This paper is mainly concerned with the generalised principal eigenvalue for timeperiodic nonlocal dispersal operators. We first establish the equivalence between two different characterisations of the generalised principal eigenvalue. We further investigate the dependence of the generalised principal eigenvalue on the frequency, the dispersal rate and the dispersal spread. Finally, these qualitative results for time-periodic linear operators are applied to time-periodic nonlinear KPP equations with nonlocal dispersal, focusing on the effects of the frequency, the dispersal rate and the dispersal spread on the existence and stability of positive time-periodic solutions to nonlinear equations.where (x, t) ∈Ω× R, Ω ⊂ R N is a bounded domain, τ > 0 is the frequency, µ > 0 is the dispersal rate, σ > 0 is the dispersal spread which characterises the dispersal range, m ≥ 0 is the cost parameter, J σ (·) = 1 σ N J( · σ ) is the scaled dispersal kernel. Throughout the paper, we will make the following assumptions on the dispersal kernel J, a family of functions {h σ } σ>0 and function a:Define the spaces χ Ω , χ + Ω , χ ++ Ω as follows:where C 0,1 (Ω × R) denotes the class of functions that are continuous in x and C 1 in t. The operator L τ,µ,σ,m Ω is then considered as an unbounded linear operator on the space C 1 (Ω × R) with domain χ Ω , namely,It may be worthwhile to point out that the time-periodic nonlocal dispersal operators of the form (1.1) include several kinds of boundary conditions, such as Dirichlet, Neumann and mixed type; See [7,28,29].The principal eigenvalues for time-periodic nonlocal dispersal operators have been studied in [2, 19, 29-31, 33, 36, 40]. In this paper, we adopt the approaches as in Berestycki, Nirenberg and Varadhan [6] and Berestycki, Coville and Vo [4] for the definition of the generalised principal eigenvalue λ p (L τ,µ,σ,m Ω ):TIME PERIODIC NONLOCAL DISPERSAL OPERATORS AND APPLICATIONS 3 Another definition for the generalised principal eigenvalue of L τ,µ,σ,m Ω is given by λ ′ p (L τ,µ,σ,m Ω ) := inf{λ ∈ R | ∃ v ∈ χ ++ Ω s.t. (L τ,µ,σ,m Ω + λ)[v] ≥ 0 inΩ × R}, motivated by the works of Donsker and Varadhan in [13] and Berestycki, Coville and Vo in [4].Shen and Vo proved in [36] that λ p = λ ′ p when λ p is the principal eigenvalue and h σ (x) ≡ 1 for all σ > 0. Our first main result proves that for general h σ , λ p = λ ′ p always holds, and λ p can be characterised as the infimum of the spectrum of −L τ,µ,σ,m Ω , whether λ p is an eigenvalue or not. Theorem 1.1. Assume that (J), (H) and (A) hold. Then λ p (L τ,µ,σ,m Ω ) = λ ′ p (L τ,µ,σ,m Ω ) = λ 1 , where λ 1 = inf{Reλ | λ ∈ σ(−L τ,µ,σ,m Ω )} and σ(−L τ,µ,σ,m Ω ) is the spectrum of −L τ,µ,σ,m Ω . Next, we turn to study the influences of the frequency τ , the dispersal rate µ and the dispersal spread σ on the generalised principal eigenvalue λ p (L τ,µ,σ,m Ω ). The following result establishes the monotonicity and asymptotic behaviors of the generalised principal eigenvalue λ p (L τ,µ,σ,m Ω ) with respect to the frequency τ : Theorem...