This project investigates the spatial and temporal characteristics of GNSS-estimated tropospheric gradients, with a focus on evaluating their consistency, sensitivity to processing strategies, and utility for meteorological applications. Using the SWEPOS GNSS network in southern Sweden (~300 stations), we analyze one year of high-precision GNSS data processed with the PRIDE-PPPAR software to estimate Zenith Total Delay (ZTD), horizontal gradients, and associated uncertainties. We aim to explore several key questions: (1) what is the spatial scale over which GNSS gradients remain correlated; (2) how reliable are gradient estimates compared to the more robust ZTD estimates; (3) how do different processing parameters (e.g., elevation angle cut-off, stochastic constraint levels) affect gradient magnitude and consistency; and (4) can GNSS-estimated gradients be validated against delay differences derived from nearby station ZTDs? To address these questions, we implement a large-scale processing pipeline in which GNSS data from hundreds of stations are processed using a variety of parameter strategies. The results are compared pairwise across baselines of varying lengths and directions to assess gradient correlation decay with distance, as well as agreement between GNSS-derived and ZTD-inferred gradient components. We apply additional atmospheric modeling (e.g., VMF3 hydrostatic/wet separation, ERA5-derived refractivity profiles) to isolate the wet component and determine its vertical structure. This study contributes to identifying best practices for reliable GNSS gradient estimation and assessing their physical consistency, with potential applications in real-time weather model assimilation and severe weather forecasting. The data volume (10k rinex data, about 2 000 GiB) and number of processing strategies necessitate high-throughput computation, motivating the use of C3SE resources for parallel execution and storage.