A critical decision process in data acquisition for mineral and energy resource exploration is how to efficiently combine a variety of sensor types and to minimize total cost. We propose a probabilistic framework for multi-objective optimisation and inverse problems given an expensive cost function for allocating new measurements. This new method is devised to jointly solve multi-linear forward models of 2D-sensor data and 3D-geophysical properties using sparse Gaussian Process kernels while taking into account the cross-variances of different parameters. Multiple optimisation strategies are tested and evaluated on a set of synthetic and real geophysical data. We demonstrate the advantages on a specific example of a joint inverse problem, recommending where to place new drill-core measurements given 2D gravity and magnetic sensor data, the same approach can be applied to a variety of remote sensing problems with linear forward models - ranging from constraints limiting surface access for data acquisition to adaptive multi-sensor positioning.