Analysis And Discrimination Model Construction Of Flower Scent Intensity Type For Flowering Crabapple

Crabapple（Malus）is an important ornamental plants in landscape application.However,at present,researchers pay more attention on flower color,flower shape and flowering period of crabapple,rather than flower fragrance.Therefore,in this study,floral scent emission characteristics and emission mechanism of crabapple（Malus’Prairie Rose’）were analyzed.Meanwhile,scent intensity type of 108 Malus crabapple flowers（including 13 species and 95cultivars）were evaluated by sensory evaluation and electronic nose technology.Fragrance compounds of Malus germplasm with different scent intensity type were detected by headspace solid phase microextraction and gas chromatography-mass spectrometry（SPME-GC/MS）,and the main differences were compared.Then,scent intensity type identification model and relative content of main compounds prediction model were established by random forest algorithm.Our aim is to establish the relationship of human nose and E-nose,which can provide a new technical method for the analysis of flower scent type.The main results are as follows:（1）Volatile compounds in M.’Prairie Rose’flowers at different flowering stages,in different floral organs and at different times of the day and night were detected by SPME-GC/MS.2-phenylethanol,geranylacetone,methylheptenone,toluene,and 2-methylnaphthalene were the floral dominant compounds of’Prairie Rose’,which smelled like rose and wood.These compounds were mainly emitted from petals,and a few of them were emited from stamens.During the day,the amount of characteristic aromas released was maximized,especially between 10–12 am,when the flowers from the initial flowering stage to the full bloom stage were in the optimal pollination period（GC peak area>7×10⁸）.Our findings are important for understanding the mechanism of insect visits to crabapple and the regulation of crabapple flower scent.（2）Sensory evaluation of flower scent intensity and preference,as well as selection of strong fragrant flower germplasm were conducted.All 30 trained evaluators showed high consistance in floral intensity sensory test（R²>0.99）.All 108 crabapple taxa had aromatic odor.Of these,42.6%cultivars had medium floral scent,16.7%cultivars had strong and very strong scent,especially M.tschonoskii,M.’Sugar Tyme’and M.’Dolgo’had very strong floral scent.With increasing of flower scent intensity（A₁→A₅）,the difference between the evaluators was smaller.Based on the sensory preference evaluation of 100 evaluators,it was found that most of crabapple taxa（56.5%）aroma were accepted by the public,especially M.’Sugar Tyme’.36.1%of crabapple taxa had no obvious preference for aroma,and 7.4%of crabapple were not popular,especially M.’Roudph’.The aroma of crabapple was more liked by the public,and the variation among the evaluation result of the evaluators was smaller.However,there was no significant correlation between aroma intensity and popular preference（R²=0.04）.According to the comprehensive evaluation of floral fragrance,50.9%of taxa not only had high scent intensity,but also were popular among the public.Among them,M.’Weeping Madonna’and M.’Sweet Tyme’were especially excellent,while M.tschonoskii and M.’Roudph’had a strong aroma,the public did not like it.（3）The optimal measurement conditions for the flower scent using the E-nose showed that the impact degree of various factors on the sensors of E-nose was:sampling time（A）>headspace equilibrium temperature（D）>flower stage（B）>sample concentration（C）>equilibration time（E）.Only the equilibration time had no significant effect on the determination.The most important factor was sampling time.The optimal condition was A₂B₃C₄D₂E₁,namely when the flowers were in blooming,4g of flowers were collected during 10:00-12:00 am,and the flower sample could be determined after equilibrating 1h under 20℃.（4）The"fingerprint"analysis of crabapple based on electronic nose technology showed that the overall"fingerprints"of 120 crabapple taxa were similar,but there was a significant difference in response intensities of same sensor for different taxa（P<0.0001）.There was also a significant difference in response intensity among different dimensions（sensors）for same taxa（P<0.0001）.Most of sensors had higher responsivity,and a few of sensors（W6S and W5S）had less responsivity.The average response value of 120 taxa were from 1.0 to 10.1.The results of PCA analysis showed that 108 taxa were classified into five major groups,and the distinguish power of five groups reached 0.75 to 0.98.Combined with the results evaluated by sensory evaluation,it was found that except for 12 taxa,the other 96 taxa were correctly classified（correct rate 88.9%）.Moreover,the E-nose showed excellent discrimination ability in species and cultivars of crabapple（distinguish power 0.99）.（5）The identification of the fragrant compounds based on SPME-GC/MS showed that 109compounds were detected in 26 crabapple taxa.Aliphatics,benzenoids and terpenes were the dominant compound classes of scented crabapple.Benzyl alcohol,linalool,cis-3-hexenol,benzyl benzoate and benzaldehyde were the major compounds of crabapple flower fragrance.The differences in relative contents of the compounds among different taxa were significant（P<0.0001）.Combined with the results evaluated by sensory evaluation,it was found that as the intensity of scent increased,the number of compound kinds showed a decreased trend.Based on the PCA analysis of relative content of 109 compounds,26 taxa were divided into five groups（a,b,c,d,and e）,and there was a significant correlation between the groups and the floral intensity type（R=0.63）.From the group a to the group e,the relative content of aliphatic and terpenoids decreased significantly（P<0.0001,P=0.047）,the relative content of benzene increased significantly（P<0.0001）,and the differences in polycyclic aromatic hydrocarbons,nitrogen and sulfur-containing compounds were not significant（P=0.181～0.700）.（6）Based on the sensory evaluation data and electronic nose response value,the predition effect of scent intensity of crabapple flower using multiple linear regression,BP artificial neural network and random forest models were compared.Meanwhile,five random forest models was established to predict the relative contents of five main chemical components of crabapple,respectively.The results showed that the prediction effect of multiple linear regression was the worst（33.3%）,and random forest model was more suitable for the prediction of flowering intensity and the discrimination of scent intensity type of crabapple,which prediction error was less than0.53,and discriminant accuracy of scent intensity type reached 81.8%.Especially,after parameter optimization by differential evolution algorithm,the prediction error of the model decreased to less than 0.15,and the prediction accuracy could reach 90.0%.The relative content of main components were predicted well by random forest model,especially the prediction error of the relative content of benzyl benzoate was less than 0.23.