论文部分内容阅读
Abstract [Objective] This study was to provide a theoretical basis for the selection of the parents of proso millet breeding in dry land of south Ningxia， with the aim to improve the production of hilly proso millet in Ningxia hilly area. [Method] The genetic diversity of 46 accessions of dryland proso millet germplasm at 11 agronomic traits were analyzed by using principal component analysis， Shannon??weaver diversity index and non??weighted matching arithmetic average clustering method （UPGMA）. Cluster analysis was also performed. [Result] The proso millet resources mainly had green inflorescence color， panicle spikes， and yellow grain color. The maximum diversity index of proso millet was 2.03. The variation coefficient of grain weight per plant was 19.23. In all the principal components， the main information concentrated on the first 4 principal components with the cumulative contribution rate reaching 83.34%. The tested materials were clustered into 3 categories at euclidean distance of 0.984 4. The first category contained 41 germplasm resources， the second contained 2， and the third category contained 3. [Conclusion] This study could provide references for the targeted use of drought??resistant parents to select hybrid combinations of proso millet in Ningxia.
Key words Dryland proso millet； Germplasm resource； Diversity
Proso millet， Paninum miliaceum L.， is one of the ancient Chinese food crops. It has the characteristics of early maturity， barren tolerance， heat resistance and drought tolerance. It is often used as the reserve crop prepared against natural disasters in areas with poor natural conditions[1]. The distribution of japonica and glutinous proso millet is relatively regular in China that the proportion of glutinous type gradually decreases while the proportion of japonica type gradually increases from east to west， and such trend is related to precipitation[2]. Due to its short growth period， strong resistances and high nutritional value， proso millet has obvious regional advantages and production advantages in semi??arid areas such as Ningxia， Shanxi and Gansu provinces， especially in the loess hilly region of southern Ningxia， where it is one of the minor grain crops and also a pioneer crop with a large planted area on newly reclaimed wasteland[3-4].
Germplasm resources are the basis for genetic improvement and breeding of proso millet varieties. The improvement of proso millet varieties lags far behind other crops. Under the new situation， in order to achieve a major breakthrough in the innovation of proso millet varieties， in addition to strengthening research on variety resources， it also needs to make breakthroughs in breeding technological research and breeding material innovation， to study effective ways and methods for the heterosis utilization of proso millet to cultivate proso millet hybrids， and the future goal of proso millet breeding is to achieve the combination of yield increase， quality improvement and resistance enhancement[5-6]. The identification of germplasm resources is focused on the much??needed characteristics of breeding， eliminating the differences caused by environmental factors and reflecting the true germplasm characteristics. The selection and improvement of proso millet varieties are carried out on the basis of existing germplasm resources. The replacement of proso millet varieties is inseparable from the utilization of high??quality germplasm resources[7-8]. Therefore， the study of genetic diversity of proso millet is particularly important[9]. However， there are few studies on the diversity of dryland proso millet germplasm resources in the mountainous areas of southern Ningxia. Thus， in order to screen out the high??quality， high??yield and special proso millet varieties suitable for planting in the mountainous areas of southern Ningxia， the genetic diversity of 46 drought??tolerant proso millet germplasm resources were analyzed using principal component analysis and cluster analysis from the growth period， morphological characteristics and yield??related factors of proso millet， with the aim to provide theoretical references for the targeted use of drought??resistant parents to select hybrid combinations， thereby providing theoretical basis for the research on proso millet breeding in Ningxia. Materials and Methods
Test materials
The names and sources of 46 dryland proso millet germplasm resources are shown in Table 1， of which the first 1-26 are japonica varieties， and No. 27-46 are glutinous varieties.
Test area overview
For three consecutive years in 2014-2016， test was carried out at Touying test base of Guyuan Branch of Ningxia Academy of Agriculture and Forestry Sciences， Guyuan City. Located at 36??44??， 106??44?? E， the test base had an elevation of 1 586 m. It is in the temperate arid zone with dry and windy spring and cold winter. It has an annual average temperature of 7.4 ??， accumulated temperature above 10 ?? of 2 500-2 800 ??， frost??free period of 130-150 d， annual precipitation of only 350 mm， which is very scarce. The rainfall is mainly distributed in August， September and October. The farming in the area is arid farming that relies totally on natural rainfall. The area has frequent occurrence of drought disasters， large temperature difference between day and night. In this study， the tested soil belonged to Hunan loess with low soil fertility level， and the main physical and chemical properties of the tested soil at the depth of 0-20 cm were as follows： organic matter of 7.38 g/kg， available N of 66.00 mg/kg， available P of 38.5 mg/kg， available K of 168.0 mg/kg， soil pH of 8.6， soil bulk weight of 1.23 g/cm3 and field capacity of 23.39%.
Test design
Using random block design， there were a total of 46 treatments with 1 variety as a treatment. The test plot was 10 m wide， 13 m long with an area of 130 m2. No repetition was set. With a row spacing of 33 cm， there were 120??104 seedlings/hm2. The preceding crop was tomato. The fertilization amounts of N， P and K were unified of 45 kg/hm2， 105 kg/hm2 and 45 kg/hm2， and all were applied as the base fertilizer before sowing. Among them， urea （N of 46%） was used as the nitrogen fertilizer， triple superphospate （P2O5 of 50%） as phosphate fertilizer， and powdered potassium sulfate as the potash fertilizer （K2O 50%）.
Statistical analysis of data
The basic data were processed by Excel 2010 to get the mean， maximum， minimum， standard deviation （SD） and coefficient of variation （CV） of characters. The differences in the characters of different varieties were expressed by coefficients of variation， and the genetic diversity index was calculated using Shannon??Weaver Information Index. The formula was H??=-??PilnPi， where Pi is the probability that the ith level of a character appears. The quantitative characters were graded and the qualitative characters were valued for quantitative and statistical analysis. The grading method for calculating the diversity index was as follows： first calculate the overall average （X） and standard deviation （d） of the test materials， and then divide them into 10 grades， 0.5 d a grade from the 1st grade of[Xi??（X-2d）] to the 10th grade of [Xi??（X+2d）]. The relative frequency of each grade was used to calculate the diversity index. The diversity index was calculated using the formula of H??=-??PilnPi， where Pi is the percentage of resource number of a character in the ith grade to the total number of resources， and ln is the natural logarithm. DPS 7.50 data processing system was used to carry out the correlation analysis and principal component analysis of 10 characters such as plant height， tiller number and 1 000??grain weight of dryland proso millet varieties， and SAS 8.2 was used to make the cluster analysis of 46 dryland proso millet germplasm resources. Results and analysis
Analysis on the diversity of 3 morphological characteristics of dryland proso millet resources
As shown in Table 2， under unified cultivation， the 46 proso millet varieties all showed differences in morphological characteristics and group structures. The inflorescence colors of proso millet were mainly purple and green， in which there were 38 varieties with green inflorescence， accounting for 82.6% and 8 varieties with purple color， accounting for 17.4%. There were 4 spike types， in which there were 5 varieties each with spadix and loose panicle types， accounting for 10.9%， respectively， 2 varieties with compact panicles， accounting for 4.3%， and 34 varieties with lateral panicles， which was the most accounting for 73.9%. There were a total of 5 grain colors， namely black， bicolor， white， red and yellow， respectively 3， 2， 6， 17 and 18 varieties， and the proportion for each color w was 6.5%， 4.3%， 13.0%， 37.0% and 39.1%， respectively， in which red and yellow resources were the most.
Analysis on the diversity of agronomic traits of dryland proso millet resources
As shown in Table 3， the 8 agronomic traits of the 46 resources showed great phenotypic genetic variation， and the coefficients of variation were in the range of 5.38-19.23%. The variation coefficients of grain weight per spike and spike weight were large， respectively 19.23%， 17.93%. The coefficients of variation of plant height and main spike length were relatively small， 7.73% and 7.33%， respectively. The average diversity index was 1.94， and the diversity indices of yield and spike stalk length were relatively larger， 2.03 and 2.02， respectively， and the diversity index of growth period was the smallest of 1.77. Based on the statistical parameters and diversity indices of 8 agronomic traits， the variation degrees of spike stalk length and yield were great， while the variation degrees of plant height and main spike length were relatively smaller. The test materials had various performances in yield per plant， while the types were relatively fewer in plant height， growth period and spike weight.
Principal component analysis of main agronomic traits of dryland proso millet germplasm resources
As shown in Table 4， different trait indices were divided into different principal components according to the absolute values of the vectors， and the position of the largest absolute value of the same index among the factors was the principal component of the index. As shown in Table 4， among the principal components， the main information was concentrated in the first 4 principal components with the cumulative contribution rate reaching up to 83.34%. The eigenvalue of the first principal component was 2.99， and the contribution rate was 37.34%. The eigenvalue of the second principal component was 1.53， and the contribution rate was 19.13%. The eigenvalue of the third principal component was 1.22， and the contribution rate was 15.19%. The eigenvalue of the fourth principal component was 0.93， and the contribution rate was 11.68%. As shown in Table 5， the contribution rate of spike stalk length was the largest in the first principal component， with the eigenvalue of 0.419 1， indicating that the germplasm resources with high first principal component value had relatively longer spike stalk. Yield had the largest contribution rate in the second principal component， while the contribution rate of spike weight was a negative value with large absolute value. Therefore， it needed to select the varieties with large spike weight when using the germplasm resources with high scores of the second principal components. The largest contribution rate in the third principal component came to the main spike length， which had the eigenvalue of 0.713 7， indicating that the germplasm resource with high scores of the third principal component had long main spikes. The contribution rate of grain weight per spike was the largest in the fourth principal component with the eigenvalue of 0.628 8， indicating the germplasm resource with high scores of the fourth principal component had large grain weight per spike.
Cluster analysis of dryland proso millet germplasm resources
As shown in Table 6 and Fig. 1， the 46 dryland proso millet germplasm resources were clustered into 3 categories at an Euclidean distance of 0.984 4. The first category contained 41 resources， which had the longest growth period， the highest plant height， the largest spike stalk length， the heaviest spike weight and moderate performances in other agronomic traits like yield. The second category contained 2 resources， which had the highest yield of 4 086.00 kg/hm2， the longest main spike length of 26.9 cm， largest grain weight per spike and 1 000 grain weight of 5.17 g and 8.25 g， respectively， and moderate performances in other agronomic traits. The third category had 3 resources， which had the lowest performances in the 7 agronomic traits except moderate performances in spike stalk length.
Conclusion and Discussion
Germplasm resources are the basis of crop breeding. To cultivate new crop varieties high quality， high yield and high resistance， it is necessary to obtain high??quality resources as parent materials. The purpose of this study is to clarify the genetic relationship between the materials and provide some theoretic references for the breeding and industry development of proso millet[11-12]. The morphological diversity index reflects the quantitative distribution of different phenotypic grade of a character， that is， the abundance and uniformity of diversity， while the coefficient of variation reflects the range of variation of a certain character， so the 2 had inconsistent expression in the same character[13]. The results of this study show that 11 agronomic traits of proso millet show different degrees of diversity in different resources. Among the morphological characteristics of the 46 tested materials， green is the major inflorescence color， accounting for 82.6%， the spike type is mainly dominated by lateral panicle type， accounting for 73.9%， and the grain colors are mainly yellow and red， accounting for 39.1% and 37%， respectively. Among the quantitative characters， the variation coefficients of grain weight per spike and spike weight are relatively larger， 19.23% and 17.93%， respectively， indicating that the tested materials have high variation potential in yield and a larger range of choice， which provides greater possibilities for increasing yield. Principal component results show that the agronomic traits of 46 tested materials can be summarized into 4 main components， namely， spike stalk length， spike weight， main spike length and grain weight per spike， and the cumulative contribution rate reaches up to 83.34%. Each principal component can objectively reflect the interrelationship between the various controlled characters. Therefore， in the breeding proso millet， the selection of parental materials should be based on the ordering of the principal components， make specific analysis and comprehensive evaluation to the advantages and disadvantages of each parental material comprehensive index， and make reasonable combination according to the breeding objectives of proso millet， so as to cultivate the ideal new proso millet varieties as soon as possible[9]. The results of cluster analysis show the 46 tested materials are clustered into 3 categories at an Euclidean distance of 0.984 4. Most of the resources are concentrated in the first and third categories. The resources of the first category have the longest growth period， the highest plant height， the largest spike stalk length， the heaviest spike weight and moderate performances in other agronomic traits like yield. Plant height and spike stalk length are important factors for increasing yield， and therefore， if the varieties are combined with drought??resistant and high??yielding parents， they have the potential to cultivate the proso millet varieties with high feeding value. The second category only contains 2 resources， which have high the highest yield， longest main spike length， largest grain weight per spike and 1 000 grain weight and moderate performances in other traits. Main spike length， spike weight and 1 000??grain weight are all important factors for increasing yield. Therefore， these varieties have the potential to cultivate high??yield varieties， and are suitable as combinations of parent materials， or continue to be studied as good lines. The third category consists of 3 resources. In addition to the spike stalk length which falls in moderate performance， the performances of all other 7 agronomic traits are the lowest Therefore， it is necessary to continue to introduce high??quality germplasm resources in future work， and the results of this study can play a guiding role in breeding proso millet. Genetic diversity reflects changes in the genetic information of biological species and can be expressed in many aspects such as molecules， cells and individuals[14]. In this study， the diversity of dryland proso millet is studied from the 11 agronomic traits， and the results show that dryland proso millet germplasm resources have rich genetic diversity. However， the results only suggest that dryland proso millet has rich diversity in agronomic traits， and there are certain limitations. Therefore， in order to reveal the genetic diversity of dryland proso millet more accurately， further study is needed from the cytological level， physiological and biochemical level and molecular level， so as to fully identify the genetic diversity of dryland proso millet.
References
[1] WANG XY. Paninum miliaceum in China[M]. Beijing： China Agricultural Press， 1996： 1-10.
[2] HU XY， LU P， HE JB， et al. Principal components and cluster analysis of agronomic traits of proso millet （Paninum miliaceum）[J].Journal of Plant Genetic Resources， 2008， 9（4）：492-496.
[3] GUO Q. Genetic diversity of Chinese scorpion （Panicum miliaceum L.） germplasm resources[D]. Taigu： Shanxi Agricultural University， 2013.
[4] OJENIYI SO. Effect of zero??tillage and disc ploughing on soil water， soil temperature and growth and yield of maize[J]. Soil & Tillage Research， 1986， （7）： 173-182.
[5] DERPSH R， SIDIRAS N， ROTH CH. Results of studies made from 1977 to 1984 to control ersion by cover crops and no??tillage techniques in Barana， Brazil[J]. Soil & Tillage Research， 1986， （8）： 253-263
[6] TANG QY， FENG MG. Practical statistical analysis and its computer processing platform[M]. Beijing： China Agriculture Press， 1997.
[7] TIAN J， ZHENG DS. Crop genetic resources in China[M]. Beijing： China Agriculture Press， 1994： 312-315.
[8] HU ZA， WANG HX. Basic principles and methods for studying genetic diversity[M]. Beijing： China Science and Technology Press， 1994： 117-122.
[9] QIAO ZJ. Present situation and developing thought of broomcorn millet industry[J].Crops， 2013（5）：25-27.
[10] HU XY， LU P， HE JB， et al. Principal components and cluster analysis of agronomic traits proso millet （Paninum miliaceum）[J].Journal of Plant Genetic Resources， 2008，9（4）492-496.
[11] WANG XY. Description and standard of Paninum miliaceum germplasm resources[M]. Beijing： China Agriculture Press， 2006： 7-24.
[12] LIU F. Genetic diversity analysis on agronomic traits of broomcorn millet[J]. Heilongjiang Agricultural Sciences， 2010， （3）： 15-16.
[13] DDONG KJ， YANG TY， HE JH. Analysis of genetic diversity of agronomic traits of local broomcorn millet[J]. Hebei Agricultural Sciences， 2012，16 （2）： 1-314.
[14] FU XY. Genetic diversity of high quality winter wheat varieties （lines） based on SSR marker[D]. Dingzhou： Hebei Agricultural University， 2006.
Key words Dryland proso millet； Germplasm resource； Diversity
Proso millet， Paninum miliaceum L.， is one of the ancient Chinese food crops. It has the characteristics of early maturity， barren tolerance， heat resistance and drought tolerance. It is often used as the reserve crop prepared against natural disasters in areas with poor natural conditions[1]. The distribution of japonica and glutinous proso millet is relatively regular in China that the proportion of glutinous type gradually decreases while the proportion of japonica type gradually increases from east to west， and such trend is related to precipitation[2]. Due to its short growth period， strong resistances and high nutritional value， proso millet has obvious regional advantages and production advantages in semi??arid areas such as Ningxia， Shanxi and Gansu provinces， especially in the loess hilly region of southern Ningxia， where it is one of the minor grain crops and also a pioneer crop with a large planted area on newly reclaimed wasteland[3-4].
Germplasm resources are the basis for genetic improvement and breeding of proso millet varieties. The improvement of proso millet varieties lags far behind other crops. Under the new situation， in order to achieve a major breakthrough in the innovation of proso millet varieties， in addition to strengthening research on variety resources， it also needs to make breakthroughs in breeding technological research and breeding material innovation， to study effective ways and methods for the heterosis utilization of proso millet to cultivate proso millet hybrids， and the future goal of proso millet breeding is to achieve the combination of yield increase， quality improvement and resistance enhancement[5-6]. The identification of germplasm resources is focused on the much??needed characteristics of breeding， eliminating the differences caused by environmental factors and reflecting the true germplasm characteristics. The selection and improvement of proso millet varieties are carried out on the basis of existing germplasm resources. The replacement of proso millet varieties is inseparable from the utilization of high??quality germplasm resources[7-8]. Therefore， the study of genetic diversity of proso millet is particularly important[9]. However， there are few studies on the diversity of dryland proso millet germplasm resources in the mountainous areas of southern Ningxia. Thus， in order to screen out the high??quality， high??yield and special proso millet varieties suitable for planting in the mountainous areas of southern Ningxia， the genetic diversity of 46 drought??tolerant proso millet germplasm resources were analyzed using principal component analysis and cluster analysis from the growth period， morphological characteristics and yield??related factors of proso millet， with the aim to provide theoretical references for the targeted use of drought??resistant parents to select hybrid combinations， thereby providing theoretical basis for the research on proso millet breeding in Ningxia. Materials and Methods
Test materials
The names and sources of 46 dryland proso millet germplasm resources are shown in Table 1， of which the first 1-26 are japonica varieties， and No. 27-46 are glutinous varieties.
Test area overview
For three consecutive years in 2014-2016， test was carried out at Touying test base of Guyuan Branch of Ningxia Academy of Agriculture and Forestry Sciences， Guyuan City. Located at 36??44??， 106??44?? E， the test base had an elevation of 1 586 m. It is in the temperate arid zone with dry and windy spring and cold winter. It has an annual average temperature of 7.4 ??， accumulated temperature above 10 ?? of 2 500-2 800 ??， frost??free period of 130-150 d， annual precipitation of only 350 mm， which is very scarce. The rainfall is mainly distributed in August， September and October. The farming in the area is arid farming that relies totally on natural rainfall. The area has frequent occurrence of drought disasters， large temperature difference between day and night. In this study， the tested soil belonged to Hunan loess with low soil fertility level， and the main physical and chemical properties of the tested soil at the depth of 0-20 cm were as follows： organic matter of 7.38 g/kg， available N of 66.00 mg/kg， available P of 38.5 mg/kg， available K of 168.0 mg/kg， soil pH of 8.6， soil bulk weight of 1.23 g/cm3 and field capacity of 23.39%.
Test design
Using random block design， there were a total of 46 treatments with 1 variety as a treatment. The test plot was 10 m wide， 13 m long with an area of 130 m2. No repetition was set. With a row spacing of 33 cm， there were 120??104 seedlings/hm2. The preceding crop was tomato. The fertilization amounts of N， P and K were unified of 45 kg/hm2， 105 kg/hm2 and 45 kg/hm2， and all were applied as the base fertilizer before sowing. Among them， urea （N of 46%） was used as the nitrogen fertilizer， triple superphospate （P2O5 of 50%） as phosphate fertilizer， and powdered potassium sulfate as the potash fertilizer （K2O 50%）.
Statistical analysis of data
The basic data were processed by Excel 2010 to get the mean， maximum， minimum， standard deviation （SD） and coefficient of variation （CV） of characters. The differences in the characters of different varieties were expressed by coefficients of variation， and the genetic diversity index was calculated using Shannon??Weaver Information Index. The formula was H??=-??PilnPi， where Pi is the probability that the ith level of a character appears. The quantitative characters were graded and the qualitative characters were valued for quantitative and statistical analysis. The grading method for calculating the diversity index was as follows： first calculate the overall average （X） and standard deviation （d） of the test materials， and then divide them into 10 grades， 0.5 d a grade from the 1st grade of[Xi??（X-2d）] to the 10th grade of [Xi??（X+2d）]. The relative frequency of each grade was used to calculate the diversity index. The diversity index was calculated using the formula of H??=-??PilnPi， where Pi is the percentage of resource number of a character in the ith grade to the total number of resources， and ln is the natural logarithm. DPS 7.50 data processing system was used to carry out the correlation analysis and principal component analysis of 10 characters such as plant height， tiller number and 1 000??grain weight of dryland proso millet varieties， and SAS 8.2 was used to make the cluster analysis of 46 dryland proso millet germplasm resources. Results and analysis
Analysis on the diversity of 3 morphological characteristics of dryland proso millet resources
As shown in Table 2， under unified cultivation， the 46 proso millet varieties all showed differences in morphological characteristics and group structures. The inflorescence colors of proso millet were mainly purple and green， in which there were 38 varieties with green inflorescence， accounting for 82.6% and 8 varieties with purple color， accounting for 17.4%. There were 4 spike types， in which there were 5 varieties each with spadix and loose panicle types， accounting for 10.9%， respectively， 2 varieties with compact panicles， accounting for 4.3%， and 34 varieties with lateral panicles， which was the most accounting for 73.9%. There were a total of 5 grain colors， namely black， bicolor， white， red and yellow， respectively 3， 2， 6， 17 and 18 varieties， and the proportion for each color w was 6.5%， 4.3%， 13.0%， 37.0% and 39.1%， respectively， in which red and yellow resources were the most.
Analysis on the diversity of agronomic traits of dryland proso millet resources
As shown in Table 3， the 8 agronomic traits of the 46 resources showed great phenotypic genetic variation， and the coefficients of variation were in the range of 5.38-19.23%. The variation coefficients of grain weight per spike and spike weight were large， respectively 19.23%， 17.93%. The coefficients of variation of plant height and main spike length were relatively small， 7.73% and 7.33%， respectively. The average diversity index was 1.94， and the diversity indices of yield and spike stalk length were relatively larger， 2.03 and 2.02， respectively， and the diversity index of growth period was the smallest of 1.77. Based on the statistical parameters and diversity indices of 8 agronomic traits， the variation degrees of spike stalk length and yield were great， while the variation degrees of plant height and main spike length were relatively smaller. The test materials had various performances in yield per plant， while the types were relatively fewer in plant height， growth period and spike weight.
Principal component analysis of main agronomic traits of dryland proso millet germplasm resources
As shown in Table 4， different trait indices were divided into different principal components according to the absolute values of the vectors， and the position of the largest absolute value of the same index among the factors was the principal component of the index. As shown in Table 4， among the principal components， the main information was concentrated in the first 4 principal components with the cumulative contribution rate reaching up to 83.34%. The eigenvalue of the first principal component was 2.99， and the contribution rate was 37.34%. The eigenvalue of the second principal component was 1.53， and the contribution rate was 19.13%. The eigenvalue of the third principal component was 1.22， and the contribution rate was 15.19%. The eigenvalue of the fourth principal component was 0.93， and the contribution rate was 11.68%. As shown in Table 5， the contribution rate of spike stalk length was the largest in the first principal component， with the eigenvalue of 0.419 1， indicating that the germplasm resources with high first principal component value had relatively longer spike stalk. Yield had the largest contribution rate in the second principal component， while the contribution rate of spike weight was a negative value with large absolute value. Therefore， it needed to select the varieties with large spike weight when using the germplasm resources with high scores of the second principal components. The largest contribution rate in the third principal component came to the main spike length， which had the eigenvalue of 0.713 7， indicating that the germplasm resource with high scores of the third principal component had long main spikes. The contribution rate of grain weight per spike was the largest in the fourth principal component with the eigenvalue of 0.628 8， indicating the germplasm resource with high scores of the fourth principal component had large grain weight per spike.
Cluster analysis of dryland proso millet germplasm resources
As shown in Table 6 and Fig. 1， the 46 dryland proso millet germplasm resources were clustered into 3 categories at an Euclidean distance of 0.984 4. The first category contained 41 resources， which had the longest growth period， the highest plant height， the largest spike stalk length， the heaviest spike weight and moderate performances in other agronomic traits like yield. The second category contained 2 resources， which had the highest yield of 4 086.00 kg/hm2， the longest main spike length of 26.9 cm， largest grain weight per spike and 1 000 grain weight of 5.17 g and 8.25 g， respectively， and moderate performances in other agronomic traits. The third category had 3 resources， which had the lowest performances in the 7 agronomic traits except moderate performances in spike stalk length.
Conclusion and Discussion
Germplasm resources are the basis of crop breeding. To cultivate new crop varieties high quality， high yield and high resistance， it is necessary to obtain high??quality resources as parent materials. The purpose of this study is to clarify the genetic relationship between the materials and provide some theoretic references for the breeding and industry development of proso millet[11-12]. The morphological diversity index reflects the quantitative distribution of different phenotypic grade of a character， that is， the abundance and uniformity of diversity， while the coefficient of variation reflects the range of variation of a certain character， so the 2 had inconsistent expression in the same character[13]. The results of this study show that 11 agronomic traits of proso millet show different degrees of diversity in different resources. Among the morphological characteristics of the 46 tested materials， green is the major inflorescence color， accounting for 82.6%， the spike type is mainly dominated by lateral panicle type， accounting for 73.9%， and the grain colors are mainly yellow and red， accounting for 39.1% and 37%， respectively. Among the quantitative characters， the variation coefficients of grain weight per spike and spike weight are relatively larger， 19.23% and 17.93%， respectively， indicating that the tested materials have high variation potential in yield and a larger range of choice， which provides greater possibilities for increasing yield. Principal component results show that the agronomic traits of 46 tested materials can be summarized into 4 main components， namely， spike stalk length， spike weight， main spike length and grain weight per spike， and the cumulative contribution rate reaches up to 83.34%. Each principal component can objectively reflect the interrelationship between the various controlled characters. Therefore， in the breeding proso millet， the selection of parental materials should be based on the ordering of the principal components， make specific analysis and comprehensive evaluation to the advantages and disadvantages of each parental material comprehensive index， and make reasonable combination according to the breeding objectives of proso millet， so as to cultivate the ideal new proso millet varieties as soon as possible[9]. The results of cluster analysis show the 46 tested materials are clustered into 3 categories at an Euclidean distance of 0.984 4. Most of the resources are concentrated in the first and third categories. The resources of the first category have the longest growth period， the highest plant height， the largest spike stalk length， the heaviest spike weight and moderate performances in other agronomic traits like yield. Plant height and spike stalk length are important factors for increasing yield， and therefore， if the varieties are combined with drought??resistant and high??yielding parents， they have the potential to cultivate the proso millet varieties with high feeding value. The second category only contains 2 resources， which have high the highest yield， longest main spike length， largest grain weight per spike and 1 000 grain weight and moderate performances in other traits. Main spike length， spike weight and 1 000??grain weight are all important factors for increasing yield. Therefore， these varieties have the potential to cultivate high??yield varieties， and are suitable as combinations of parent materials， or continue to be studied as good lines. The third category consists of 3 resources. In addition to the spike stalk length which falls in moderate performance， the performances of all other 7 agronomic traits are the lowest Therefore， it is necessary to continue to introduce high??quality germplasm resources in future work， and the results of this study can play a guiding role in breeding proso millet. Genetic diversity reflects changes in the genetic information of biological species and can be expressed in many aspects such as molecules， cells and individuals[14]. In this study， the diversity of dryland proso millet is studied from the 11 agronomic traits， and the results show that dryland proso millet germplasm resources have rich genetic diversity. However， the results only suggest that dryland proso millet has rich diversity in agronomic traits， and there are certain limitations. Therefore， in order to reveal the genetic diversity of dryland proso millet more accurately， further study is needed from the cytological level， physiological and biochemical level and molecular level， so as to fully identify the genetic diversity of dryland proso millet.
References
[1] WANG XY. Paninum miliaceum in China[M]. Beijing： China Agricultural Press， 1996： 1-10.
[2] HU XY， LU P， HE JB， et al. Principal components and cluster analysis of agronomic traits of proso millet （Paninum miliaceum）[J].Journal of Plant Genetic Resources， 2008， 9（4）：492-496.
[3] GUO Q. Genetic diversity of Chinese scorpion （Panicum miliaceum L.） germplasm resources[D]. Taigu： Shanxi Agricultural University， 2013.
[4] OJENIYI SO. Effect of zero??tillage and disc ploughing on soil water， soil temperature and growth and yield of maize[J]. Soil & Tillage Research， 1986， （7）： 173-182.
[5] DERPSH R， SIDIRAS N， ROTH CH. Results of studies made from 1977 to 1984 to control ersion by cover crops and no??tillage techniques in Barana， Brazil[J]. Soil & Tillage Research， 1986， （8）： 253-263
[6] TANG QY， FENG MG. Practical statistical analysis and its computer processing platform[M]. Beijing： China Agriculture Press， 1997.
[7] TIAN J， ZHENG DS. Crop genetic resources in China[M]. Beijing： China Agriculture Press， 1994： 312-315.
[8] HU ZA， WANG HX. Basic principles and methods for studying genetic diversity[M]. Beijing： China Science and Technology Press， 1994： 117-122.
[9] QIAO ZJ. Present situation and developing thought of broomcorn millet industry[J].Crops， 2013（5）：25-27.
[10] HU XY， LU P， HE JB， et al. Principal components and cluster analysis of agronomic traits proso millet （Paninum miliaceum）[J].Journal of Plant Genetic Resources， 2008，9（4）492-496.
[11] WANG XY. Description and standard of Paninum miliaceum germplasm resources[M]. Beijing： China Agriculture Press， 2006： 7-24.
[12] LIU F. Genetic diversity analysis on agronomic traits of broomcorn millet[J]. Heilongjiang Agricultural Sciences， 2010， （3）： 15-16.
[13] DDONG KJ， YANG TY， HE JH. Analysis of genetic diversity of agronomic traits of local broomcorn millet[J]. Hebei Agricultural Sciences， 2012，16 （2）： 1-314.
[14] FU XY. Genetic diversity of high quality winter wheat varieties （lines） based on SSR marker[D]. Dingzhou： Hebei Agricultural University， 2006.