The Study Of Identification Method On Weak Geochemical Anomaly Of Stream Sediment

The 1:200000 stream sediment geochemical survey since 1970 s in our country has been carried out in nearly all the mountains and mid-levels.The 1:200000 stream sediment geochemical survey is inexpensive and efficient,which plays a key role to identify prospecting areas(belt),ore field and medium and large deposits,especially nonferrous and precious metal deposits.With the deepening of mineral exploration in our country,the amount of outcrop and shallow ores reduces gradually.Therefore the focus of the prospecting gradually turns to the concealed deposits,half-concealed deposits and ores that are difficult to be identified.The huge amounts of high-quality 1:200000 stream sediment geochemical data in our country will still play an important role in geological prospecting in the future.Although almost all the "high,big and complete" geochemical anomalies of 1:200000 stream sediment have been assessed,the data still contain a lot of weak geochemical anomaly information with high ore-prospecting value which needs to be further mined.Many researches show that the size and strength of mineralized anomalies are controlled by many factors,so the big mine does not necessarily shows a big anomaly and vice versa.In recent years a series of ore deposits are found in areas with weak anomalies in China,which proves the prospecting value of weak geochemical anomaly.At present,there is no perfect identification method of weak geochemical anomalies of stream sediment,so it is of practical significance to explore identification theory and method of weak geochemical anomaly.The 1:200000 stream sediment data of Hunan province are taken as emphasis,and geochemical data of other landscapes are taken as references.This research studies the geochemical characteristics of stream sediments in Hunan province,the influencing factors of stream sediment geochemical background and the eliminating method of influencing factors.And then we put forward anomaly identification methods based on normal distribution and multivariate non-normal population,respectively.The chemical compositions of stream sediments have obvious spatial differences in Hunan province.The study area can be divided into six different geochemical areas through the k-means clustering method.And the element background values in these geochemical areas have a certain difference.The study finds that the element geochemical background values of stream sediments are influenced by many factors such as dimensions,lithology background,sample particle size and the landscape,the bedrock lithology is an especially significant factor.In the area with complex lithologies,stream sediments come from multiple "parents",so probability distributions of elements usually show the mixed distribution and they often show multiple peaks on probability distribution figures.Geological background(especially the background of lithologic)influences the element background values,which may lead the valuable weak geochemical anomaly be covered up,at the same time may lead the geological body with high background regarded as anomaly mistakenly.In order to eliminate the influence of lithology background,we propose a geochemical background influencing factor eliminating method based on k-means clustering and robust regression.According to the rock-forming elements that can characterize lithologic characteristics in stream sediments,divide the samples into several subclasses by k-means clustering algorithm,and then check the correlation relationships of elements and main oxides(Si O2,Fe2O3)in each subclass.For the element that has no obvious correlation with the main oxides,first of all,transform the element data to make them normal,calculate mean and standard deviation after eliminating outliers,and then standardize the data through Z-scores.For the element that is significantly related with the main oxides,establish the regression equation between element and the main oxide,solve the undetermined coefficients using robust regression method,calculate regression value and standard residual error,and then standardize the data.Finally we can obtain the standardized normal data.The 1:200000 stream sediment data of Guiyang and Liuyang in Hunan Province are taken as examples.It is found that the obtained data are close to standard normal distribution with mean of 0 and the standard deviation of 1,which indicates that the influence of the geological background is eliminated to a large degree.The standardized data highlight some weak geochemical anomaly and suppresses the high background,which makes a better agreement of the anomalies and the known deposits.For the factors such as sample particle size,system errors and the landscape,the standardization method in individual map can be used.We discuss the weakness of traditional anomaly contour map and propose combination anomaly color-block mapping method based on normal standardized data.The main ore-forming elements and associated elements are used,the sampling point is expressed by a color-block,each color-block is divided into 9 small grids,and each grid represents an element,and then adopt corresponding multiples to represent anomaly intensity.Research shows that the method can convey more information about anomaly strength and erosion degree of ore deposit,it can also improve the accuracy of anomaly delineating.The method fully considers the hydrothermal deposit vertical zoning regularity and different mineral anomaly element combination characteristics in the study area.Compared with the traditional method,this method can convey a large amount of information,obtain more accurate anomaly location,and it is easy to be used,etc.We improve traditional Mahalanobis background and anomaly division method on the basis of the standardized data,and put forward a background and anomaly division method based on the multivariate normal population.And we also put forward the anomaly mapping method based on Mahalanobis distance.The map is simple and intuitive,it can show anomaly element combination information of each point.Through the case study,it is found that the anomalies identified by this method coincide better with the known ore deposits and the weak anomalies are highlighted.Using multidimensional data information and characterizing anomalies from the perspective of multiple elements are suitable for weak anomaly recognition and can improve the prediction accuracy of anomalies.Study shows that the contents of some elements in stream sediments,even eliminated the influence of lithology background,their standardized data can still not meet or approximate normal distribution.In this case,the geochemical anomaly identifying method based on the theory of normal distribution is inappropriate.Therefore,the projection pursuit method based on real-coding accelerating genetic algorithm which is suitable for high-dimensional non-normal data is put forward.The high-dimensional data are projected to one dimension and then the geochemical anomalies can be identified according to the projected values.This method not only solves normal distribution limitations of the above method,it can also make full use of the information of depleted elements in the process of mineralization.Combined with the whole dispersion degree and partial condensation degree,the method can make the abnormal distribution high-dimensional data projected to one dimension space,make full use of effective information in the data and avoid the interference of other information.The case study shows that the projection pursuit method based on the background-eliminated standard data can identify the weak anomaly information that traditional methods cannot.The identification of weak stream sediment geochemical anomaly is the key problem in the process of geochemical data processing.The proposed method in this research has a certain innovation and supplement for background elimination method.The multivariate geochemical anomaly identification method based on the standardized data which eliminated the background can effectively identify weak anomaly and suppress high background.Geochemical anomaly identification method integrated with multivariate data conforms the development trend of information integration.The method proposed in this study is of certain theoretical and practical significance for weak geochemical anomaly identification.