# BASIC LINEAR GEOSTATISTICS PDF

A model obtained via kriging provides a false idea of the variability of the true grades as the estimates were calculated minimising the error variance, and therefore smoothing the grades (Armstrong, ). A comprehensive presentation on kriging is presented by Matheron ( [Dr. Margaret Armstrong] Basic Linear Geostatistics - Ebook download as PDF File .pdf), Text File .txt) or read book online. Basic Linear Geostatistics. Linear Geostatistics covers basic geostatistics from the underlying statistical assumptions, the variogram calculation and modelling Download book PDF.

Author: | LEANA SEEMER |

Language: | English, Portuguese, Arabic |

Country: | Moldova |

Genre: | Religion |

Pages: | 418 |

Published (Last): | 06.06.2016 |

ISBN: | 534-4-25566-588-9 |

ePub File Size: | 28.63 MB |

PDF File Size: | 19.77 MB |

Distribution: | Free* [*Registration needed] |

Downloads: | 48200 |

Uploaded by: | CHERI |

Linear Geostatistics covers basic geostatistics from the underlying statistical DRM-free; Included format: PDF; ebooks can be used on all reading devices. Download and Read Free Online Basic Linear Geostatistics Margaret Armstrong Basic Linear Geostatistics by Margaret Armstrong Free PDF d0wnl0ad, audio. Basic Linear Geostatistics by Margaret Armstrong, , available at Book Depository with free delivery worldwide.

Collect and check data Variance of a point within a volume Ordinary kriging ofa block Ordinary kriging Kriging the value of the mean Dispersion versus block size An iron ore deposit Screen effect Testing the quality of a kriging configuration Symmetry in the equations For the lkm grid. Iron ore deposit. For the sOOm grid.. How the choice of the variogram model affects kriging Similar looking variograms Block kriging using a large neighbourhood..

Point kriging using a large neighbourhood. What is causing the ugly concentration oflines? Relationship to the dispersion variance. What size blocks can be kriged?

Grid size for kriging. The effect of the choice of the nugget effect Negative weights.. When the limits of the orebody are not known a priori. Direct composition of tenns.. Can kriging be used to estimate global reserves?

Area known to be mineralized Kriging small blocks from a sparse grid. Adding extra samples improves the quality of the estimate.. Point kriging using smaller neighbourhoods Extension variance.. Composition by line and slice tenns.. Optimal sampling grids.

How to eliminate these concentrations of contour lines AU Means and variances Al What maths skills are required in linear geostatistics Table of Contents xi Appendix 1: Review of Basic Maths Concepts.

Due DUigence and its Implications Lastly some case studies comparing geostatistics with other estimation methods are reviewed. Its application to rhe petroleum industry is more recent.

Danie Krige. After reading an early paper written by Krige. It illustrates rhe need for good estimators. This estimation procedure is called "kriging" after the South African engineer. Its use has been extended to other fields such as environmental science.

Georges Matheron. The basic tool in geostatistics. Once a mathematical function has been fitted to the experimental variogram. Before going into detail about the variogram and rhe different types of kriging.

The economic impact of the support and information effects on reserve calculations is stressed. TIlis is one of the advantages of geostatistics over traditional mefuods of assessing reserves. So once fue variogram has been selected for a particular deposit or region. Geostatistics can help tile mine planner get accurat. So it is important to know how serious fuis error is. Here a block might represent fue production for a shift. In addition to estimating fue ore tonnage and the average grade of mining blOCKS.

For iron ore. For coal these include ash content. As well as giving fue estimated values. TI1is makes it possible to evaluate fue estimation variance for a wide variety of possible sample patterns wifuout actually doing fue drilling. L and calorific value. This means that if a numerical model of a deposit is being set up to test various proposed mine plans. Over the past 25 years the petroleum industry has been turning more and more to kriging for this.

This leads in to nonlinear geostatistics. When the sample grid is about the same size as the selection blocks their grades can be estimated individually with reasonable accuracy. As this text deals only with linear geostatistics. More recently environmental scientists have also started using geostatistics. In the subsequent chapters we go on to see what the variogram is and how kriging is used to estimate values and to obtain the estimation variance.

After this. But if the blocks are much smai1er than the grid size as is usually the case at the feasibility stage. Similar problems arise in soil rehabilitation work where scientists have to predict the total amount of material that is contaminated.

Having seen some of the possible applications of geostatistics in the rnmmg industry.

In this case a conditional simulation of the deposit should be used. This has the advantage of being more accurate than other methods of evaluating grid node values. The best that can be done is to predict the proportion of selection units that will be recovered.

More information on when to use simulations rather than kriging is given in Chapter 9. Introduction 3 1. These are simply not accurate enough. Figure 1. First let us see the data. Then several comparative case studies on ore evaluation techniques are reviewed. The values of the other 48 samples will be used later for comparison purposes. First we present a simplified example showing the financial impact of poor block estimates. Now it is your turn to design a way of estimating each of Ihese 16 values.

The simplest way of estimating each of the 16 block grades is by equating the grade of the sample in each one to the block estimate.. The grades of 64 blocks of size I x 1 were available in an area 8 x 8. You may choose whatever method you like.. This exercise has been designed to highlight the economic impact of estimation errors. Write your estimates in Ihe space provided on the right of Fig.. This is called the polygonal method. Many people wonder whether kriging really does give better results than other methods.

Sixteen samples of size 1 x 1 to be used to estimate mining blocks of size 2x2 These 16 values will be used as the "samples" to estimate the values of mining blocks of size 2 x 2 i. So mining a block with a grade of leads to a profit of 1 unit. Kriged estimates of block grades A third set of estimates was obtained by having a geostatistician krige the block values Fig.

For the present. Introduction 45 95 75 20 32 20 5 b 8 Fig 1. Kriging is just a special sort of weighted moving average. You are not expected to understand how these numbers were obtained yet.

Suppose that in this case the economic cutoff is For the time being we are going to ignore any geometric constraints due to the mining method. For the polygonal method.

Their real grades are The company could well end up in serious financial difficulties. Repeat these calculations for kriging. Clearly only three blocks would have been selected The true grades of the 2 x 2 blocks are given in Fig.

Note the difference between true and estimated grades 1. Shaded blocks with a grade above are scheduled for mining.

Lastly repeat the calculation for your own estimator and note the results. So the actual profit would be: Show that only two blocks are scheduled for mining and that the actual profit is compared to a predicted profit of There are five of them.

See Fig. Compared to this. We shall work through this together for the polygonal estimation and then you can repeat it for the other two. So the expected profit is: These ore blocks are correctly estimated as ore. What we actually wanted was the blocks whose true grade is above The cloud of points has been represented as an ellipse.

To visualize this. Introduction 7 illusory. A horizontal line drawn at Y Crossplot of true grade versus estimated grade. We will see that the problems met when estimating blocks are due to two effects: The blocks to the right of this line are selected for mining. They form a cloud of points which has been represented here as an ellipse.

Unfortunately they do not. To show this graphically. The blocks above this line should have been mined. This divides the whole area into four zones: Blocks with an estimated grade above are scheduled for mining whereas those blocks actually above should be mined When selecting blocks for mining. Now it is interesting to see why kriging works better.

## Browse more videos

They correspond to the upper right part of the diagram. Ideally the estimated grade would be equal to the true one. We have only estimates for the block grades instead of the real ones. Kriging gave a much more realistic prediction of compared to an actual profit of Le. For kriging the slope of the regression is approximately 1. The reader can go back to the two scatter diagrams and see the misallocated blocks for each of the estimation methods in the upper left and lower right quadrants.

In Chapter 8 we shall see that the criteria for judging an estimator include the slope of the regression line of the!. Now look at the "fatness" of the two clouds. This confirms that kriging is better. These blocks lie in the upper left part of the diagram. Here the samples have aim x 1m support while blocks are 2m x 2m. Although the two means are the same. These ore blocks have been considered to be waste. In general. The true grades of the sixteen 2m x 2m blocks and of the sixty-four 1m x 1m blocks are shown in Figs.

They lie in the lower left part of the diagram.. This second type of estimation error does not cancel out the preceding one and can have expensive consequences for the mine. Going back to our example. These waste blocks are correctly estimated as waste.

Kriging effectively gives a "thinner" cloud. These blocks are in the lower right part of the diagram. These waste blocks have been considered to be ore. Histograms of the grades. For the cutoff more ore will be recovered if 1m x 1m blocks are mined rather than 2m x 2m ones. The true grades of the sixty-four 1m x 1m blocks lbeir histograms Fig. As the polygonal method equates the grades of the samples Le. Although the means remain the same the variances are different and so are the shapes.

Introduction 45 20 30 70 33 95 80 40 35 13 ] 75 95 20 35 32 95 20 35 53 2 45 58 90 9 Fig 1. Both were favorably impressed by the results given by kriging.

Four estimation methods were considered: The tests were carried out in two parts of the mine one with sets of data values. The frrst part of this course will deal with the variogram. As this is more complicated than ordinary kriging.

The aim of his study was to predict the seam width at a distance of 18m one pillar in advance of the workings in the No 2 seam at the Witbank Mine. We can see that the way in which we combine data in the neighbourhood of the block to be estimated is important.

The differences between the estimated values and the actual production figures were calculated for all four methods Table 1. We will first look at two case-studies on coal. He then compared his estimates with the actual production figures. In the second part of the book the variogram is used to calculate the weights to be used when estimating blocks for example.

To answer this question we will have a look at some case-studies where the predictions made using geostatistics were compared with actual production figures. This showed that the kriged estimates were consistently closer to the true values than the other methods considered. Wood did not run into the problem of a trend in the data in his study of South African coal.

It is important to note that this deposit had a marked trend in the sulphur values. The average relative error between the estimates and the actual value was 9. Sabourin estimated the sulpbur content of blocks using channel samples. We now know some of the properties that a good estimator should have. He was therefore able to use ordinary kriging.

If the regression line is not at this angle. As the data were on a very close grid 25 ft. Rendu then took the central one of the 25 grades as the "sample" and estimated the block grades by kriging using the "samples". Lognormal kriging with a known mean came closer to this than any of the other methods.

Two interesting ones are Rendu and Krige and Magri When that occurs. Ideally the regression line should be at 45 degrees. Introduction 11 Table 1. Since the data had a three parameter lognormal distribution. Rendu set out to test whether geostatistical predictions were verified in practice.

He had about gold grades from one section of the Hartebeestfontein Mine. Mean square of the standardised estimation errors. To present his results he calculated the regression of the true grade against the estimated one on a bi-Iogarithmic scale for all the estimation methods considered. By moving the center of the "sample" grid. In the other paper the authors described how they used both DDH data and blasthole data when kriging blocks.

Two particularly interesting ones were those by Raymond and Armstrong who worked on a porphyry copper deposit and by Blackwell and Johnston who studied a low grade copper molybdenum deposit. The mineral reserves results are easily duplicated by different mine personnel. Raymond and Armstrong found a very close agreement between the grade of milled ore over a 17 month period.

The improved mineral reserve permits better long and short range planning and allows the operator flexibility when dealing with downtime. In their conclusion they cited three advantages of using geostatistical methods: Since these comparative studies confirm the superiority of kriging over other commonly used estimation methods for deposits ranging from coal to gold.

Calculate the average for the 16 block grades for each method and for the true grades. The second comparative study by Krige and Magri was on the gold grades of a very variable reef in the Lorraine gold mine and on the lead grades in the Prieska copper-zinc mine. Geostatistics effectively improves estimated grades. Their findings confirmed those by Rendu. Which of these estimators are unbiased? Plot the scatter diagrams of the true grade on the vertical axis against the estimated grade.

They used lognormal kriging which is a special form of kriging designed for skew data with a lognormal distribution. As we have seen. We shall start by seeing how to use geostatistics to model these types of variables. Grades estimated by three different methods a.

Introduction 13 Table 1. One of the most intuitively appealing ways of developing a mathematical model is by modeHing the genesis of the phenomenon. These problems led researchers to give up this approach at the time. The underlying hypotheses second order stationarity and the weaker intrinsic hypothesis are introduced.

Not surprisingly it proved difficult to get meaningful estimates of these from limited sample data. As sedimentary processes are amongst the simplest to describe. Summary In tills chapter the basic definitions in geostatistics including the concepts of random function and regionalized variable are presented. There are many ways of setting up models. The relationsillp between the variogram and the spatial covariance is derived but the rest of the variogram properties are left to the next chapter.

Several will be discussed. Some of the basic properties of the spatial covariance are introduced in tills chapter as they are helpful in deciding on the degree of stationarity. The variogram and the spatial covariance are defined. The problem of how to decide whether to treat a variable as stationary.

Recent work by Hu. Genetic models. The flfst task in a geostatistical study is to identify these structures. The geology of reservoirs and deposits is too complicated and not yet well enough known for this approach to work. At points where no measurements have been made. The common statistical models including trend surfaces put all the randomness into the error term while all the structure is put into the deterministic term. InSisting on having uncorrelated errors means that the function has to twist and tum a lot.

Trend surfaces. The equation is very complicated and contains many terms like sines. Its mean is called the drift. The geostatistician can go on to estimate or simulate the variables. This suggests that it might be better to allow for correlations between values different distances apart. A better way of representing the reality is to introduce randomness in terms of fluctuations around a fixed surface which Matheron called the "drift" to avoid any confusion with the term "trend".

They can also be thought of as being the outcomes or realizations of the corresponding random variable Z x. The implicit assumption underlying these types of regression methods is that the surface under study can be represented.

By the late sixties. The difficulty with this approach can be seen from Table 2. The problem is that most geological variables display a considerable amount of short scale variation in addition to the large scale trends that can reasonably be described by deterministic functions. So at the same time that lacod and loathon were working on reservoir genesis.

Unfortunately this is not realistic for geological phenomena. This is the basic idea behind geostatistics. Fluctuations are not "errors" but rather fully fledged features of the phenomenon. Here "random" means that the error is uncorrelated from one place to another and does not depend on the function. The term regionalized variable was coined by Matheron Regionalized Variables 17 Table 2.

It would be impossible to do anything with this model unless we are prepared to make some assumptions about the characteristics of these distributions. A random function bears the same relation to one of its realizations as a random variable does to one of its outcomes. SU ]2. The next section. A random function is characterized by its finite dimensional distributions.

**READ ALSO:**

*WILLIAM DALRYMPLE PDF*

Metal grades. In hydrology. BOX No 1: Variables that can be modeUed by random functions.

## Non-Linear Geostatistics for Reservoir Modelling

Porosity and permeability. In fishery science. Tree density in tropical forests. In the same way. In other words. Before going into these in detail. For soil science. Box No I1ists some of the more common ones.. Quality parameters e. This is called ''weak'' or second order stationarity. Topographic variables such as seam thickness.

In its strictest sense stationarity requires all the moments to be invariant under translation. Rock type indicators e. For any increment h. Geochemical trace element concentrations in soil samples and stream sediments. Another branch of geostatistics has been developed to handle "nonstationary" regionalized variables. Z x exist and are independent of the point x. Most estimators used in the earth sciences are linear combinations i.

**You might also like:**

*BASICS NEUROANATOMIE PDF*

Interested readers could consult Matheron or Delfiner This is why Matheron In practice. For the moment we shall only consider cases where the mean is constant. This is true for the inverse distance method. By using intrinsic regionalized variables instead of just stationary ones.

We will show later that the variance of linear combinations can be calculated only if the sum of the weights is O. It assumes that the increments of the function are weakly stationary: That is. Regionalized variables that are stationary always satisfy the intrinsic hypothesis but the converse is not necessarily true. It is outside the scope of this text. In contrast to the stationary case.

Later in this chapter we will see that if a regionalized variable is stationary. On both theoretical and practical grounds it is convenient to be able to weaken this hypotheSis. A particularly startling practical example of this was described by Krige for the gold grades in South Africa. It is the basic tool for the structural interpretation of phenomena as well as for estimation.

Clearly when there is a marked trend the mean value cannot be assumed to be constant. Regionalized Variables 19 That is. This assumption of quasi-stationarity is really a compromise between the scale of homogeneity of the phenomenon and that of the sampling density. With samples at these distances. This means that at this scale the sulphur content could be considered as a locally stationary or. This limit could be the extent of a homogeneous wne within a deposit or the diameter of the neighbourhood used in kriging i.

There is plenty much closer. In practical situations the variogram is only used up to a certain distance. Diagrammatic representation of sulphur grades and a blow-up of the central section. Over the whole 8km length. The problem is to decide whether we can find a series of moving neighbourhoods within which the expected value and the variogram can be considered to be constant and where there are enough data to give meaningful estimates. In practice the blocks of coal to be estimated are about m x m for underground mines and m x 60m in strip mining operations.

But over shorter sections it can be considered as being locally stationary because the fluctuations dominate the trend Consider the sulphur content of coal along a transect Fig Samples are generally on a m x m grid for wide spaced holes.

However looking at a blow-up of the central section. Over the total distance shown 8 kIn there is a clear increase from left to right This can best be seen from an example. Regionalized Variables 21 Three important properties are listed below.

C h Proof.. Figure 2. This shows that the corresponding covariance is obtained by "turning the variograrn upside down". The proof starts out from the detinition of the variograrn: Proofs are given in Box No 2. Whereas the variogram starts from zero and rises up to a limit. The covariance is. For stationary variables. First property. Proofs of the properties of the covariance. It can be shown mathematically that variograms with an upper bound come from stationary regionalized variables.

It would be more accurate to say that only stationary regionalized variables have bounded variograms. Its spatial covariance is denoted by C h. One of the key steps in geostatistics is expressing the variance of a linear combination a weighted average in terms of the weights and the covariance function.

Regionalized Variables 23 Clearly this is possible only when the variogram is bounded above. Express its variance in terms of the weighting factors and its covariance C h. The first exercise develops the basic formula. This result means that if the variogram rises more rapidly than a quadratic for large h.

Interested readers can consult Matheron Otherwise it can be considered to be stationary or intrinsic. Show that its variance can be written in either of the following ways: This is helpful in deciding whether a variable is stationary or intrinsic or whether it has to be treated as nonstationary.

When the angle is changed. For a fixed angle. The common variogram models are presented. The formula for calculating the variance of a linear combination of regionalized variables in terms of the variogram is proved. For example. Z x is zero.

Figure 3. Images of variables having some of these variograms have been simulated to highlight the differences between the models. The reason why only positive definite functions can be used as models for the variogam is stressed. In this one. A typical variogram which reaches a limit called its sill at a distance called the range It presents the following features: This merely reflects the anisotropy of the phenomenon. What is more.

It could be discontinuous just after the origin. This occurs when there are several nested structures acting at different distance scales. The range need not be the same in all directions.. This is one fundamental difference between the variogram and the covariance. The properties of the variogram will now be treated in detail.

Alternatively it could just go on rising. After the variogram has reached its limiting value its sill there is no longer any correlation between samples. This critical distance. The latter only exists for stationary variables and is bounded. Examples of anisotropy and nested structures will be given later. Four types of behaviour near the origin are shown in Fig.

This means that the variable is highly irregular at short distances. The regionalized variable is then continuous but not differentiable. This indicates that the regionalized variable is highly continuous.

The term "nugget effect" is also applied to short range variability even when it is known to be due to some other factor such as micro-structure. A quadratic shape can also be associated with the presence ofa drift. Pure randomness or white noise. The grade passes abruptly from zero outside the nugget to a high value inside it.

**Other books:**

*IMPULSE ELLEN HOPKINS PDF*

In fact it is differentiable. Particles of pyrite randomly distributed in coal lead to erratic changes in its sulphur content. Gold is not the only substance that contains nuggets. Discontinuous at the origin i. Bounded and unbounded variograms 3. It is. This is the limiting case of a total lack of structure. But it is even more important to study its behaviour for small values of h because this is related to the continuity and the spatial regularity of the variable.

It is called a nugget effect because it was ftrst noticed in gold deposits in South Africa where it is associated with the presence of nuggets of gold. The variograms of most geological variables.

## download for others

If this does not occur. Two different types of anisotropy can be distinguished: The quadratidc shape a indicates a high degree of continuity in the variable. Discontinuities at the origin c. On the left. The Variogram 29 y h direction 1 h Fig 3. If the curve is an ellipse in 2-D. In these cases a simple change of coordinates transforms the ellipse into a circle and eliminates the anisotropy. Ellipses showing the major and minor axes in the case of geometric anisotropy This transformation is particularly simple when the major axis of the ellipse coincides with the coordinate axes as is shown in Fig.

Then if the equation of the variogram in direction 1 is Yl h. Elliptic or geometric anisotropy We can draw a diagram showing the range or the slope as a function of direction. Isotropic component Vertical component: In such cases the sills of the variograms are not the same in all directions.

To be more specific. It is standard practice to split the variogram into two components. If the variogram was calculated in only two perpendicular directions it would be possible to miss the anisotropy completely. The experimental variogram. This indicates the presence of a drift Fig. Variogram shape in presence of a drift However. Nested structure composed of a shon range structure and a longer one Nested structures indicate the presence of processes operating at different scales.

The variograms for different zones have the same shape but the sill in rich zones is much higher than in poor ones. This often occurs with lognonnally distributed data. As the sill often. At the petrographic scale Le. The Variogram 31 provides an estimate of O. If they are not. These two coincide only if the increments have a zero mean.

This change is obvious when the two ranges are quite different. At the level of strata or mineralized lenses i. The shoner range can be distinguished by the characteristic change in the curvature. In Fig. An example of this is presented in Chapter 5. But it is more common to find that periodicity is an artefact due to human activity rather than Mother Nature.

It is important to check that the effect is real and not merely an artefact. This regular. The reason for this is that variograms have to satisfy certain conditions.

Folded strata could exhibit periodicity. The calcite crystals tended to be separated by intervals roughly proportional to their size. Otherwise there is always a risk of finishing up with a negative variance which would be totally unacceptable. As this "hump" in the variogram corresponds to a hole in the covariance. One case where periodicity can occur.

First we consider a stationary variable Z x with covariance C h. But more generally bumps of this type are due to natural fluctuations in the variogram or to statistical fluctuations because too few pairs of points were used in calculating the experimental variogram.

The variograms calculated perpendicular to the ridges and valleys can show the periodicity but those parallel to the ridges do not. In some cases. Whereas it is natural to find periodic phenomena when dealing with time series. One such example is the variogram obtained by Serra from thin sections of iron ore from the Lorraine region in France.

Sometimes this shape can be explained geologically. We can only be sure that this exists for linear combinations of increments. By detinition its variance [ As the covariance need not exist for intrinsic random functions. The situation is slightly different when the variable is intrinsic but not stationary. A function C h satisfying this condition is said to be positive detinite. Conversely any combination satisfying this condition can be written as a linear combination of increments..

Box 3 gives this proof and the formula for its variance in terms of the weights and the variogram model. The Variogram 33 [ In this case the variance of an arbitrary linear combination need not exist. Combinations are said to be "admissible" if the sum of the weights is zero.

Calculating the variance of admissible linear combinations. Firstly we want to show that any linear combination whose weights sum to 0. By choosing an arbitrary point as origin. Aj [y x.

Var[Z xi. Var[Z x. Var I A.. The range of admissible variogram models is more restricted for the stationary hypothesis but any weighting factors may be used. Consequently the class of admissible variogram models is richer than for covariances.

This condition is weaker than the preceding one for covariances which had to hold for all possible weights. It contains the bounded variograms associated with covariances and also unbounded ones which have no covariance counterpart. By this we mean that you cannot choose one model up to a cerlain distance then a different one from there onwards as shown in Fig. The Variograrn 35 [3.

So there is a trade-off between the two hypotheses. The intrinsic hypothesis allows us to use a wider range of variograms but the weights must sum to O. B Admissible models We have seen that in order to ensure that the variance of any linear combination never goes negative. Nor can they be combined piecewise. As it is not easy to recognize functions that have this properly or to test for it. For more information on how to test for positive definiteness see Armstrong and Diamond A list of the common models is given in the next section.

Example of a function that is NOT allowable as a variogram model In order to work out whether a cerlain function is or is not positive definite.

Covariances must be positive definite functions. These can be added to obtain other admissible models because this is equivalent to adding independent random functions. Piecewise linear model that is admissible in 10 but not in 20 or higher dimensions Having been warned of the dangers of trying to invent their own variogram models. For example the piece-wise linear function shown in Fig. This list is not exhaustive. Those with a sill correspond to stationary regionalized variables while the unbounded models are associated only with intrinsic variables.

Several exercises at the end of the chapter illustrate this procedure. Exercise 3. The resulting model must. Basically they were developed by mathematically constructing a random function and calculating its variogram theoretically.

The tangent at the origin intersects the sill at a point with an abscissa 2a This can be useful when fitting the parameters of the model. The differences are quite obvious. The exponential rises more rapidly initially but only tends towards its sill rather than actually reaching it. The Variogram 37 3.

As both the spherical and the exponential models are linear for small distances. It has a simple polynomial expression and its shape matches well with what is often observed: The tangent at the origin intersects the sill at a point with an abscissa a. The gaussian model represents an extremely continuous phenomenon. The value of A. Experience shows that numerical instabilities often occur when this is used without a nugget effect.

Working in degrees. The cardinal sine model 3. When calculating this on a pocket calculator. The Variogram 39 3. The last two models were developed to model different types of gravimetric and magnetic anomalies. It corresponds to very continuous structures.

Note the mirror image around the y axis. These variograms all have sills so they correspond to stationary variables.

If a pure nugget effect had been used to generate a simulation. So it is important to understand the implications of this choice in terms of the continuity of the variable or. Comparing the four figures. Write down the equation for the spherical model with a range of m and a sill of 2. Variograms that are quadratic near the origin come from highly continuous variables. The images are pixels x pixels. Unless very closely spaced data are available. For a spherical and an exponential with the same sill and the same range.

Simulation methods are beyond the scope of linear geostatistics. Are they very similar or do they have obviously different fealures? To highlight the differences between the variograms. A zero nugget effect was used throughout. It is important to keep the relation between the variogram model and its realizations in mind when fitting models to experimental variograms and later when kriging. The darker and lighter patches in the figures are elongated with these dimensions on average.

This obvious difference is due to the fact that the spherical and the exponential models are linear near the origin whereas the other two are parabolic. When most people look at the equations. Before fitting models to experimental variograms it is important to become more familiar with their properties. The EW range or the practical range was set to 20 pixels whereas the NS one is half that.

New in Description Based on a postgraduate course that has been successfully taught for over 15 years, the underlying philosophy here is to give students an in-depth understanding of the relevant theory and how to put it into practice. This involves going into the theory in more detail than most books do, and also discussing its applications.

It is assumed that readers, students and professionals alike are familiar with basic probability and statistics, as well as the matrix algebra needed for solving linear systems; however, some reminders on these are given in an appendix.

Exercises are integrated throughout, and the appendix contains a review of the material. Product details Format Paperback pages Dimensions x x Table of contents 1 Introduction.

Collect and check data. Ordinary kriging of a block. Adding extra samples improves the quality of the estimate.

## Explore other courses under this discipline:

Review of Basic Maths Concepts.As all the other logs lay in the range from By detinition its variance [ In Fig. Margaret A. site Giveaway allows you to run promotional giveaways in order to create buzz, reward your audience, and attract new followers and customers.