province—situated in traditional major oilseed rape producing areas,as the study area,and employed multisource remote sensing data acquired from EOS-MODIS,ENVISAT-MERIS and Chinese HJ-CCD/IRS sensors,to explore the method of assessing oilseed rape freezing injury from multiple perspectives.The historical freezing injury events occurred in the study area,which spanned from 2003-2004,2009-2010 and 2010-2011 growing seasons,were selected as the case studies.The main research contents and the main achievements are as follows:(1)The judgment of freezing injury-prone areas based on low-temperature indicatorThis research achieved the judgment of freezing injury-prone areas from the foremost damage-inducing factor—low temperature,according to the national standard of daily minimum temperature for different freezing-damage levels,and the solution of multisource data(MODIS data,geographic elements data and temperature observation data)combination was applied.Taking the freezing damage of oilseed rape occurred in the last 10-day of January and first 10-day of February,2004 as a case study,the thesis explored the methodologies of obtaining the gridded daily minimum temperature under clear and cloudy weather conditions.For clear-sky regions,the spatial distribution of daily minimum temperature was derived from multiple linear regression with land surface temperature(LST),NDVI and Julian day(DOY),and the daily acquired data in the winter months during 2000-2014 were taken as the input.Results showed that the equation established based on Terra-LST acquired in the previous night had the best performance,followed by daytime Terra-LST,daytime Aqua-LST and nighttime Aqua-LST of the specific date.For overcast areas,the daily minimum temperature under cloudy sky was estimated step by step through annual 10-day low temperature background + low-temperature departure from annual level + residual spatialization approach.The annual low temperature level of each 10-day in winter months(from early December to early March)was simulated based on the model established from the latitude,longitude,altitude,slope and aspect data of the study area.The discrepancies of 8-day composite LST(MOD11A2)in late January and early February,2004 from the annual average level(2000-2014)were regarded as the temperature departures in the freezing-experienced year,with the support of HANTS filtering.The residual between daily minimum temperature on each cold date and the low-temperature level of the corresponding 10-day was gidded by means of geographic elements data,and daily minimum temperature observed by 78 county-level meteorological stations throughout the study area was adopted.Multiple linear regression and random forest simulation were applied successively and their performance in estimation effect was also investigated,results showed that random forest model outperformed multiple linear regression obviously,it was then implemented so as to interpolate temperature residual to the whole study area.For the case of partly cloud coverage,the output give priority to the results derived from clear-sky algorithms,and the estimated result under cloud was taken as supplementary,the daily minimum temperature in full coverage could thus obtained,1 km gridded.Considering that the duration of low-temperature weather has not been defined by national standard,this research employed cold accumulated temperature during low-temperature process to judge the damage-prone areas.Results showed that it could reveal the distribution of extremely low temperature in Suzhou area(northern Anhui)and Hefei,Chuzhou regions(Yangtze-Huaihe area),and good effect was achieved in terms of cold accumulated temperature estimation,by correlation analysis with the observed values(r = 0.810,P<0.01,RMSE = 8.6?).(2)The identification and extraction of oilseed rape by remote sensingThe greenness of oilseed rape declines from initial flowering to full-bloom stage,and thus a frame of 'trough' appears on the temporal profile of vegetation index,which is different from those of other winter crops(mainly wheat)growing simultaneously,the growing regions of oilseed rape could be extracted according to the unique change trend of greenness between these two phenological phases.In light of the different crop phenology at medium to large spatial scales,the start time of these key phenological phases were first identified through time-series analyses adopting the S-G filtered NDVI time series data in 8-day composite,and then the oilseed rape was mapped accordingly,the results were validated by the recorded sown area of oilseed rape at the municipal level provided by statistical yearbooks.Research shows that the oilseed rape plant suffered severe freezing injury would likely get a lower height and have sparse flowers,and the overall tone in flowering phase is less prominently yellow,which is easily confused with winter wheat at 30 m-resolution pixel scale,causing obvious underestimation of the actural acreage.Crop damage assessment has the requirement of timeliness,the thesis proposed a so-called 'benchmark growing regions + overwintering-crop mask adjustment' method to obtain the growing regions of oilseed rape in damage-experienced growing season,combining with the actual situation of local agricultural practices.This method is based on the theoretical assumption that only winter crops would grow after emergence until reaching the wintering period,and the oilseed rape distribution in the target growing season could be simulated with the benchmark growing regions in the adjacent 'normal' growing season(freezing-damage free).The application effect was examined based on the sown area of oilseed rape from statistical yearbooks.(3)The evaluation of freezing damage intensityFor the freezing damage occurred in early 2004,The ratio relationship was adopted to investigate the sensitivity of the selected four widely-used vegetation indices to freezing injury at MODIS-pixel scale.Results of temporal profile analyses showed that NDVI and GNDVI outperformed the other ?s,and NDVI had a better performance in revealing the influence of freezing damage,which was more in line with the reality.MODIS data has a strong time series continuity,it was appropriate for freezing injury assessment by adopting vegetation index anomaly method.The average growth level during the same time span in the adjacent freezing damage-free growing seasons(2001-2006,excluding 2004)was taken as the benchmark,the fluctuations of growth under normal circumstances was represented by the annual standard deviation of NDVI,and the degree of damage was described by the discrepancy between freezing-damage affected growth level and the benchmark in percentage.The different damage levels were mapped and the remote sensing-derived damage extent was supported by the correlation between average degree of damage at county level and cold accumulated temperature observed by meteorological stations,due to the lack of field-measured damage intensity.A total of 47 oilseed rape producing counties were selected,and significant correlation was observed(r = 0.378,P<0.01),which demonstrated its rationality in a sense.The coarse resolution of MODIS sensors makes it more difficult to show spatial details of crop damage.For the freezing injury occurred in early 2010 and 2011,the data acquired from Chinese HJ sensors were applied to damage assessment.Eight vegetation indices generated from the four working bands assembled on HJ-CCD sensors were taken as candidate damage indicator.Taking into account the availability of remote sensing data,this research adopted different strategies to represent the baseline growth level,so as to evaluate the degree of damage.The sensitivity of each VI to cold stress was investigated by means of the histogram curve of its normalized deviation(post-relative to pre-damage variation)from the baseline variation(2008-2009 growing season).Results showed that NDVI and GNDVI were more sensitive to freezing injury as well at 30 m pixel-size scale,and GNDVI had better performance compared with NDVI,which was qualified to be the optimal freezing injury indicator.The degree of damage was described by the discrepancy of post-damage GNDVI from the baseline level,and the remote sensing-derived damage intensity was validated by the field-measured freezing injury index for the damage occurred in 2011.The data from nine crop monitoring sites in Hefei were adopted to implement correlation analysis,and a relatively high correlation coefficient was gained between field-measured and remote sensing-derived damage levels(r =-0.698,P<0.05).The gray relationship analysis and statistical analysis approaches were applied to evaluate the influence degree of several natural and crop factors to damage intensity,results showed that pre-freeze growth,soil moisture,LST of the coldest day,and altitude were in descending order of importance in determining the degree of damage.Furthermore,the oilseed rape planted on south-and west-oriented facing slopes would like to suffer more serious damage.(4)Response pattern of remote sensing physiological indicators to cold stressThis research investigated the performance of several remote sensing indicators of plant biophysical and biochemical parameters,i.e.,the photosynthetic rate,canopy water status,canopy chlorophyll content,leaf area index(LAI)and the red edge position(REP),which were derived from MODIS and MERIS bands,in their responses to freezing damage.Results showed that the photosynthetic rate indicator—Photochemical Reflectance Index(PRI)responded strongly to cold stress,which signified the great sensitivity of photosynthesis to freezing damage;the abnormally increased NDWI after the attack of cold wave indicated that freezing injury caused canopy water loss.MERIS-LAI showed a slow and lagging response to low temperature and restored rapidly in the recovery phase,REP and the indicator of canopy chlorophyll content—MERIS Terrestrial Chlorophyll Index(MTCI)were not reduced by cold stress,it is deduced that the level of canopy chlorophyll content might not decline.However,low temperature suppressed the increase of MTCI,while REP did not appear to be influenced by freezing injury.Based on the above findings,this research selected seven bands from MODIS-L1B product which were responsive to photosynthesis and canopy water status,among them bands 5,11 and 12 were singled out through the OIF method,with the support of random sampling points selected from the oilseed rape fields corresponding to different damage levels.A new index was proposed adopting these three bands,which was termed as the MODIS Freezing Injury Sensitive Index(MFISI).New index responds to both photosynthesis and canopy water status,which is of explicit physiological significance;the impact of freezing injury could be revealed in view of the great sensitivity of both physiological variables to cold stress,and rich ground details could also be captured;in addition,MFISI is superior to the well-recgnized NDVI in terms of the pre-judgment of freezing injury and the prediction of its development tendency.