Small satellite clustered systems are getting increasingly affordable as a sensing infrastructure for earth observation. New earth observation applications are being conceived and capability of existing ones gets augmented by such systems. A small satellites constellation flying in low earth orbit (LEO) offers high fidelity spatial and temporal data which has applications in various urban management tasks, like detecting and monitoring constructions, green cover, traffic management, etc. A region can potentially be imaged by multiple satellites, based on their tracks. Given a set of such regions, the ground control assigns and schedules collection tasks to the satellites in the constellation. Given multiple constraints related to kinematics of the satellites, windows of observations, and communication to ground station; the problem of observation task assignment is known to be computationally hard. Several optimization techniques are proposed in literature. However, these optimizations do not take into account the environmental disturbances which may occlude sensor view and render collected data useless for processing. Tolerance to such external disturbances can be mitigated through redundancy in schedule, but at the cost of generating and transferring larger data volume to ground stations. In this paper, we propose a randomized heuristic to induce redundancy to balance these two conflicting aspects.