Abstract
In practical applications of area-to-point spatial interpolation, inequality constraints, such as non-negativity or more general constraints on the maximum and/or minimum attribute value, should be taken into account. The geostatistical framework proposed in this paper deals with the spatial interpolation problem of downscaling areal data under such constraints, while: (1) explicitly accounting for support differences between sample data and unknown values, (2) guaranteeing coherent (mass-preserving) predictions, and (3) providing a measure of reliability (uncertainty) for the resulting predictions. The formal equivalence between Kriging and spline interpolation allows solving constrained area-to-point interpolation problems via quadratic programming (QP) algorithms, after accounting for the support differences between various constraints involved in the problem formulation. In addition, if inequality constraints are enforced on the entire set of points discretizing the study domain, the numerical algorithms for QP problems are applied only to selected locations where the corresponding predictions violate such constraints. The application of the proposed method of area-to-point spatial interpolation with inequality constraints in one and two dimension is demonstrated using realistically simulated data.
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs10109-006-0036-7/MediaObjects/10109_2006_36_Fig1_HTML.gif)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs10109-006-0036-7/MediaObjects/10109_2006_36_Fig2_HTML.gif)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs10109-006-0036-7/MediaObjects/10109_2006_36_Fig3_HTML.gif)
![](http://media.springernature.com/m312/springer-static/image/art%3A10.1007%2Fs10109-006-0036-7/MediaObjects/10109_2006_36_Fig4_HTML.gif)
Similar content being viewed by others
References
Arbia G (1989) Spatial data configuration in statistical analysis of regional economic and related problems. Kluwer, Dordrecht
Barnes RJ, Johnson TB (1984) Positive Kriging. In: Verly G et al (eds) Geostatistics for natural resources characterization, Vol. 2. D. Reidel Publishing, Dordrecht, pp 231–244
Barnes RJ, You K (1992) Adding bounds to Kriging. Math Geol 24(2):171–176
Chilès JP, Delfiner P (1999) Geostatistics: modeling spatial uncertainty. Wiley, New York
Cressie NAC (1990) Reply to comment on Cressie’s geostatistics. Am Stat 44(3):256–258
Cressie NAC (1993) Statistics for spatial data, Revised edn. Wiley, New York
Deutsch CV (1996) Correcting for negative weights in Ordinary Kriging. Comput Geosci 22(7):765–773
Deutsch CV, Journel AG (1998) GSLIB: geostatistical software library and user’s guide, 2nd edn. Oxford University Press, New York
Dubrule O (1983) Two methods with different objectives: splines and Kriging. Math Geol 15(2):245–257
Dubrule O, Kostov C (1986) An interpolation method taking into account inequality constraints: I Methodology. Math Geol 18(1):33–51
Duchon J (1977) Splines minimizing rotation-invariant semi-norms in Sobolev spaces. In: Constructive theory of functions of several variables. Springer, Berlin Heidelberg New York, pp 85–100
Dyn N, Wong WH (1987) On the characterization of non-negative volume-matching surface splines. J Approx 51(1):1–10
Galli S, Murillo E, Thomann J (1984) Dual Kriging-It’s properties and its uses in direct contouring. In: Verly G et al. (eds) Geostatistics for natural resources characterization, Vol. 2. D. Reidel Publishing, Dordrecht, pp 621–634
Gill P, Murray W, Wright M (1981) Practical optimization. Academic, London
Goovaerts P (1997) Geostatistics for natural resources evaluation. Oxford University Press, New York
Hillier FS, Lieberman GJ (2001) Introduction to operations research, 7th edn. McGraw-Hill, New York
Hutchinson MF, Gessler PE (1994) Splines-more than just a smooth interpolator. Geoderma 62:45–67
Isaaks E, Srivastava R (1989) An introduction to applied geostatistics. Oxford University Press, New York
Journel AG (1986) Constrained interpolation and qualitative information. Math Geol 18(3):269–286
Journel AG, Huijbregts CJ (1978) Mining geostatistics. Academic, New York
Kostov C, Dubrule O (1986) An interpolation method taking into account inequality constraints: II. Practical approach. Math Geol 18(1):53–73
Künzi HP, Tzschach HG, Zehnder CA (1971) Numerical methods of mathematical optimization. Academic, Dordrecht
Kyriakidis PC (2004) A geostatistical framework for the area-to-point spatial interpolation. Geograph Anal 36(3):41–50
Kyriakidis PC, Yoo E-H (2005) Geostatistical prediction and simulation of point values from areal data. Geograph Anal 37(2):124–151
Limic N, Mikelic A (1984) Constrained Kriging using Quadratic Programming. Math Geol 16(4):423–429
Luenberger DG (1969) Optimization by vector space methods. Wiley, New York
Martin D, Tate NJ, Langford M (2000) Refining population surface models: experiments with Northern Ireland census data. Trans GIS 4(4):343–360
Matheron G (1971) The theory of regionalized variables and its applications. Fasc. 5, Ecole Nat. Sup. des Mines, Paris
Matheron G (1981) Splines and Kriging: their formal equivalence. In: Down-to-Earth Statistics: Solutions Looking for Geological Problems. Syracuse University, Geology Contributions, pp 77–95
Meyer CD (2000) Matrix analysis and applied linear algebra. Society for Industrial and Applied Mathematics (SIAM), Philadelphia
Michalak AM, Kitanidis PK (2005) A method for the interpolation of nonnegative functions with an application to contaminant load estimation. Stoch Environ Res Risk Assess 19:8–23
Nagle N (2005) Continuous field statistical methods for spatial analysis in the social sciences. Ph.D. dissertation. Department of Geography, University of California Santa Barbara, Santa Barbara
Olea RA (ed) (1999) Geostatistics for Engineers and Earth Scientists. Kluwer, Dordrecht
Pardo-Iguzquiza E, Dowd PA (2005) Empirical maximum-likelihood Kriging: the general case. Math Geol 37(5):477–492
Prenter PM (1975) Splines and variational methods. Wiley, New York
Royer JJ, Vieira PC (1984) Dual formalism of Kriging. In: Verly G et al (eds) Geostatistics for natural resources characterization, Vol 2. D. Reidel Publishing, Dordrecht, pp 691–702
Saito H, Goovaerts P (2000) Geostatistical interpolation of positively skewed and censored data in a dioxin-contaminated site. Environ Sci Technol 34(19):4228–4235
Szidarovszky F, Baafi EY, Kim YC (1987) Kriging without negative weights. Math Geol 19(6):549–560
Tobler WR (1979) Smooth pychnophylactic interpolation for geographical regions. J Am Stat Assoc 74(367):519–530
Villalobos M, Wahba G (1987) Inequality-constrained multivariate smoothing splines with application to the estimation of posterior probabilities. J Am Stat Assoc 82(397):239–248
Wackernagel H (2003) Multivariate Geostatistics: An Introduction with Applications. Springer, Berlin Heidelberg New York
Wahba G (1990a) Comment on Cressie’s Geostatistics. Am Stat 44(3):255–256
Wahba G (1990b) Spline models for observational data. In: CBMS-Regional conference series in Applied Mathematics, Society for Industrial and Applied Mathematics (SIAM), Philadelphia
Walvoort DJJ, de Gruijter JJ (2001) Compositional Kriging: a spatial interpolation method for compositional data. Math Geol 33(8):951–966
Wong WH (1980) An analysis of the volume-matching problem and related topics in smooth density estimation. Ph.D. dissertation, Department of Statistics, University of Wisconsin-Madison, Wisconsin
Acknowledgments
Financial support for this study was provided by the National Science Foundation Grant BCS \(\sharp 0352879.\) The second author would also like to acknowledge partial funding provided by the National Geospatial-Intelligence Agency under the project Strategic Enhancement of NGA’s Geographic Information Science Infrastructure.
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix
Derivation of bounded ordinary Kriging (BOK) variance
Let C be a ((K + 2) × (K + 2)) matrix that is partitioned as
where \({\bf C}_{11} = \left[\begin{array}{ll} {\bf C}_{ss} & {\bf 1} \\ {\bf 1}^{\rm T} & 0 \end{array}\right]\) \({\bf c}_{12} = \left[\begin{array}{l}{\bf z}_s \\ 0 \end{array}\right]\) c 21 = [z T s 0], and c 22 = [0].
Consider now a ((K + 2) × 1) vector v containing a (K × 1) vector of covariances c s p between areal data and unknowns, the sum of weights 1, and the lower bound value z l . The vector v can also be partitioned into two vectors v 0 = [[c s p ]T 1]T and z l , as
We assume that the inverse of C exists, and is denoted as \({\bf B} = {\bf C}^{-1} = \left[\begin{array}{ll} {\bf B}_{11} & {\bf b}_{12} \\ {\bf b}_{21} & {\bf b}_{22} \end{array}\right].\) Then, the BOK variance of Eq. 15a can be re-written as
According to the Partitioned Matrix Lemma (Meyer 2000), matrix B contains the following entries, provided that both C 11 and b 22 are nonsingular:
-
b 22 = [c 22 − c 21 C −111 c 12]−1 = [0 − c 21 C −111 c 12]−1 = −[c 21 C −111 c 12]−1
-
b 12 = −C −111 c 12[c 22 − c 21 C −111 c 12]−1 = −C −111 c 12 b 22
-
b 21 = − b 22 c 21 C −111
-
B 11 = C −111 + C −111 c 12 b 22 c 21 C −111
Let \(\iota = {\bf c}_{21}{\bf C}_{11}^{-1}{\bf c}_{12}\) and recall that the predicted value at location u p is expressed as \(\hat z({\bf u}_p) = {\bf V}_0^{\rm T}{\bf C}_{11}^{-1}{\bf c}_{12}.\) The BOK variance of Eq. 36 then becomes:
where \({\hat \sigma}_{OK}^2({\bf u}_p)\) denotes the unconstrained OK variance at location u p .
Dual formalism of Kriging with a trend model
Universal Kriging (Matheron 1971) or Kriging with a trend (Goovaerts 1997; Deutsch and Journel 1998), is a general type of Kriging where the mean (trend component) of the random function varies in space. That trend component (or drift) is often modeled as a smoothly varying parametric function of the coordinates vector u:
where f m (·) is a deterministic function known at any point u, and β m denotes its mth unknown parameter; there are (M + 1) such unknown parameters, which are determined by the data.
The other part of the Universal Kriging (UK) model is a second-order stationary residual, characterized by a zero mean and covariance function C R (h). The dual UK prediction at location u p using the K areal data can thus be written as
with:
where w denotes a (K × 1) vector of coefficients of the stochastic interpolator, f p = [f m (u p ), m = 0, . . . , M]T denotes a ((M + 1) × 1) vector of deterministic function values at location u p , and \({\varvec{\upbeta}} = [\beta_m, m = 0,\ldots,M]^{\rm T}\) denotes a ((M + 1) × 1) vector of trend coefficients. Note that the covariance vector c s p in the above equation is derived by regularization from the point residual covariance C R (h).
In the context of interpolation, area-to-point Dual Kriging with a trend model can be viewed as the following problem: “find the coefficients w and \({\varvec{\upbeta}}\) in Eq. 38, satisfying the K areal data reproduction constraints and the (M + 1) unbiasedness constraints”:
where |a| denotes the elementary area represented by the pth prediction location, u p ; this is used for numerical integration purposes. Term f m (s k ) denotes deterministic function values associated with the k-th areal support at location u p .
The vector w of weights and the vector \({\varvec{\upbeta}}\) of trend coefficients in Eq. 38 are derived by solving the following system of normal equations:
where F = [f 0 ... f m ... f M] is a (K × (M + 1)) matrix of (M + 1) low-order drift functionals, with f m = [f m (s k ), k = 1, . . . , K]T. Note that the covariance matrix C R ss in the above equation is again derived by regularization from the point residual covariance C R (h).
QP formalism of Kriging
4.1 A classical QP problem
QP problems involve the optimization of a quadratic objective function (maximization or minimization) subject to a number of constraints. More precisely, consider a quadratic minimization problem subject to the following linear inequality constraints:
where x is a (K × 1) vector of decision variables whose respective values need to be determined, and C and f denote, respectively, a (K × K) matrix and a (K × 1) vector with the mathematical expression (function) of the variable considered. To obtain a unique solution of the QP problem, it is assumed that C is a symmetric and positive semi-definite matrix. The Q l inequality constraints are imposed on the solution vector x by a (Q l × K) matrix A and the lower bound vector b; this forces the solution to be greater than the lower bound values.
The duality in QP problems often provides a convenient strategy for arriving at the same solution vector x. The dual form of Eq. 40 can be written as
where t is a (Q l × 1) vector of Lagrange multipliers associated with the Q l inequality constraints.
The above dual QP problem in Eq. 41 can be simplified based on the fact that x is a linear combination of t, i.e.: x = C −1(A T t − f). By simply replacing x by the function of t, the dual form of the minimization QP problem in Eq. 40 can be represented as the maximization problem:
which is again equivalent to the minimization problem:
The dual form of QP problems with equality constraints alone also leads to the the same minimization problem as Eq. 42. The only difference between QP problems with equality and those with inequality constraints is that there are no positive constraints on the Lagrange multipliers in the former case.
4.2 QP problem formulation of CDOK
In Sect. 2.2.3, we showed that CDOK is equivalent to a constrained interpolator, which is determined by its coefficients; these coefficients are obtained by solving a quadratic minimization problem subject to linear constraints. In what follows, we will convert the dual form of area-to-point Kriging under constraints into a simplified quadratic programming problem. To keep the problem simple, and without loss of generality, we consider the task of predicting the unknown value at location u p using the K areal data z s = [z(s 1), . . . , z(s K )]T under lower bound inequality constraints only.
The primal QP problem is written in this case as
where \({\varvec{\upomega}}\) is a ((K + Q l ) × 1) vector of weights applied to the K areal data (associated with equality constraints) and the Q l prediction points (associated with inequality constraints), and \({\varvec{\Upsigma}}\) is a ((K + Q l ) × (K + Q l )) matrix of covariances among all constraints. A s denotes the (K × (K + Q l )) matrix of equality constraints, and \({\varvec{\upmu}}_s = {\bf G}{\bf 1}_P \beta_0\) and \({\varvec{\upmu}}_l = {\bf 1}_{Q_l} \beta_0\) denote, respectively, a (K × 1) and and (Q l × 1) vector of mean values for the areal data RVs and the lower bound RVs. A l denotes the (Q l × (K + Q l )) matrix of inequality constraints, and z l represents the lower bound values at Q l point locations.
The dual form of the constrained DOK problem of Eq. 43 is then written as
where the objective function is a maximization problem which involves a ((K + Q l ) × 1) vector \({\varvec{\uppsi}} = [{\varvec{\uppsi}}_s^{\rm T} \quad {\varvec{\uppsi}}_l^{\rm T}]^{\rm T}\) of Lagrange multipliers, as well as data and bound values z = [z T s z T l ]T and their means \({\varvec{\upmu}} = [{\varvec{\upmu}}_s^{\rm T} \quad {\varvec{\upmu}}_l^{\rm T}]^{\rm T}.\) The new constraints in Eq. 45 for the dual form of the objective function combine both equality and inequality constraints in the primary form of Eq. 43. The constraints matrix A is written as A = [A T s A T l ]T, where A s is a (K × (K + Q l )) matrix of equality constraints and A l is a (Q l × (K + Q l )) matrix of inequality constraints. The corresponding Lagrange multipliers for each constraint are denoted by \({\varvec{\uppsi}}_s\) and \({\varvec{\uppsi}}_l\) respectively. Thus, Eq. 45 is a combination of the following two equations:
As shown in the case of the classical QP problem, the constrained DOK problem can be simplified by the following two steps. First, convert the minimization problem to a maximization problem, and note that \({\varvec{\upomega}}\) is a function of \({\varvec{\uppsi}}\) written as \({\varvec{\upomega}} = {\varvec{\Upsigma}}^{-1}{{\bf A}^{\rm T}{\varvec{\uppsi}}}\) from Eq. 45. Therefore, the simplified version of the dual form of QP problems for constrained DOK can be reformulated as
By letting \({\bf Q}^* = ({\bf A}{\varvec{\Upsigma}}^{-1}{\bf A}^{\rm T})\) the simplified dual form of the CDOK maximization problem in Eq. 47 can be converted to the following minimization problem:
where Q * is a ((K + Q l ) × (K + Q l )) matrix of covariances involving K support-to-support covariances where equality constraints are assigned, and Q l point covariances where inequality constraints are imposed.
Rights and permissions
About this article
Cite this article
Yoo, EH., Kyriakidis, P.C. Area-to-point Kriging with inequality-type data. J Geograph Syst 8, 357–390 (2006). https://doi.org/10.1007/s10109-006-0036-7
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10109-006-0036-7