Unmanned Aerial Vehicles (UAVs) have been extensively utilized in environmental modeling tasks, including forest monitoring, agricultural mapping, and disaster assessment. These applications benefit from UAVs’ ability to collect high-resolution data for purposes such as deforestation detection, crop management, and damage evaluation. However, the common assumption that UAVs are homogeneous leads to significant inaccuracies due to differences in UAV platforms and sensor characteristics. This paper investigates the deployment of heterogeneous UAVs equipped with various sensors in fire-affected regions, modeled via a point of interest framework. The goal is to minimize measurement errors by explicitly leveraging UAV heterogeneity. We introduce a weighted frame potential metric based on matrix orthogonality, and develop a randomized greedy algorithm to optimize UAV positioning. The proposed algorithm guarantees at least a 50% approximation of the optimal solution and offers a theoretical error bound on the metric. Experimental results demonstrate that the proposed method outperforms joint greedy and determinant-based algorithms in reducing measurement errors.