Page 192 - Modern Spatiotemporal Geostatistics
P. 192

Modifications  of  BME  Analysis              173

        EXAMPLE   9.4:  As  general  knowledge  we  consider  the  groundwater  flow  law
        (Eq.  3.25, p. 81). The  flow  equation  offers  a  physical  basis  for  relating  the
        stochastic  moments  of  hydraulic  head  and  hydraulic  log  conductivity.  A pos-
        sible  discretization  of  this  law  is  given  in  Equation  3.26.  Specificatory  data
        include  the  hard data {Xi,j-i,k,  X»-i,j,fc.  X»,j+i,fc. V'i-iAj.fc.  ^»j+i/2,fc} and
        the soft  (interval) data {Xi+i,j,k,  ^i,j-i/2,fc. i>i+i/2,j,k}•  An estimate of X(p)
        is sought  at the point p  = (i,j, k).  Then, a possible term  of the corresponding
        BME  equation  may  be as follows












        where the vector  I  denotes the interval  data  domains.  As was mentioned  in
        Example  3.10  (p.  81),  in  a  realistic  flow  analysis  several  terms  of  the  form
        of  Equation  9.21 will  be  included  in  the  BME  equation.  Depending  on  the
        situation  (objectives  of  the  analysis,  statistics  available,  etc.),  other  forms  of
        the  ^-operator  are  also  possible.  The  BME  equation  should  be  solved  for
        Xi,j,ki  usually numerically.


            The  analysis  above suggests  an interesting  approach of studying stochas-
        tic  algebraic  and  differential  equations  representing  physical laws.  Generally,
        given  that a natural  variable X  obeys an equation  of the form  D(X,  Y)  =  0,
        where  Y—  (Yi,...,Yfc)  are observed variables, we wish  to  find  solutions  of
        the  corresponding  stochastic  expectation  equation,  say D(X,  Y)  =  0.  The
        traditional  approach is either  to  solve the  original  physical equation  for X  and
        then  take  the  expected  value of  the  solution,  or  to  solve  the  corresponding
        expectation  equation  directly  for  the  moment  of  interest.  Alternatively,  BME
        analysis  suggests  another  way, as  follows:  Include  the  expectation  equation
        in  the  BME  mapping  process together  with  any  other  form  of  general  and
        specificatory  knowledge available.  BME will  search for  solutions  such that  the
        expectation  equation  is an identity.  A  similar approach can  be used  in the  case
        of a series of physical  equations


        Transformation     laws

        Now we consider another interesting application  of multivariable  BME analysis.
        Assume that  a natural variable X(p)  can be expressed in terms  of  a secondary
        variable Y(p)  by  means  of  a transformation  law of  the  form
   187   188   189   190   191   192   193   194   195   196   197