This paper introduces a novel adaptation of the Unsupervised Image-to-Image Translation (UNIT) framework, specifically designed to address the challenges of translating complex climate data across domains. Traditional UNIT models often fail to handle climate datasets due to their intricate geospatial constraints and multi-variable nature. To overcome this limitation, the proposed approach integrates land masks to maintain geospatial specificity, accurately representing geographical boundaries such as coastlines. We propose a flexible loss calculation strategy incorporating diverse reconstruction loss functions, including Frechet Inception Distance (FID), Structural Similarity Index (SSIM), and Learned Perceptual Image Patch Similarity (LPIPS) functions. This combination allows the model to better accommodate the diverse characteristics of climate data, enhancing accuracy and feature preservation. Furthermore, the framework introduces a mechanism for managing multi-variable datasets by selectively applying different loss functions to individual data channels, effectively balancing the importance of each variable. Experimental results demonstrate that the adapted UNIT framework significantly outperforms traditional methods, achieving superior geospatial feature preservation and more precise climate data translation. This advancement in machine learning provides a robust tool for environmental analysis, contributing to more reliable climate modelling and prediction.