The Determine of Desert area Portion in Production of Falling Dust by Discriminate Analysis (case study: Yazd city)

Introduction
Dust haze phenomenon is dust that cover large distance and it originated is of arid and semi-arid area.Environmental effects of dust including dispersion, transport and sediment become is large concerns in early of 1990s. The researches done associated to the frequency of dust days showed that most dust day's frequency is related to central holes of Iran. The main impact of origin sites is created via wind erosion. The Yazd province with more than percent fifty of desert and sand area is located in Yazd – Ardakan plain. Therefore always is exposed to wind erosion and difficult due to it especially dust storms. The critical focuses of wind erosion in Yazd-Ardakan plain is including Sebkha, Kalut & Yardang, Hill, Glacis Epandage Plain and water sediment. In determine of sediment source for the reason that using of traditional methods is very difficult so fingerprinting method pay attention is as appropriate and alternative method based sediment properties. In this method, most important principle is use of chemical, physical and organic properties and Comparethese characteristicswith the samecharacteristicsinsedimentsamples. The method is no many of theproblems oftraditional methods. The main advantages ofthismethod are including high speed, economic and the abilitytoobtaininformationabout the type ofsediment sources andlocation ofsediment sources. Investigation of reference showed that many studies is associated identify of dust source using of fingerprinting   but in country there is any study in this case. The aim of this study is determine of falling dust origin using of fingerprinting in Yazd- Iran.
Material and Method
Study area
Yazd, the largest city in Yazd Province with the latitude as N 31° 53' 50" and longitude as E 54° 22' 3" and population of over 582682 people and approximately within 140 km2. Yazd located in Yazd – Ardakan plain. The climate in this area is arid and semi-arid. Yazd city and Yazd – Ardakan plain are selected for sampling of falling dust and determine of dust origin respectively.
Sampling and Chemical analyses
Falling dust samples were collected from 33 different locations almost covering Yazd city area (roofs of buildings with a height 4 meter were selected for the fixing of the dust collectors). The dust particles were sampled using Marble Dust Collector (MDCO) method for six month from January 2012 to June 2013 (winter and spring seasons). The sampling of falling dust source was including Sebkha, Kalut & Yardang, Hill, Glacis Epandage Plain and water sediment of top soil (5cm) by plot (20*20 cm) with 3-8 repeat in Yazd – Ardakan plain. Then ten heavy metals including Cr, Pb, Cu, Ni, Bi, Zn, Ag, Cd and Se were analyzed by Atomic Absorption Flame Spectrophotometer (Analytic jene-350 model, Germany).
Determine the origin using discriminant analysis method 
The each heavy metal ability was investigation in separation of dust source use of statistical analysis such as One - Way ANOVA and Kruskal-Wallis (P < 0.05) and criteria of strong linear multivariate (Tolerance ≥ 0.1 and VIF≤ 10). Then using of Discriminant analysis was selected the optimal combination of tracers with ability to separation of dust sources.
Determination the contribution of dust sources
In new fingerprinting it is assumed combination of tracer proprieties is linear. Therefore can be wrote combination model for each of tracer specifications according equation (1).
Xi =               (1)
 In the equation (1):
Xi: estimated valueof i tracer (i-=1, 2… m)
aij: mean value of i tracer in j source (j= 1, 2… n)
bj : j source contribution
n: source number, m: number of tracer characterizes
The equation is repeated for each of tracers thus a multivariate mixing model was subsequently used to estimate the relative contribution of the potential sediment sources to a dust sample. For obtain the optimal results in determine of sources contribution can be use optimization methods. In this method, the proportions P contributed by the m individual sources s are established by minimizing the sum of the squares of the residuals (Res) for the n tracer properties involved, where:
)                                         (2)
In the equation (2):
Cssi :the concentration of tracer property i in the dust sample,
Csi : the mean concentration of tracer property i in source group of sediment
bs: the relative proportion from source group of  sediment
The optimal results for sediment sources are achieved by minimum of above equation and repeat, trial and error operation and considerthe following two conditions:
1)      The values of contribution coefficient between zero and one eachofthe deposition sources
 
2)      Total coefficients equal one ofeach of thedepositionsources
 
Evaluation the results of a multivariate mixing model
The measure of the relative error, Coefficient of Performance Model and field observation was utilized in order to investigation of model accuracy.
3)( ME = 1-
TheME valuesiscloser to one indicating of High PerformanceModel.
Result 
Determination the origin by Discriminant analysis method
The results of statistical analysis and criteria of strong linear multivariate was showed all heavy metal except Fe (VIF>10) is usable for Discriminant analysis. The stepwise Discriminant function analysis was employed to select the optimum composite fingerprinting.
The comparison of different mean showed different was significant for Ag and ‌Zn between of groups.
 
Table1. Various steps of import elements to model





Cumulative %


Sig


Wilks Lambda


Element


Step




100.00


0.004**


0.208


Zn


1




100.00


0.000**


0.091


Ag


2





 
** Significantly in the 0.01 level
The results of table 1 showed to added each element was unchanged Cumulative percentage but wilks Lambda was declined  and Significant level was better therefore was increased separation ability between groups. 
The power of detection function is evaluated with results ofthe audit functioncanonical (Table 2).
 
 
 
Table2. Results ofthe audit functioncanonical





Function


Eigenvalues


Percentageof  variance


% cumulative of variance


Canonical correlationcoefficient




1


1.42


100.00


100.00


0.77





 
Table(3) b oflinear regressioncoefficientsoffunctionsis presented.
Table3.Auditfunctionscanonicalcoefficients





Tracer elements


Function




Zn


0.919




Ag


-0.849





In finally Discriminant function was defined according to Canonic Discriminant Function Coefficients (equation 4).
F1 = 0.919 Zn – 0.849 Ag                            (4)
To determine the roleof each of theresourcesfallingdust using theresults of thedetection function is in the function average concentrationof heavy metalsinthe monthwasin the function.The results are showed most likelybelongingtodust is associated to Sebkha in the six months.Therefore most contribution of falling dust of originsuburbanarea is Sebkha in Yazd – Ardakan plain.
The best result was obtained of scenario with two groups including Sebkha - Kalut & Yardang and Hill - Glacis Epandage Plain. Therefore were defined discriminate analysis based on the scenario.
The sources contribution in sediment production 
The according to mixed multivariate model was obtained sources contribution 99.9 and 0.1 percent respectively.  Therefore major contribution of falling dust is related to Sebkha and Kalut & Yardang. The results of minimizing the sum of the squares of the residuals are indicative the best portion for falling dust sources. The results showed portion of groups for production of falling dust are 100 and 0 percent respectively.  These results almost are corresponded with results of mixed multivariate model. The assessments of this model showed percent of the relative error are between 0.0001-3.41 for all samples. The coefficient of performance model variable is between 0.71 – 0.99 for samples. 
Conclusions
Most occurrences of severe sand storms and wind with speeds that is more than 100 km/h are mainly severe in February to June and it events sometimes the black storms and thick clouds of dusts in Yazd Province, so it selected winter and spring seasons for research.
The investigation of low relative error and high coefficient of performance model is indicating the accuracy and performance of model. The results of this model are in agreement with field observation completely. The high sensitive of Sebkha and Kalut against the wind and fine soil in this area are indicating major role this area in production of falling dust. The results of investing wind erosion in faces of Yazd – Ardakan plain is showed Sebkha and Kalut – Yardang among other of faces are the highestshare in production of falling dust because Sebkha are Crustofclay–salt therefore due tohighsalinity andsodiumishighly sensitivetoerosion and The soilof thislandisa sensitive andhighly susceptible to erosion. The Neogene hills are thehigherresistance againstwind erosion because they cover ispebblesand rubble.
The researchin case of wind erosion in Yazd – Ardakan plain showed area involving Sebkha and Kalut despite the slight area than other area is highest proportion in wind erosion and production of dust.