High Throughput Phenotyping Using UAV Based Remote Sensing And Its Application For QTL Mapping In Bread Wheat

Phenotyping is increasingly perceived as a bottleneck due to inefficiency in application of phenotyping platforms and high cost.Field phenotyping must be resource-efficient,robust,precise and digitally logged to support selection in breeding.Prediction of the complex traits like grain yield in bread wheat can be improved by adding information on the dynamics of its constituent secondary components.Accurate and precise phenotyping of novel traits from diverse populations is also necessary for high resolution linkage mapping and genomic selection in wheat improvement.Commercial availability of unmanned aerial vehicles(UAV)and the continued improvement of sensors have opened new opportunities to adopt high-throughput phenotyping in multi-environment trials.There are two types of canopy traits that can be measured from UAV platforms.First,in the form of multispectral vegetation indices using spectral bands(NIR,RED,Red-edge,Blue,Green)that are surrogate to green biomass,different pigments and chlorophyll contents.Second types of canopy traits are digital surface model base traits like plant height.Therefore,introduction of different vegetation indices linked with important crop traits by using different proportion of spectral band could also help to increase the choice of traits for breeders during selection.This study aims to evaluate a multispectral sensor mounted on unmanned aerial vehicle for filed high through phenotyping of wheat physiological traits such as red-edge index(NDREI)along with four other spectral vegetative indices(SVIs);green normalize difference vegetation index(GNDVI),simple ratio(SR),green chlorophyll index(GCI),red-edge chlorophyll index(RECI)for within season perdition of grain yield and time-series senescence rate for accurate selection of potential genotypes.While,a plant height model(PHM)was also developed for assessment of UAV-based plant height.A time series plant height development was assessed using UAV and validated from ground truth data,while accuracy of UAV-based plant height was evaluated for quantitative genomic analysis.UAV-based data was used to validate with ground truth information.In Trial I,a set of 32 bread wheat genotypes were planted under two water treatments(full and limited irrigation)with three replications at Beijing(Shunyi district),China 2016-2017.In Trial II,198 doubled haploid(DH)lines derived from Zhongmai 895/Yangmai16 was sown in three replications at two locations in Henan province for two years 2016-2017 and 2017-2018.The key conclusions are presented below:1.Significant correlations between UAV-based NDVI and Greenseeker-based NDVI ranging from R2=0.38 to 0.90 were observed across the developmental stages under both water treatment.UAVbased NDVI had shown significant phenotypic variations among the genotypes in predicting within season grain yield from early to mid-grain filling with broad sense heritabilties ranging from h2=0.71 to 0.95.Grain yield was predicted from UAV-based NDVI through linear regression analysis,there were high R2 value at EGF(R2=0.86),MGF(R2=0.83)and LGF(R2=0.89)stages,and results were consistent with GS-NDVI.Whereas,four genotypes(Nongda 211,Nongda 5181,Zhongmai 175 and Zhongmai 12)were selected as high yielding genotypes across the water treatments using UAV-NDVI during maturation.2.Time-series senescence rate was predicted from UAV-based normalized difference red-edge index (NDREI)along with four other spectral vegetative indices(SVIs)such as GNDVI,SR,GCI,RECI calculated through four different spectral bands captured from canopy reflectance.Senescence rate was estimated by decreasing values of SVIs from their peak values at heading,while variance for senescence rate among genotypes could be explained by SVIs variations.All five SVIs had explained significant variations for chlorophyll content,leaf area index and canopy temperature with high R2 values ranging from 0.69 to 0.78 with handheld measurements with high heritability under both full and limited water treatments.SVIs also had significant correlations raging from r = 0.23 to0.63 with thousand grain weight and grain yield across the water treatments.NDREI had shown similar trend in senescence rate as observed for four other SVIs.Whereas,principle component analysis corroborated the negative correlation between high senescence rate with thousand grain weight and grain yield.Some genotypes(Beijing 0045,Nongda 5181,and Zhongmai 175)were also selected with low senescence rate and stable grain yield under both full and limited water treatments.3.UAV-based plant height had shown significant phenotypic variations among the two parents and DH lines by high R2 value(0.96)with ground truth observations and low root mean square error at mid grain filling stage.Similar quantitative genomic analysis result in QTL identification and genomic prediction were obtained in both UAV-based and ground truth plant height observations.Two QTL on chromosome 4B corresponding to Rht-B1 and on 4D linking with Rht-D1 on had explained quite similar phenotyping variation 61.3% and 64.5% when validated UAV data with ground truth observations.Whereas,two stable QTL on chromosome 6DL(383.7-391.1c M)were also identified at the booting stage using UAV-based data.Genomic prediction accuracy for both UAV and ground-based data sets was significantly high,ranging from r = 0.47-0.55.UAV-based plant height phenotyping allowed temporal estimation of plant height and its quantitative genomic analysis that could increase the power of genomic selection in crop breeding.Conclusively,UAV-based high through phenotyping is effective,cost-efficient approach which can help in selection of correct traits,increase the selection intensity and accuracy and precise genomic analysis to increase the genetic gain.