An inverse problem in population dynamics

DOI10.1080/01630569708816762MaRDI QIDQ4351392zbMATH OpenOpenAlexFDO

Publication date 26 November 1997

Published in Numerical Functional Analysis and Optimization (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1080/01630569708816762

continuous dependence results determination of delay kernels depending on time linear parabolic integrodifferential identification problems

Mathematics Subject Classification ID

Nonlinear parabolic equations (35K55) Inverse problems for PDEs (35R30) Integro-partial differential equations (45K05) Population dynamics (general) (92D25)

Review text

The author considers the following identification problem related to an interacting species model accounting for diffusion phenomena: determine a vector-valued function \({\mathbf u}=(u_1,u_2)\in C^1([0,T];L^p(\Omega;{\mathbb{R}})) \cap C([0,T];Y_2)\) and a matrix--valued function \(K=(k_{i,j})_{i,j=1,2}\) with \(k_{i,j}\in L^1((0,T);{\mathbb{R}})\) \((i,j=1,2)\) such that \[ u_i \geq 0,\quad \text{in\;}{\overline \Omega} \times [0,T],\;i=1,2 \qquad k_{i,j} \geq 0\quad \text{a.e.\;in\;} [0,T],\;i,j = 1,2, \] \[ D_t{\mathbf u} + A{\mathbf u}= {\mathbf F}({\mathbf u},K) \quad \text{in } {\overline \Omega} \times [0,T],\qquad u(0) = u_0, \] \[ \phi_j[u_i](t) := \int_{\Omega} u_i(x,t)\omega_j(x) dx = g_{i,j}(t) \qquad t \in [0,T],\;i,j = 1,2. \] Here \(A\) denotes the diffusion differential matrix defined by \[ A = \begin{pmatrix} -d_1\Delta & 0 \\ 0 & -d_2\Delta\end{pmatrix} \] with domain \[ m = \Biggl\{ {\mathbf z} \in W^{2,p}(\Omega;{\mathbb{R}}^m): {\partial {\mathbf z}\over \partial \nu} = 0\;\text{ on } \partial \Omega\Biggr\}, \] \(d_1\), \(d_2\) and \(\nu\) being, respectively, positive constants and the outward normal unit vector field to \(\partial \Omega\). The boundary condition in the definition of \(Y\) expresses the absence of interactions with the exterior. The rate function \({\mathbf F}\) is given by \[ {\mathbf F}({\mathbf u},K) := \Biggl( a_1-\sum_{s=1}^2 (b_{1,s}u_s+k_{1,s}*u_s)u_1, a_2-\sum_{s=1}^2 (b_{2,s}u_s+k_{2,s}*u_s)u_2 \Biggr)^\ast \] The coefficients \(a_i\), \(b_{i,i}\) and \(b_{i,j}\) \((i\neq j)\) are positive and nonnegative constants, respectively, while the symbol \(*\) stands for the time convolution product over \((0,t)\). Moreover, \(u_0 := (u_{0,1},u_{0,2})^\ast \in Y_2\) prescribes the initial population densities so that \(u_{0,i}\geq 0\) in \({\overline \Omega}\). Finally, \(w_j\in Y_1\) are suitable nonnegative functions and \(g_{i,j}\in W^{2,1}((0,T);(0,+\infty)\) are known. Under obvious consistency conditions the author proves that the solution \(({\mathbf u},K)\) to the above problem continuously depends on the data provided the following basic condition is satisfied: \[ \inf_{\text{supp}(\omega_1)} u_{0,1} \inf_{\text{supp}(\omega_2)} u_{0,2} - \sup_{\text{supp}(\omega_1)} u_{0,2} \sup_{\text{supp}(\omega_2)} u_{0,1} > 0 \] and we have an a-priori information of the form \[ \sum_{i,j}^2|k_{i,j}|_{L^1((0,T);{\mathbb{R}})} \leq N. \] Moreover, the latter assumption can be replaced by the following one involving the unknown \({\mathbf u}\): \[ \inf_{\text{supp}(\omega_j)} u_i(\cdot,t) \geq \beta \qquad \forall t\in [0,T],\;i,j=1,2, \] \(\beta\) being a positive constant.

Recommendations

Cites work

Cited in

(15)

This page was built for publication: An inverse problem in population dynamics

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4351392)