In the realm of geographic data, implicit neural representations (INRs) often incorporate location embeddings to enhance performance in downstream tasks. Recent advancements in this field have introduced location encodings based on Fourier domain decomposition. However, given the localized nature of many geographic signals, these approaches may introduce significant biases against certain geographic subgroups. To this end, we propose an alternative encoding mechanism, lifting the theoretical guarantees of spherical wavelets. Leveraging the FAIR-Earth dataset, we demonstrate that our novel encoding successfully mitigates biases in localized regions while simultaneously maintaining competitive global performance. Our approach represents a significant step forward in creating more equitable and accurate INRs for geographic data.