Abstract
A new parametric class of covariance functions of wide-sense homogeneous and isotropic random fields on Euclidean spaces is presented. While the new class possesses uncoupled fractal and Hurst effects, it is shown to encompass the Generalized Cauchy and Dagum models. On the basis of the new functions, product-based covariance models are developed, along with parametric conditions guaranteeing a decoupling of fractal and Hurst effects.
Similar content being viewed by others
References
Mateu, J., Porcu, E., Nicolis, O.: A note on decoupling of local and global behaviours for the Dagum random field. Probab. Eng. Mech. 22(4), 320–329 (2007). https://doi.org/10.1016/j.probengmech.2007.05.002
Stein, M.: Statistical Interpolation of Spatial Data: Some Theory for Kriging. Springer, New York (1999). https://doi.org/10.1007/978-1-4612-1494-6
Shen, L., Ostoja-Starzewski, M., Porcu, E.: Harmonic oscillator driven by random processes having fractal and Hurst effects. Acta Mech. 226(11), 3653–3672 (2015). https://doi.org/10.1007/s00707-015-1385-4
Shen, L., Ostoja-Starzewski, M., Porcu, E.: Bernoulli-Euler beams with random field properties under random field loads: fractal and Hurst effects. Arch. Appl. Mech. 84(9), 1595–1626 (2014). https://doi.org/10.1007/s00419-014-0904-4
Laudani, R., Ostoja-Starzewski, M.: Fracture of beams with random field properties: fractal and Hurst effects. Int. J. Solids Struct. 191, 243–253 (2020). https://doi.org/10.1016/j.ijsolstr.2019.12.002
Nishawala, V.V., Ostoja-Starzewski, M., Leamy, M.J., Porcu, E.: Lamb’s problem on random mass density fields with fractal and Hurst effects. Proc. Math. Phys. Eng. 472(2196), 20160638 (2016). https://doi.org/10.1098/rspa.2016.0638
Nishawala, V.V., Ostoja-Starzewski, M.: Acceleration waves on random fields with fractal and Hurst effects. Wave Motion 74, 134–150 (2017). https://doi.org/10.1016/j.wavemoti.2017.07.004
Laudani, R., Zhang, D., Faouzi, T., Porcu, E., Ostoja-Starzewski, M., Chamorro, L.: On streamwise velocity spectra model accounting for the fractal and long-memory effects. Phys. Fluids 33, 035116 (2017). https://doi.org/10.1063/5.0040453
Jetti, Y.S., Ostoja-Starzewski, M.: Scaling in anti-plane elasticity on random shear modulus fields with fractal and Hurst effects. Fractal Fract. 5(4), 255 (2021). https://doi.org/10.3390/fractalfract5040255
Ostoja-Starzewski, M., Laudani, R.: Violations of the Clausius–Duhem inequality in Couette flows of granular media. Proc. Math. Phys. Eng. 476, 20200207 (2021). https://doi.org/10.1098/rspa.2020.0207
Jetti, Y.S., Ostoja-Starzewski, M.: Elastic contact of random surfaces with fractal and Hurst effects. Proc. R. Soc. A. 478(2268), 20220384 (2022). https://doi.org/10.1098/rspa.2022.0384
Malyarenko, A., Ostoja-Starzewski, M.: Tensor-Valued Random Fields for Continuum Physics. Cambridge University Press, UK (2019). https://doi.org/10.1017/9781108555401
Malyarenko, A., Ostoja-Starzewski, M.: Tensor- and spinor-valued random fields with applications to continuum physics and cosmology. Probab. Surv. 20, 1–86 (2023)
Malyarenko, A., Ostoja-Starzewski, M.: Polyadic random fields. ZAMP 73, 204 (2022)
Gneiting, T., Schlather, M.: Stochastic models that separate fractal dimension and the Hurst effect. SIAM Rev. 46(2), 269–282 (2004)
Berg, C., Mateu, J., Porcu, E.: The Dagum family of isotropic correlation functions. Bernoulli 14(4), 1134–1149 (2008). https://doi.org/10.3150/08-BEJ139
Zhang, X., Malyarenko, A., Porcu, E., Ostoja-Starzewski, M.: Elastodynamic problem on tensor random fields with fractal and Hurst effects. Meccanica 57(4), 957–970 (2022). https://doi.org/10.1007/s11012-021-01424-1
Schoenberg, I.J.: Positive definite functions on spheres. Duke Math. J. 9, 96–108 (1942). https://doi.org/10.1215/S0012-7094-42-00908-6
Daley, D.J., Porcu, E.: Dimension walks and Schoenberg spectral measures. Proc. Am. Math. Soc. 142(5), 1813–1824 (2014). https://doi.org/10.1090/S0002-9939-2014-11894-6
P’olya, G.: Remarks on characteristic functions. In: Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability, pp. 115–123 (1949). https://www.jstor.org/stable/2235784
Berg, C.: Stieltjes–Pick–Bernstein–Schoenberg and their connection to complete monotonicity. In: Positive Definite Functions: From Schoenberg to Space-Time Challenges, pp. 15–45 (2008)
Schilling, R.L., Song, R., Vondracek, Z.: Bernstein functions. In: Bernstein functions, de Gruyter, Berlin (2012) https://doi.org/10.1515/9783110269338
Gneiting, T.: Criteria of P’olya type for radial positive definite functions. Proc. Am. Math. Soc. 129(8), 2309–2318 (2001). https://doi.org/10.1090/S0002-9939-01-05839-7
Porcu, E., Mateu, J., Zini, A., Pini, R.: Modelling spatio-temporal data: a new variogram and covariance structure proposal. Stat. Probab. Lett. 77(1), 83–89 (2007). https://doi.org/10.1016/j.spl.2006.05.013
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Competing interests
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix: Proofs
Appendix: Proofs
Some background is necessary for a self-contained exposition. A function \(f:[0,\infty ) \rightarrow {\mathbb {R}}\) is called a Bernstein function [21] if its first derivative is completely monotonic.
Proof of Proposition 3.1:
According to the Schoenberg theorem [18], a candidate mapping \(\varphi : [0,\infty ) \rightarrow {\mathbb {R}}\) belongs to the class \(\varPhi _{\infty }\) if and only if the mapping \(0 \le x \mapsto \varphi (x)\) is completely monotonic. Hence, our assertion is proved by a direct inspection that allows to check for complete monotonicity. To provide a proof based on direct construction, we define
and
Direct inspection shows that the function \(\psi _{\alpha ,\beta }(\cdot )\) can be expressed as follows:
We now invoke the fact that completely monotone functions are a convex cone that is closed under pointwise convergence [21]. Hence, the product of two completely monotone functions is still completely monotonic. We now observe that the function \(\theta _{\beta }(\cdot )\) is completely monotonic for \(\beta \ge 0\). Hence, we need prove that \(\varphi _{\alpha }^{-\beta }(\cdot )\) is completely monotonic, implying that \(\psi _{\alpha ,\beta }(x)\) is completely monotonic as well.
For a completely monotonic function, g and a Bernstein function, f, the composition \(g\circ f\) is a completely monotonic function [21]. Direct inspection shows that the function \(g(z):=z^{-\beta }\) is completely monotonic for \(\beta \ge 0\). We now show that \(\varphi _{\alpha }(\cdot )\) is a Bernstein function, and this fact will complete the proof.
Consider the following function:
For \(0<\alpha <1\), \(\rho _{\alpha }(x)\) is a complete Bernstein function ([22], p. 220). Now, consider the following expression:
Here, \(\varphi _{\alpha } (x) = 1+\rho _{\alpha }(x) > 0\) is a Bernstein function for \(0<\alpha <1\) as \(\rho _{\alpha }(x)\) is a Bernstein function for \(0<\alpha <1\). It can be easily verified that \(\psi _{1,\beta }(x)=\theta _{\beta }(x)\) and therefore \(\psi _{\alpha ,\beta }\) is completely monotonic for the case \(\alpha =1\) as well. \(\square \)
Proof of Positive definiteness of \(\psi _{\alpha ,\beta }\) in \({\mathbb {R}}^3\) The sufficient conditions for the positive definiteness can be obtained by applying Pólya theorem [20] for radial functions and the corresponding extension to \({\mathbb {R}}^n\) as given in [23]. Define the following function:
for \(x>0\) and k a nonnegative integer. From Gneiting’s theorem, if there exists \(m \ge \frac{1}{2}\) such that
is convex for \(x>0\) and l a nonnegative integer, then \(\psi \) is positive definite on \({\mathbb {R}}^n\). The above-mentioned arguments are taken from [24]. By setting \(k=0\), \(l=1\), and \(m=1\), if we can show that
is convex, then (3.1) is valid on \({\mathbb {R}}^3\). To show this, we take the second derivative of Eq. (A.1), which has the following expression:
Here,
\(f_{\alpha ,\beta }^{\prime \prime }(x)\) has to be positive for \(f_{\alpha ,\beta }(x)\) to be convex. It can be easily shown that \(f_{\alpha ,\beta }(x)\) is positive when \(\alpha >0\) and \(\beta >0\). For \(f_{\alpha ,\beta }^{\prime \prime }(x)\) to be positive, the expression,
has to be positive when \(\alpha >0\) and \(\beta >0\).
The expression \(Q(x,\alpha ,\beta )\) is quadratic in x with a positive coefficient of \(x^2\). \(Q(x,\alpha ,\beta )\) is always positive if its discriminant is negative.
Similarly, the expression \(R(x,\alpha ,\beta )\) is quadratic in x with a positive coefficient of \(x^2\). \(R(x,\alpha ,\beta )\) is always positive if its discriminant is negative.
The expression \(P(x,\alpha ,\beta )\) as well is quadratic in x, however, with a negative coefficient of \(x^2\). So, \(P(x,\alpha ,\beta )\) can either be completely negative or mostly negative.
Consider \(S(x,\alpha ,\beta )\) when \(0<\alpha <1\) and \(0<\alpha \beta <1\) as \(x\rightarrow 0\) and as \(x\rightarrow \infty \)
Therefore, consider \(S(x,\alpha ,\beta )\) when \(0<\alpha <1\) and \(0<\beta <1\):
The expression \(T(x,\alpha ,\beta )\) is much simpler compared to \(S(x,\alpha ,\beta )\). It is minimized using Wolfram Mathematica with the constrains \(\{0<\alpha<1; 0<\beta <1\}\).
\(\square \)
Proof of Proposition 3.2:
Consider the following volume integral for the case \(n=3\) over \(V\in {\mathbb {R}}^3\) to check \(L^1-\) integrability,
For \(R\le 1\)
For \(R> 1\),
By similar computations, it can be shown that \(\psi _{\alpha ,\beta }(x)\) is \(L^1-\) and \(L^2-\) integrable, respectively, under the conditions \(\beta > n\) and \(\beta >n/2\) for \(n=1,2,3\). \(\square \)
Proof of Proposition 3.3:
To estimate the fractal dimension for \(\psi _{\alpha ,\beta }\), consider the following:
Setting \(\delta = \alpha \) implies
Therefore, the fractal dimension is identically equal to \(D = n+1-\alpha /2\). To estimate the Hurst parameter for the above covariance function, consider the following:
Consider \(\omega = \beta \), this implies that
Therefore, the Hurst parameter is \(H = 1-\beta /2\). \(\square \)
Proof of Proposition 3.6:
To obtain the fractal dimension for the MOSP model corresponding to \(\psi _{\alpha ,\beta ,\gamma }\) as in Eq. (3.2), consider the following
Setting \(\delta = \gamma _1\alpha _1 \le \gamma _2\alpha _2\) implies
Therefore, the fractal dimension is identically equal to \(D=n+1-\text {min}(\gamma _1\alpha _1,\gamma _2\alpha _2)/2\). To obtain the Hurst parameter, consider the following:
Considering \(\omega _1 = \gamma _1\beta _1\) and \(\omega _2 = \gamma _2\beta _2\), we get
Therefore, the Hurst parameter is \(H = 1 - (\gamma _1\beta _1 + \gamma _2\beta _2)/2\). \(\square \)
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Jetti, Y.S., Porcu, E. & Ostoja-Starzewski, M. New decouplers of fractal dimension and Hurst effects. Z. Angew. Math. Phys. 74, 123 (2023). https://doi.org/10.1007/s00033-023-02010-z
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00033-023-02010-z