BIAGI_final:Layout 6 ANNALS OF GEOPHYSICS, 57, 1, 2014, S0190; doi:10.4401/ag-6367 S0190 The collocation approach to Moho estimate Riccardo Barzaghi, Ludovico Biagi* Politecnico di Milano, Dipartimento di Ingegneria Civile e Ambientale, Milan, Italy ABSTRACT In this paper, the collocation approach to Moho estimate is presented. This method is applied to the inversion of gravity data that can be com- plemented by Moho depth information coming from e.g. seismic infor- mation. In this context, a two layers model is considered and discussed in order to give a general theoretical framework for the inversion method. A body with two inner constant density layers and an inner separation sur- face between is considered and a uniqueness theorem is proved for the es- timability of the separation surface given the gravity outside the body itself. Based on this result, a discussion is given on the estimation of the Moho depths based on terrestrial gravity observations. The observation equation is presented and its local planar approximation is derived. The application of the collocation method to the estimate of Moho depths is then studied and discussed in relationship to the planar observation equation. Also, numerical tests are presented. To this aim, the collocation inversion algorithm is implemented and tested on simulated data to prove its effec- tiveness. The results show that the proposed method is reliable provided that proper data reductions for model discrepancies are taken into account. 1. Introduction The transition layer between the upper mantle and the lower part of the crust is defined as the Mohorovicic discontinuity (Moho) and ideally considered as a surface [Turcotte and Schubert 1982]. On a global scale, this sur- face has sharp variations, ranging from over 60 km to less than 5 km depth. The density variation occurring across the Moho between the upper mantle and the lower crust is generally set at 0.5 g/cm3, based on the assumption that the upper mantle density is 3.4 g/cm3 and the lower crust density is 2.9 g/cm3 [Anderson 1989]. The Moho transition layer causes seismic wave refraction and re- flection and the related gravity signal as measured on the Earth surface can have a standard deviation ranging from 50 to 100 mGal. Moho depths can be estimated using both seismic and gravimetric inversion methods [Parker 1972, Lebedev et al. 2013]. Seismic methods can give more accurate estimates that are not however homoge- neously distributed over the Earth. On the contrary, grav- ity is quite homogeneously distributed over the entire Earth and only relatively small areas are un-surveyed [Pavlis and Rapp 1990]. Furthermore, satellite dedicated gravity missions [Reigber et al. 1999, Albertella et al. 2002, Tapley et al. 2004] have made available a large amount of data that can be profitably used in combination with ground based gravity data [Shin et al. 2007]. Recently, an approach based on the integration of gravity and seismic information has been proposed by Eshagh et al. [2011]. They devised a method based on a stochastic combina- tion of seismic and gravity Moho models. In this paper, the gravity estimation of the Moho based on the collo- cation method is presented [Krarup 1969, Moritz 1989, Barzaghi et al. 1992] which also allows integration of gravity and seismic information available in the investi- gated area. The basic idea of this method is to propa- gate the covariance structure of the Moho depth to the covariance of the observed gravity. Given gravity ob- servations and Moho depth information (e.g. from seis- mics), the Moho collocation estimate can be obtained as a linear combination of these data. Also, collocation allows the computation of the variance of the estima- tion error that can be used as an indication of the esti- mated depth precision. This approach is here devised for a two layers model and a uniqueness theorem is proved as a theoretical background for the proper ap- plication of the method (see Appendix). In this context, the collocation inversion method for the two layers model is presented and the impact of model discrep- ancies on the Moho estimate is discussed. Computation problems are addressed too. The method effectiveness is then tested using simulated gravity data. 2. The observation equation for the gravity effect of a two layers body In the following the two layers body is considered. This is defined as a body with an inner volume and an external layer: the two parts have different densities and Article history Received June 14, 2013; accepted November 15, 2013. Subject classification: Moho, Gravity collocation, Gravimetric inverse problem. are separated by a unique surface. Particularly, we con- sider a spherical two layers body. We also assume that this spherical body has a radius R = 6380 km and that the surface separating the external layer from the inner volume is at a depth ranging from 5 to 60 km. In a very simplified way, this can be assumed to be the reference model for the Earth and the Moho. Given this model, a new method is developed to obtain a Moho estimate on regional areas, e.g. areas having a 6° × 6° extension. Furthermore, the method and the equations are given in planar approximation. Gravity observations are supposed to be known on the surface of the sphere representing the Earth. In addi- tion, Moho depths can be given at scattered points. The density of the outer layer tC and the inner density tN are assumed as known and independent from the radial com- ponent. So, tC and tN depend only on the spherical co- ordinate v, v = ({,m). This is a special case in the theorem proved in the Appendix and thus it still holds that the inner separation surface is unique. The potential gener- ated from this body in a point P on the external surface is expressed by V(P) = NN tN + NC tC , where NN and NC are the Newtonian operator applied to the inner volume and the outer layer respectively. If we define a normal poten- tial U(P) = NB tN = NN tN + NC tN (NB is the Newtonian operator extended to the whole body), we can define the anomalous potential T as T(P) = V(P) − U(P) = NC (tC − tN) = NC (t), where we have put t = tC − tM . This defi- nition implicitly poses the density of the normal potential equal to the mantle density. Other choices can be given and are available in literature, for example by putting the ‘normal’ density equal to the Earth mean density. How- ever, in our application the gravity anomaly is derived by T and a different choice of density just affects the mean dg that in any case will not be used. Thus T(P) is given by [Moritz 1990, Sjoberg 2009] P is a point on the external surface, RP , vP are the ra- dial and angular coordinates of P: hypothesizing a spher- ical body, RP = . rPQ = (]PQ is the angle between OP and OQ, O being the center of the sphere). The term is the depth of the separation surface between the two layers. The equa- tion for gravity dg = −^T/^R is then Since we can express dg in the following way with QR and QR−H running over the outer sphere sur- face and the inner separation surface. The planar ap- proximation of these equations can be obtained in a reference system having the Z axis along the plumb line in the midpoint O of the investigated area and the (X,Y) plane tangent to the outer sphere in O (see Fig- ure 1). We define n' as the inner surface normal vec- tor and n as the unit vector along the radial direction. If we name dS as the area element of the inner sepa- ration surface, we have dSn' · n. The vector n can be further decomposed into ez + dex , being ez its component along the Z axis and dex the component onto the (X,Y) plane. It follows that n' · n = n' · eZ + n' · deX . If we assume to consider areas having maximum width of about 6°, dex , which is the tangent of the angle between n and ez , has a maximum value that is 1/20 of ez . This implies that, at the first order, we can approx- imate n' · n � n' · eZ , that is � dSn' · eZ. The second term in the equation giving dg can be thus simplified accordingly This integral can be further simplified by assum- ing � 1 (this is true with an error which is, at maximum 1%). In this way, it follows that 2 cosR RQ Q PQ 2h h ]+ -2R ( )R H v- T P Q r Q Q G r Q dB G d d PQ C N N PQ C C Q Sphere Q R H R 2 t t v hh t t- = - = = v- ^ ^ ^ ^ ^ ^ h h h h h h ### ## # ,g R G d d R r 1 P R R S H Q PQ 2 2 2 d v vt v h h=- v- ^ ^ ^ h h h ## # 3 2 R r r r r r r r r r R r R r 1 1 2 R PQ Q PQ Q Q PQ PQ Q PQ Q PQ Q PQ Q PQ Q R PQ Q PQ Q PQ Q 2 2 3 2 3 2 2 3 2 3 2 2 2 2 2 2 2 h h h h h h h h h h h =- - = =- - + = = - =- + h h h h c ` c c c c m j m m m m ( )d R H 2v - = ( )d R H 2v - , , R G d r R H R G dxdy x y r R H x y PQR H S R PQR H 3 2 , , vt v v t - - - - - - ^ ^^ ^ ^^ h hh h hh ## ## ( ) /R H R- , R G d r R H G d d r 1 PQ S R PQ 3 R H R H2 , , vt v v p it p i - - - - - ^ ^^ ^ h hh h ## ## ,g R G d r R R G d r R H P S PQ PQR H S 2 3 R d v vt v vt v v = + - - - ^ ^ ^ ^^ h h h hh ## ## BARZAGHI AND BIAGI 2 (2.1) (2.2) (2.3) (2.4) (2.5) (2.6) 3 The same holds for the first integral. Hence that is where . This is a non-linear integral equation between H(x,y) and dg. In order to get it in a form that is suitable for applying collocation, we have to linearize it. To this aim, we define (E[+] is the mean oper- ator) and we set H(x,y) = . More- over, let assume that : note that in some areas this is not true and the resulting approximation could represent one possible limitation of this approach that must be carefully considered in its application. In any case it is needed for getting dg(x,y) as a linear func- tional of f(x,y) as required for applying collocation: more investigations on the approximation consequences will be discussed in the tests of Section 4. Therefore, considering only the linear terms in f, we obtain If we further consider the density contrast between the crust and the upper mantle, we set and . Thus, the equa- tion for dg(x,y) can be written as with In its linearized form, the dg(x,y) observed gravity is given as the sum of four terms that will be analyzed in the following. Integrals I2, I3 are linear terms in f and dt while I4 contains the product f(p,i)dt(p,i). This last integral is thus a second order term which is likely to be smaller than I2. So, in a first approximation, we assume that I2 + I4 � I2. If we now disregard I4 , we have E[dg] = E[I1] + E[I2] + E[I3]. In our hypotheses on f and dt, it holds that , which implies that . So, in principle, one can estimate from observed gravity. This is indeed possible only in theory but not in practice due to the standard defini- tion of observed gravity anomalies (there is a mismatch in the reference density between our definition and the standard one based on the normal ellipsoidal field). Thus we assume to have an estimate of as given in- dependently (e.g. from seismic information). The last term to be discussed is I3. It gives the grav- ity effect due to density variations over the thickness Hmax %f ( , ) ( ) ( )d x yxy 2 2p i p i= - + - ( , )H E H x y= 6 @ [ ( , )]E x yt t= , , , , , , , , , , , , , , , ,g x y H H x y H x y I H x y I x y I I 0 4 1 2 3 d t t f dt dt f = + + + + ^ ^ ^ ^ ^ h h h h h [ ( , , , , )] [ ( , , , )] 0E I H Hx y E I x y2 3t f dt= = [ ] / (2 )H E g Gd r t= H , , , , g x y G d d r G d d r 0 PQ R PQ R H R R2 2 d p i t p i p i t p i = + - - ^ ^ ^ h h h ## ## , , , , , , g x y G d d d d H 0 1 1 / R xy xy 2 2 1 2 2 ,d p i p i p i p i p i + - + ^ ^ ^ ^ ^ h h h h h ; 6 @ E ## , , , , , , , , g x y G d d d H d G d d H d H 0 1 1 / / R xy xy R xy 2 1 2 2 3 2 2 2 t t d p i p i p i p i p i p i p i f p i = + - + + + + 2 2 ^ ^ ^ ^ ^ ^ ^ h h h h h h h ; 6 6 @ E @ ## ## , , , , , , , , , , , , , , , , I H G d d d H d G H I H G d d H d H I H G d d d H d I H G d d H d H 1 1 2 1 1 / / / / R xy xy R xy R xy xy R xy 1 2 1 2 2 2 3 2 3 2 1 2 3 2 3 2 2 2 2 2 t p it p i p i r t t f p it p i f p i dt p idt p i p i p i dt f p idt p i p i f p i = + - + = = + = + - + = + 2 2 2 2 ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ h h h h h h h h h h h h h h ; 6 6 ; 6 6 @ E @ @ E @ ## ## ## ## H THE COLLOCATION APPROACH TO MOHO ESTIMATE Figure 1. Spherical geometry and planar approximation. ( , ), [ ] 0H x y Ef f+ = [ ] ( , , , ) 2E g I H x y G H1d t r t= = ( , ) ( , ), [ ] 0x y x y Et t dt dt= + = (2.7) (2.8) (2.9) (2.10) (2.11) layer. The impact of this term can be large as it can be assumed that, in the mean over the whole Earth, � 30 km. Hence, this term must be locally modeled based on available crustal density variation data in the inves- tigated area. The amplitudes of the four integrals terms and the impact on the estimated f values will be further discussed in the numerical simulations given in para- graph five. 3. The collocation estimate method It is well known [Menke 1989, Tarantola 2005] that the gravity inversion is unstable. Thus the estimation algorithm to be applied should be stable and smooth with respect to high frequency gravity signal compo- nents. Collocation method can be successfully applied to this aim. It is a stochastic filtering/predicting method that filters out high frequencies based on the covariance structure of the data. It has been widely applied in Ge- odesy [Tscherning 1985, Moritz 1989] particularly in view of the use of heterogeneous data. In our applica- tion, this implies that dg data can be used together with f known values (coming from e.g. seismic) to get an integrated estimate of f. This method can be properly applied to the ob- servation equation . Thus, in order to get unbiased estimates, we need to model and remove the signals related to the I3 and the I4 terms that account for density variations. As it will be proved in the simulations, these terms give a sig- nificant signal having also a frequency signature that couples with I2. However, it can be reasonably as- sumed that I4 ≪ I2 and I4 ≪ I3. Thus, the proposed es- timation method is performed according to the following steps. 1) Estimates of and dt(x,y) are given using avail- able information on t(x,y). 2) On observation points, the I3 term effect is esti- mated based on the given values of dt(x,y) and (this last term can be derived from known geological infor- mation and/or from seismic profiles). This effect can be straightforwardly computed using the closed for- mulas of the gravity effect of a prism [MacMillan 1958]. 3) Reduced gravity data are estimated as dgI(x,y,0) = dg(x,y,0) − E[dg] − and, assuming that dgI(x,y,0) � are estimated by col- location. 4) Using the estimated values, I4 is com- puted and subtracted from dgI(x,y,0) to get dgII(x,y,0) = dgI(x,y,0) − . 5) Then, the last computation step is accomplished using dgII(x,y,0) as input data and getting, by colloca- tion, . Thus, in the end, In order to see how collocation can be applied to get the estimate of f, a brief presentation is given in the following. A more detailed description can be found in Barzaghi et al. [1992], Biagi [1998] and Fors- berg [1984]. We assume that the following linear relationship between gravity (dg) and depth (f) holds Assuming that and are known, the colloca- tion estimate of f can be obtained based on dg obser- vations and f values given in some scattered points. To get these estimates, we further have to assume that: 1) f is a weak stationary stochastic process, ergodic in the mean and in the covariance 2) the noises in gravity and depth, ng and nf are un- correlated zero mean signals 3) the cross-correlation between signals and noises is zero. Auto and cross covariances can then be summa- rized as follows Using these covariances and observations plus noise we can get the following auto-covariance matrices of the observations The collocation estimate of f [Moritz 1980, Barza- ghi et al. 1992] in a point Pk = (xk ,yk) is given by with (3.9) ( , , 0) [ ] ( , , , , )g x y E g I H x y2d d t f- = t H ( , , , )I H x y3 dt ( , )x yIft ( , , , , )I H x yI4 dt ft ( , ) ( , )H x y H x yf= + =t t( , )x yIIft tH , , , , g x y H H G d d d 0 / xyR 2 2 3 2 2 d p it p i f p i = + ^ ^ ^ h h h 6 @ ## , , ; , , ; C g g C P P C g g C C P P C i j g g i j j i i j i j j i d d d d f f f f = - = = - = d d ff ^ ^ ^ ^ ^ ^ h h h h h h , , , ; , , C g C P P C g C n n C n g C n 0 i j g i j j i i j ij i 2 d d d v d f f f = - = = = = fd^ ^ ^ ^ ^ ^ h h h h h h , , , , ..., , , , ..., g g n n g g x y i N x y i N 1 1i OBS g i i i g OBS i i=f f d dd d f f = + = = = + = f f ^^ ^^ h h h h C g g g g n Cn I E E ET g gOBS OBS OBS T g g g g T n 2 g v d d d d = = = + = + d d d d d d d 6 6 6 @ @ @ C n C Cn E E ET OBS OBS OBS T T nn f f ff = = = + = + ff ff f f f 6 6 6 @ @ @ ( , ) .H x yIIf+ t H c c C lg T T ll 1f= fd ff -t 6 @ , ,,c cC g Cg k OBSi k OBSii if d f f= =fd ff^ ^h h BARZAGHI AND BIAGI 4 (3.1) (3.2) (3.3) (3.4) (3.5) (3.6) (3.7) (3.8) ( , , , , ), ( , )I H x y x y2 f t fIt 5 To compute this estimate, we first have to esti- mate the empirical covariance function of dg which must be then modeled with proper positive definite model functions [Moritz 1980]. Also, models for auto and cross-covariances must be defined in order to com- pute the matrices defining the f estimator. This can be done via covariance propagation law applied to linear functionals [Moritz 1980]. In our ap- proach we assumed where LP [f] indicates the linear functional relating f to dg. It follows by covariance propagation law that A suitable model for C ff can be (J0 : zero order Bessel function) which implies [Biagi 1998] Another model that can be used for the auto-co- variance of dg is (J1 : first order Bessel function) which gives The A and the a values contained in these models are tuned to fit the empirical estimated covariance val- ues of dg. As stated above, models for C ff and C fdg can be then derived, allowing the computation of the ma- trices contained in the estimation formula for f. 3.1. Some comments on the efficient computation of the estimation formula One numerical major problem in computing the collocation estimate of f is the inversion of the covari- ance matrix of the data Cll. Some standard data config- urations are discussed here. We firstly assume that the number of known Moho depths NH coming from seis- mic is at maximum 200. We then consider three differ- ent gravity data distributions: 1. A quite poor gravity data coverage and depth in- formation: 2. High gravity data coverage and no depth infor- mation: 3. High gravity data coverage and depth informa- tion: Case 1. Can be handled using standard techniques such as Cholevsky decomposition [see e.g. Press et al. 1992]. The efficient numerical computations of case 2 and 3 can be obtained if we assume that gravity data are regularly gridded. In this case, the covariance matrix has a Toeplitz/Toeplitz structure: based on this partic- ular structure, in case 2, Fast Collocation can be applied [Bottoni and Barzaghi 1993]. Case 3 must be discussed in more details. Also in this case, we assume to have gridded gravity data. On the other hand, depth infor- mation are assumed to be given on scattered points. Under these assumptions, an efficient solution can be obtained by partitioning the solution. We have to solve the following system The partitioned solution is given by with In this way, the term m can be computed via Fast Collocation being dg on a grid. The same holds in com- , , , ,C P Q C P Q L L C P Q L C P Qg g g P Q P = = fd ff d d ff ^ ^ ^ ^ h h h h 6 6 @ @ ,C P Q AJ P Q0 a= -ff^ ^h h , 2 , C P Q G AJ P Q C P Q G e AJ P Q e 2 2 H H g g g 0 2 0 2 r t a r t a = - = - d a fd d - -^ ^ ^ ^ ^ ^ h h h h h h C r x AJ x g g 1 a a = d d ^ ^ h h , , 2 C P Q G A dkke J P Q C P Q G A dkke J P Q k k 2 kH kH g 2 2 0 0 0 0 r ta r ta = - = - ff a f a ^ ^ ^ ^ ^ h h h h h # # 1000, ( ) [ , ]dimN C N N N N 1200g ll g gH H1 1= + + 1000, ( ) [ , ]dimN C N N 1200g ll g g2 2= , ( ) [ , ]dimN C N N N N1000 1200g ll g H g H2 2= + + v v v C C C C gTg g gg g OBS OBS 1 OBS OBS d f = = f d d fd fd ff - ; = ;E G E v c v v N N N OBS g T N N N Ng 1 1 1 1 1 1 H H H H g gg f m m X C = - = - # # # # # # fdf f - ^ ^ h h 6 6 6 6 6 6 @ @ @ @ @ @ I g C C C C C C C C C C C 0 0nn nn OBS OBS g g g g g g g ll g g OBS OBS d f = = = = = d d fd d ff d d fd d f ff f d f ; = = = E G G G , , , , g x y G d d H d H L 0 / R xy P2 2 3 2 2 d p it p i f p i f = = + = ^ ^ ^ h h h 6 6 @ @## C g C C N gg N N OBS N N g g N N g N N T 1 1 1 g N H g g H g g g 1g g C m d= = # # # # #d d fd - - # d d dd 6 6 6 6 6 6@ @ @ @ @ @ C CN N N N N N N NgH H H H H g Hg CX = -# # # #ff fd6 6 6 6@ @ @ @ THE COLLOCATION APPROACH TO MOHO ESTIMATE (3.10) (3.11) (3.12) (3.13) (3.14) (3.15) (3.16) (3.17) (3.18) (3.19) (3.20) puting C, since we can apply Fast Collocation NH times in solving Ci = . , where ci is the i-th column of , i = 1,..., NH. The memory allocation for this method is equal to (Ng + NH)(1 + 2NH) since we have to store the first row of the Toeplitz/Toeplitz matrix Cdgdg, the full C ff and C fdg matrices, the C and X matrices and the vector m. This memory occupation is smaller than those implied by the full Cll matrix storage. Further- more, in terms of computation effort we have (NH + 1) Fast Collocation solutions having Ng dimension and 2 Cholesvky based solutions having NH dimension. The computation time is thus proportional to (NH + 1)Nd 2 g logN dg + 2NH 3 which is less than the one require for solv- ing the full Cll system by standard Cholevsky method. A last remark is in order. Usually collocation is a stepwise procedure since it is an adaptive filter which is tuned by the empirical covariance. Thus, in the first step, the low frequency component is dominant while in the further steps the high frequencies become dom- inant. At each step, residuals on observation points are obtained as the difference between observed and collo- cation predicted values. The iterative process stops when residuals (e.g. gravity residuals) are uncorrelated. 4. The simulated tests The devised method has been tested via numeri- cal simulations that aim at verifying: - the properties and the stability of the inversion procedure under model/data consistency, - the impact of known HOBS data in the inversion, i.e. the improvements that one can get with respect to the inversion based on gravity data only, - the biases induced in the depth estimates when model inconsistencies are present, like for example, er- rors in the gravity contrast model or in the mean Moho depth. A surface depth has been simulated on a regular (x,y) grid with 10 km grid knot over a rectangular area 600 km × 400 km. The depth values are given with respect to mean depth while the f values have been generated according to the pro- cedure described in Barzaghi et al. [1992]. The a-priori covariance structure of the f values has been fixed ac- cording to the following parameters The simulated depths (Figure 2) have the follow- ing statistics v(f) = 6.4 km min(H)= −43.67 km max(H)=−16.00 km As one can see in Figure 2, the resulting depth sur- face is smooth and regular which seems to be quite far from the real Moho behavior, at least in some particular areas where high frequencies features are present (e.g. the Alpine region). However, this is not so relevant since in the devised tests we are aiming at defining the impact of model inconsistencies on the inversion procedure. This can be critically evaluated even if high frequency pattern is not present in the target surface. Generally, the simulated depth does not respect the condition adopted for the (2.9) linearization, i.e. : this will allow a sensitivity assessment of the proposed al- C g T fd C cg g i 1 d d - ( , ) ( , )H x y H x yf= + , 55 km , 0.0160 kmC AJ r Ar 0 2 1a a= = = ff -^^ hh 31.51 kmH =- Hmax %f BARZAGHI AND BIAGI 6 Figure 2. Simulated depth. 30 kmH = (4.1) 7 gorithm with respect to this critical approximation. Starting from the simulated depths, the gravimet- ric effect has been computed according to two different density contrast models. In the first case a constant den- sity contrast t(x,y) = t0 = −0.52 g/cm 3 has been con- sidered. In the second case, it is assumed that the density contrast varies linearly with x: In both the cases (Figure 3), the gravimetric effect has been computed at z = 0 by prism integration for- mula [MacMillan 1958] over the same grid used for the depth values. In the following, the two computed grav- ity fields will be named GC(x,y) (constant t0 value) and GL(x,y) (linearly varying tL value). Furthermore, three depth sections have been computed based on the H(x,y) grid. In this way, we simulated 2501 H(x,y) grid values, 2501 gravity grid values and a total of 100 scattered depth observations on the three distinct sections. These preliminary simulations are aimed at the discussion of the theoretical model and the analysis of the implemented software: therefore, no noise has been added to the simulated observations. Three different inversions have been carried out: 1) Integrated inversion based on GC(x,y) gravity , ; . g/cm , . g/cm ; . g/cm x y x x y0 56 0 50 0 53 L L 0 3 3 0 3 # # t t a t t = + - - =- ^ ^ h h THE COLLOCATION APPROACH TO MOHO ESTIMATE Figure 3. Gravity field with constant density model (top) and differences of the gravity field with the linear density model (bottom). and known depth values assuming a constant gravity model t0. This is the case when data and density are consistent. 2) Inversion based on GC(x,y) gravity data only as- suming the constant gravity model t0. Also in this case data and density are consistent. 3) Integrated inversion based on GL(x,y) gravity and known depth values assuming that density is lin- early varying following the tL(x,y) rule. Once again data and gravity model are consistent. 4) Integrated inversion based on GL(x,y) gravity and known depth values assuming that density is con- stant and equal to t0. This is the critical case where we have a mismatch between data and density model. In all the inversions, the model covariance func- tion of the gravity field has been fitted on the empirical covariance function of the simulated gravity observa- tions [Barzaghi and Sansò 1983]: therefore, it has been propagated by (3.14) and (3.16) to model the cross co- variance between gravity and depth and the covariance of depth. Data inversions 1) and 2) are made for testing the impact of known depths in the collocation inver- sion procedure. Test 3) is performed to check for the ef- fectiveness of the iterative inversion procedure described in paragraph 4, assuming to have a known density model consistent with the available data. Finally, test 4) is set to estimate the impact of improper density information on the integrated estimated depths. Inversion 1) should provide the benchmark results, for three reasons. The adopted density model in the in- version is consistent with the simulated model, all the simulated observations are used and the constant den- sity model does not require either approximations or iterations to manage I3 and I4 terms. In all the inversion procedures, as it would be in a real case, the mean depth has been estimated using the 100 given depth values. The obtained value is . In this way, the inversions are based on a biased mean depth. So, the impact of this bias on the results can be investigated too. Furthermore, as a general remark on all the inver- sion procedures, it must be underlined that two itera- tion steps were necessary to get the final estimates since first step residuals were still spatially correlated (which was not the case for the second step residuals). The estimated depths have been then compared with the simulated ones in order to evaluate their pre- cisions and accuracies (to avoid edge effects that can give biased statistics, comparisons are carried out in the 500 × 300 inner area). Statistics of the differences for all the four inversion procedures are listed in Table 1. Sta- tistics over the 100 known depths are also listed when such data are used in the different inversion procedures. Figures 4 to 6 report the results provided by inversion 3, that are the most interesting. The following remarks are in order. The results ob- tained for inversion 1 (see row one in Table 1) proved that depths can be correctly estimated even if the a-priori mean depth is biased. If scattered depth data are avail- able, they can correct the initial bias and the final mean km.H 32 56=- BARZAGHI AND BIAGI 8 E(D) (km) v(D) (km) Min(D) (km) Max(D) (km) Inversion 1: statistics over the inner grid points 0.5 0.4 −1.1 1.4 Inversion 1: statistics over the 100 scattered points 0.5 0.4 −0.7 0.9 Inversion 2: statistics over the inner grid points 2.2 0.4 0.2 4.1 Inversion 3 (I3 only): statistics over the inner grid points 0.5 0.5 −2.0 1.5 Inversion 3 (I3 and I4): statistics over the inner grid points 0.5 0.4 −1.3 1.4 Inversion 3 (I3 and I4): statistics over the 100 scattered points 0.5 0.4 −0.6 0.9 Inversion 4: statistics over the inner grid points 0.4 0.9 −2.6 2.3 Inversion 4: statistics over the 100 scattered points 0.2 1.1 −2.5 1.7 Table 1. Statistics of the residuals D = Hsimulated − .Hestimatedt 9 discrepancy between estimated and simulated depths is 0.5 km, which is negligible if compared to the mean depth of 30 km. If gravity data only are used (inversion 2) the a-priori bias remains in the final estimates, because no depth observations are given to tune the mean. In this case the mean of the differences is 2.2 km (see row three THE COLLOCATION APPROACH TO MOHO ESTIMATE Figure 4. Integrated inversion 3 (linear density model and consistent correction of I3 and I4 terms): prediction errors. Figure 5. Integrated inversion 3: gravity residuals and their covariance function. in Table 1). However, it must be stressed that the f val- ues are correctly estimated as it results from the standard deviation which is equal to the one obtained in inversion 1. The iterative procedure used in inversion 3 to correct for density heterogeneities leads to satisfactory results. It must be remarked that the computation and correc- tion for the I4 term improves the results, even if this term is remarkably smaller than I3 (Table 1, row four are sta- tistics base on modeling I3 term only; in row five, I3 and I4 are used in the estimation procedure). Generally, in cases where no model error is present in the inversion (inversions from 1 to 3) the standard de- viation of the residuals (0.4-0.5 km) is less than 10 % of the standard deviation of the simulated depths (6.4 km); moreover, the approximation adopted to linearize the ob- servation equations does not affect significantly the re- sults: also on this regard, results are completely satisfying. The adoption of an improper model for density heterogeneities has a significant impact on the esti- mated depths (see Table 1, row seven). This is true even in the case of the integrated inversion based on gravity and depth observations. The impact of density model mismatch can be evaluated in the residuals computed on the observed depths that show high discrepancies. Hence, the observed depth values cannot compensate for an improper density model thus confirming the im- portance of a reliable correction for existing density heterogeneities (see also row eight in Table 1; even the residuals on given depth points are worse than in the previous cases). 5. Conclusions The collocation method can be profitably used to combine gravity and depth information in order to get reliable Moho estimate. In this framework, a theoreti- cal background has been provided through a unique- ness theorem for a two layers body. Under quite general hypotheses on the body inner density distribution, it can be proved that the separation surface between the two layers can be uniquely estimated. Based on this the- orem, the estimation of the Moho is discussed and the equations relating gravity and Moho depth are derived in planar approximation. In this context, a procedure to Moho estimate based on collocation approach is pre- sented. This iterative approach has been then tested via numerical simulation. A simulated Moho is computed and two density models are assumed to compute the direct gravity effect to be used in the inversion proce- dures. These tests gave relevant insight in the devised inversion method. Reliable results have been obtained when density models are consistent with gravity simu- lated data. It can be thus assumed that, under unbiased conditions, the method is stable and can be applied to Moho depth estimate. Also, the impact of depth infor- mation proved to be relevant. Collocation allows easily the integration of different data types and this can im- prove the solution. Particularly, this applies to the bias in the mean depth that is reduced if depth data are con- sidered. On the contrary, the solution based on gravity only cannot compensate for this bias. Another impor- tant remark is related to density information. It was BARZAGHI AND BIAGI 10 Figure 6. Integrated inversion 3: residuals on the depth observations. 11 clearly proved that density plays an important role in the inversion. If proper density information are avail- able, the iterative inversion procedure based on collo- cation can give reliable results. On the other hand, when assuming inconsistency between density and sim- ulated gravity data, we obtained remarkable distortions in the estimated depths. Furthermore, in this case, depth information are not effective in reducing the bias and even on the given depth data a significant discrep- ancy is present. Thus, it can be concluded that the collocation based procedure can be applied for getting Moho depth estimates from gravity and depth information (from e.g. seismic). However, as it is for any other inversion method, improper density information can induce se- rious biases in the estimates even if depth data are in- cluded in the inversion. Two possible evolution lines of the collocation inversion method can be devised. One line is to investigate its behavior in local areas where high frequency Moho patterns are present. Being col- location quite a stable and regularizing technique [Barzaghi et al. 1992], this doesn’t seem to be a too crit- ical issue. The second possible application of this methodology is on global Moho estimates based on gravity data coming from the gravity dedicated satel- lite missions. Since proper spherical model equations can be quite easily derived and linearized [see e.g. Es- hagh et al. 2011], global applications of the collocation method seem to be feasible. Acknowledgements. The suggestions and the corrections of the anonymous reviewers really have helped the authors and im- proved the clarity of the paper. References Albertella, A., F. Migliaccio and F. Sansò (2002). GOCE: the earth gravity field by space gradiometry, Celes- tial Mechanics and Dynamical Astronomy, 83 (1-4), 1-15. Anderson, D.L. (1989). Theory of the Earth, Blackwell Scientific Publication, Boston-Oxford-London- Edinburgh-Melbourne. Barzaghi, R., and F. Sansò (1983). Sulla stima empirica della funzione di covarianza, Bollettino di Geodesia e Scienze affini, 42 (4). Barzaghi, R., and F. Sansò (1988). Remarks on the in- verse gravimetric problem, Geophysical Journal of the Royal Astronomical Society, 92 (3), 505-511. Barzaghi, R., A. Gandino, F. Sansò and C. Zenucchini (1992). The collocation approach to the inversion of gravity data, Gephysical Prospecting, 40, 429-451. Biagi, L. (1998). La stima della Moho mediante il me- todo della collocazione, Tesi per il conseguimento del titolo di Dottore di Ricerca in Geofisica, X Ciclo, Consorzio tra le Università di Genova, Modena e Torino. Bottoni, G.P., and R. Barzaghi (1993). Fast Collocation, Bulletin Géodésique, 67 (2), 119-126. Eshagh, M., M. Bagherbandi and L. Sjöberg (2011). A combined global Moho model based on seismic and gravimetric data, Acta Geod. Geoph. Hung., 46 (1), 25-38. Forsberg, R. (1984). A study of terrain reduction, den- sity anomalies and geophysical inversion methods in gravity filed modelling, Reports of the Depart- ment of Geodetic Science and Surveying, no. 355, The Ohio State University, Columbus, Ohio. Krarup, T. (1969) A Contribution to the Mathematical Foundation of Physical Geodesy, Meddelelse no. 44, Geodætisk Institut, København. Lebedev, S., J.M.-C. Adam and T. Meier (2013). Map- ping the Moho with seismic surface waves: A review, resolution analysis, and recommended inversion strategies, Tectonophysics; http://dx.doi.org/10.10 16/j.tecto.2012.12.030. MacMillan, W.D. (1958). The theory of the potential, Dover Publications Inc., New York. Menke, W. (1989). Geophysical data analysis: discrete inverse theory, International Geophysical Series, vol. 45, Academic Press, San Diego. Moritz, H. (1980). Advanced Physical Geodesy, Wich- mann, Karlsruhe. Moritz, H. (1990). The figure of the Earth: theoretical ge- odesy and the Earth’s interior, Wichmann, Karlsruhe. Parker, R.L. (1972). Geophys. J. R. Astr. Soc., 31, 447-455. Pavlis, N.K., and R.H. Rapp (1990). The development of an isostatic gravitational model to degree 360 and its use in global gravity modeling, Geophys. J. Int., 100, 369-378. Press, W.H., S.A. Teukolsky, W.T. Vetterling and B.P. Flannery (1992). Numerical recipes in Fortran. The art of scientific computing, 2nd edition, Cambridge University Press, Cambridge. Reigber, C., R. Casper and W. Päffgen (1999). The CHAMP Geopotential Mission, IAA 2nd Interna- tional Symposium on Small Satellites for Earth Ob- servation (Berlin, April 12-16), 25-28. Sansò, F. (1980). Internal collocation, Atti della Accad- emia nazionale dei Lincei, Memorie, Classe di Scien- ze Fisiche Matematiche e Naturali, serie 7, vol. 16, sezione I, fascicolo I. Shin, Y.M., H. Xu, C. Braitenberg, J. Fang and Y. Wang (2007). Moho undulations beneath Tibet from GRACE-integrated gravity data, Geophys. J. Int., 170 (3), 971-985. Sjöberg, L.E. (2009). Solving Vening Meinesz-Moritz problem in isostasy, Geophys. J. Int., 179, 654-662. THE COLLOCATION APPROACH TO MOHO ESTIMATE Tapley, B.D., S. Bettapur, M. Watkins and C. Reigber (2004). The gravity recovery and climate experi- ment: Mission overview and early results, Geophys. Res. Lett., 31 (9), L09607. Tarantola, A. (2005). Inverse problem theory and model parameter estimation, SIAM (Society for Industrial and Applied Mathematics), Philadelphia. Tscherning, C.C. (1985). Local approximation of the gravity potential by least squares collocation, In: K.P. Schwarz (ed.), Proceedings of the International Summer School on Local Gravity Field Approxima- tion (Beijing, China, August 21 - September 4). Turcotte, D.L., and G. Schubert (1982). Geodynamics: applications of continuum physics to geological problems, John Wiley&Sons. *Corresponding author: Ludovico Biagi, Politecnico di Milano, DICA, c/o Laboratorio di Geomatica del Polo Territoriale di Como, Como, Italy; email: ludovico.biagi@polimi.it. © 2014 by the Istituto Nazionale di Geofisica e Vulcanologia. All rights reserved. Appendix. A uniqueness theorem for the two layers model The gravity inversion problem can be defined as fol- lows: to estimate the inner density distribution of a body given gravity observations in the spatial domain outside the body. As it is well known, this problem is an ill-posed problem [Moritz 1990] because uniqueness is not, in gen- eral, guaranteed. As an example, concentric spheres hav- ing the same total mass homogeneously distributed can vary their densities and volumes to give the same poten- tial in the space outside the larger considered sphere. Thus, by inverting the gravity potential no unique solu- tion can be estimated. To have uniqueness in the gravity inversion problem one has to impose proper regulariza- tions and/or restrictions to the required solution [see e.g. Sansò 1980, Moritz 1990, Sjöberg 2009]. In the framework of the gravimetric Moho estimation problem, a unique- ness theorem for the two layers model is given in the fol- lowing. This theorem can be considered as an extension to a more general case of the one proved in Barzaghi and Sansò [1988]. It can be used as a reasonable model for the gravity inversion to Moho estimate in the real case. Two bodies B1 and B2 are given with two layers inner mass distributions and the same bounding surface S. The two inner volumes are Bi1 and Bi2 and the two crust layers are Be1 and Be2. The two separation surfaces between the inner volumes and the crust layers Si1 and Si2 are considered as regular and are parameterized as Sik = Rik (v), with k = 1,2 and v = ({,m) (see Figure A.1). Let ti1 and ti2 be the inner part densities and te1 and te2 the crust densities of the two bodies. These densities are given as a function of (r,v). Assume that two func- tions ti(r,v), te(r,v) are defined inside S and satisfy the following conditions Moreover, the following conditions hold: If V1 = V2 outside S it can be proved that Si1 and Si2 are the same surface, i.e. that Proof We consider a body B having a density distribution which is obtained by subtracting the B2 from the B1 den- sity in every point. Outside S, the potential implied by 0 , , , , , r r r r 0 e i e ir r2 2 1 1 # # t v t v t v t v ^ ^ ^ ^ h h h h , , , , , ,P r P P r P P r P P r P B B B B i i i i e e i i e e e e 1 1 1 2 2 1 2 2 6 6 6 6 ! ! ! ! t t v t t v t t v t t v = = = = ^ ^ ^ ^ ^ ^ ^ ^ h h h h h h h h R Ri i1 2v v=^ ^h h BARZAGHI AND BIAGI 12 (A.1) (A.2) (A.3) 13 this body is equal to zero since where N(t) is the Newtonian operator. The density function t has the following properties: Furthermore, having by hypothesis that , it follows We now define the two surface Si and Se are the bounding surface of B. We can further define Se+ and Si+ as those parts of Se and Si that bound the domain in B where t is positive. Similarly, we define Se− and Si− as those parts of Se and Si that bound the domain in B where t is negative. Further- more, we define v+ and v− as the projection of Se+ and Se− onto the unit sphere v. The potential generated by B is harmonic down to Se since the mass of this body is inside this surface. As stated before, the potential of B is equal to zero outside S. Due to the maximum and minimum property of the harmonic function, it is also equal to zero outside Se. It follows that t is in the orthogonal complement of LH 2 B We now define a function k(v) onto Se , satisfying the following condition where H v 1/2 is the Sobolev space ½ of the function hav- ing domain on the unit sphere. Based on the hypothesis on Ri1 and Ri2 , we can also assume that Re and Ri are regular surfaces. It fol- lows that . Furthermore, in the hypothe- sis stated for k(v), there is a unique harmonic function h(v) defined in B that on Se has the values assigned to k(v). If , it follows that . Hence Being the following decomposition for I holds where in in S S H 0 0 /e e 1 2 2 !l v l v l v= v + - + ^ ^ ^h h h' ( ) H /Se 1 2!l v ( )r h P Lr HB 22 ! ( )h P LHB 2! ,, 0 I P r h P h r dB d r r dr Re B r r P Ri 3 2 2 t v t v v = = = = v ^ ^ ^ ^ h h h h= G ### ### I I I I I I1 2 3 4 5= - + - - 0V N V VNB B B B B B1 2 1 2t t t= = - = - =^ ^h h 0P B BP B B i ie e 1 21 2 kk j6 !t =^ ^ ^h h h P P P P B B P P P P B B 0 0 e i i e i e e i 1 2 1 2 k k 6 6 1 2 ! ! t t t t t t = - = - - + ^ ^ ^ ^ ^ ^ h h h h h h 0 , 0 ,r r2 2$ #t t+ - , ,S R R S R R R Sup R Inf e e i i ii i i 1 2 21v v v v v v = = = =^ ^ ^ ^ ^ ^ h h h h h h 6 6 @ @ " " , , 0P f P dB f L B P HB 26 !t =^ ^h h### r e i i2 2#t t P r h P P r h P r P h P r h P P3 r r r r r 3 3 2 3 2 2 2 2 2 t t t t = + - - ^ ^ ^ ^^ ^ ^ ^ ^ h h h hh h h h h , , , I d R R R R h R e e i i i 1 3 3 v t v v l v t v v v = + - v+ + + + + + + + + ^ ^ ^ ^ ^ ^ h h h h h h 6 @ ## THE COLLOCATION APPROACH TO MOHO ESTIMATE Figure A.1. The two bodies with two layers inner mass distributions and the same bounding surface. (A.4) (A.5) (A.6) (A.7) (A.8) (A.9) (A.10) (A.11) (A.12) (A.13) (A.14) Since, obviously, I5 can be written in the following way It holds that I5 = 0. We suppose now that Re(v) differs from Ri(v). Under this hypothesis, we want to define the values assumed by the remaining integrals. Since h(p) is har- monic in B, it can assume its minimum or maximum values only on the boundary, i.e. onto Se. In the limit for l+(v) → 1, we obtain 0 ≤ l(v) ≤ 1, that is . It follows that By making use of the properties of ^rt− , we have that I4 < 0 which implies I3 − I4 > 0. The I2 term is pos- itive. If we define III as we have I2 < III which implies I1 − I2 > I1 − III . III can be further be decomposed in the following way Hence, we obtain that where M+ is the positive mass included in the domain where t is positive. It follows that If we now consider the limiting value for l, we can write the following Since we finally have I1 − III > 0. So, if we assume that Re(v) doesn’t coincide with Ri(v), it holds that I = I1 − I2 + I3− I4− I5 > 0. However, I must be equal to zero. This implies that Re(v) and Ri(v) must coincide, i.e. that the interface surfaces in the two bodies are the same. , , I I d R R h R M1 II i i i 1 3v t v v v - = - +- v + + + + + + ^ ^ ^^h h h h6 @## , ,d R R h R 1 0i i i 3 2v t v v v -- v + + + + + ^ ^ ^^h h h h6 @## , 3 ,I d r r r r dr Re II r Ri 3 22v t v t v= - v + + + ^^ ^hh h= G## # , ,I d r h r r dr Re Ri r 3 2 2v v t v= v+ +^ ^h h= G## # , ,I d r h r r dr Re r Ri 3 4 2v v t v= v- -^ ^h h= G## # 3 , ,I d drr r h r Re Ri 5 2v t v v= v ^ ^h h## # I h dBP P3 P B 5 t= ^ ^h h## 0 ( ) 1h P P B61 1 ! , ,I d R RR h 0i ii3 3 2vt v v v=- v - - - --^ ^ ^h h h## ,I d r r dr Re II r Ri 32v t v= v + + ^ h= G## # , , , , I d R R R R R h R e e e i i i 3 3 3 v t v v l v t v v v = + - v - - - - - - - - - ^ ^ ^ ^ ^ ^ h h h h h h 6 @ ## , , , , , I d R R R R d r r dr d R R R R M 3 Re II e e i i Ri e e i i 3 3 2 3 3 v t v v t v v v t v v t v v t v v = + + - = = + - - - v v v + + + + + + + + + + + + + + + + + ^ ^ ^ ^ ^ ^ ^ ^ ^ h h h h h h h h h 6 = 6 @ G @ ## ### ## , , , , , , , , I I d R R R R h R d R R R R M d R R R R h R M 1 1 II e e i i i e e i i e e i i i 1 3 3 3 3 3 3 l l v t v v v t v v v v t v v t v v v t v v v t v v v - = + - + - - - + = = - + - - + v v v + + + + + + + + + + + + + + + + + + + + + + + + + + + ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ^^ ^ ^ ^^ h h h h h h h h h h h h h h h h h h 6 6 6 @ @ @ ## ## ## BARZAGHI AND BIAGI 14 (A.15) (A.24) (A.25) (A.26) (A.16) (A.17) (A.18) (A.19) (A.20) (A.21) (A.22) (A.23) << /ASCII85EncodePages false /AllowTransparency false /AutoPositionEPSFiles false /AutoRotatePages /None /Binding /Left /CalGrayProfile (Dot Gain 20%) /CalRGBProfile (sRGB IEC61966-2.1) /CalCMYKProfile (U.S. Web Coated \050SWOP\051 v2) /sRGBProfile (sRGB IEC61966-2.1) /CannotEmbedFontPolicy /Warning /CompatibilityLevel 1.3 /CompressObjects /Tags /CompressPages true /ConvertImagesToIndexed true /PassThroughJPEGImages true /CreateJobTicket false /DefaultRenderingIntent /Default /DetectBlends true /DetectCurves 0.1000 /ColorConversionStrategy /LeaveColorUnchanged /DoThumbnails false /EmbedAllFonts true /EmbedOpenType false /ParseICCProfilesInComments true /EmbedJobOptions true /DSCReportingLevel 0 /EmitDSCWarnings false /EndPage -1 /ImageMemory 1048576 /LockDistillerParams true /MaxSubsetPct 100 /Optimize false /OPM 1 /ParseDSCComments true /ParseDSCCommentsForDocInfo true /PreserveCopyPage true /PreserveDICMYKValues true /PreserveEPSInfo true /PreserveFlatness true /PreserveHalftoneInfo false /PreserveOPIComments false /PreserveOverprintSettings true /StartPage 1 /SubsetFonts true /TransferFunctionInfo /Apply /UCRandBGInfo /Preserve /UsePrologue false /ColorSettingsFile (None) /AlwaysEmbed [ true /AndaleMono /Apple-Chancery /Arial-Black /Arial-BoldItalicMT /Arial-BoldMT /Arial-ItalicMT /ArialMT /CapitalsRegular /Charcoal /Chicago /ComicSansMS /ComicSansMS-Bold /Courier /Courier-Bold /CourierNewPS-BoldItalicMT /CourierNewPS-BoldMT /CourierNewPS-ItalicMT /CourierNewPSMT /GadgetRegular /Geneva /Georgia /Georgia-Bold /Georgia-BoldItalic /Georgia-Italic /Helvetica /Helvetica-Bold /HelveticaInserat-Roman /HoeflerText-Black /HoeflerText-BlackItalic /HoeflerText-Italic /HoeflerText-Ornaments /HoeflerText-Regular /Impact /Monaco /NewYork /Palatino-Bold /Palatino-BoldItalic /Palatino-Italic /Palatino-Roman /SandRegular /Skia-Regular /Symbol /TechnoRegular /TextileRegular /Times-Bold /Times-BoldItalic /Times-Italic /Times-Roman /TimesNewRomanPS-BoldItalicMT /TimesNewRomanPS-BoldMT /TimesNewRomanPS-ItalicMT /TimesNewRomanPSMT /Trebuchet-BoldItalic /TrebuchetMS /TrebuchetMS-Bold /TrebuchetMS-Italic /Verdana /Verdana-Bold /Verdana-BoldItalic /Verdana-Italic /Webdings ] /NeverEmbed [ true ] /AntiAliasColorImages false /CropColorImages true /ColorImageMinResolution 150 /ColorImageMinResolutionPolicy /OK /DownsampleColorImages true /ColorImageDownsampleType /Bicubic /ColorImageResolution 300 /ColorImageDepth -1 /ColorImageMinDownsampleDepth 1 /ColorImageDownsampleThreshold 1.10000 /EncodeColorImages true /ColorImageFilter /DCTEncode /AutoFilterColorImages true /ColorImageAutoFilterStrategy /JPEG /ColorACSImageDict << /QFactor 0.15 /HSamples [1 1 1 1] /VSamples [1 1 1 1] >> /ColorImageDict << /QFactor 0.15 /HSamples [1 1 1 1] /VSamples [1 1 1 1] >> /JPEG2000ColorACSImageDict << /TileWidth 256 /TileHeight 256 /Quality 30 >> /JPEG2000ColorImageDict << /TileWidth 256 /TileHeight 256 /Quality 30 >> /AntiAliasGrayImages false /CropGrayImages true /GrayImageMinResolution 150 /GrayImageMinResolutionPolicy /OK /DownsampleGrayImages true /GrayImageDownsampleType /Bicubic /GrayImageResolution 300 /GrayImageDepth -1 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 1.10000 /EncodeGrayImages true /GrayImageFilter /DCTEncode /AutoFilterGrayImages true /GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict << /QFactor 0.15 /HSamples [1 1 1 1] /VSamples [1 1 1 1] >> /GrayImageDict << /QFactor 0.15 /HSamples [1 1 1 1] /VSamples [1 1 1 1] >> /JPEG2000GrayACSImageDict << /TileWidth 256 /TileHeight 256 /Quality 30 >> /JPEG2000GrayImageDict << /TileWidth 256 /TileHeight 256 /Quality 30 >> /AntiAliasMonoImages false /CropMonoImages true /MonoImageMinResolution 1200 /MonoImageMinResolutionPolicy /OK /DownsampleMonoImages true /MonoImageDownsampleType /Bicubic /MonoImageResolution 1200 /MonoImageDepth -1 /MonoImageDownsampleThreshold 1.08250 /EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode /MonoImageDict << /K -1 >> /AllowPSXObjects false /CheckCompliance [ /None ] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile (None) /PDFXOutputConditionIdentifier () /PDFXOutputCondition () /PDFXRegistryName (http://www.color.org) /PDFXTrapped /Unknown /CreateJDFFile false /SyntheticBoldness 1.000000 /Description << /ENU (Use these settings to create PDF documents with higher image resolution for high quality pre-press printing. The PDF documents can be opened with Acrobat and Reader 5.0 and later. These settings require font embedding.) /JPN /FRA /DEU /PTB /DAN /NLD /ESP /SUO /NOR /SVE /KOR /CHS /CHT /ITA >> >> setdistillerparams << /HWResolution [2400 2400] /PageSize [595.000 842.000] >> setpagedevice