Soil water content (SWC) is a key variable in hydrology, agriculture, and climate research, but large-scale measurements remain challenging due to spatial heterogeneity and logistical limitations. Stationary Cosmic Ray Neutron Sensing (CRNS) provides intermediate-scale estimates (~200m footprint), yet covers only local areas. Mobile Rail-CRNS platforms overcome this by enabling continuous SWC mapping along hundreds of kilometers of railway networks. In 2024, the UFZ operated five such Rail-CRNS systems, collecting data up to hundredth of kilometer daily across diverse landscapes in Germany. However, rail roving multiplies exposure to dynamic environmental influences (e.g., tunnels, bridges, parallel tracks, urban areas, water bodies, roads, topography, biomass/forest types), which can systematically bias neutron signals. Further, inaccuracies in GPS positioning can cause the measurement positions to be several meters off the track. At this data volume, manual screening is infeasible, automated detection, flagging, and quantitative scoring of these influences are required for data quality control and correction.

Here we present a fully automated, Python-based pre-processing pipeline that evaluates measurements at both point and segment levels. GPS positions are first snapped to OSM railway tracks (nearest-points projection) to correct for localization errors. Each point is then queried for proximity to OSM features, tree species from the German Aerospace Center and DEM-derived topography, using configurable minimum feature sizes (e.g. length of a river, tunnel), influence radii, and weights (e.g., tunnel > bridge). These parameters can be flexibly adjusted and regionally adapted. To address the integral nature of mobile measurements, we introduce segment-based scoring: Intervals between consecutive points are subdivided into subsamples (minimum 3, additional every ~10 m for longer segments), incorporating direction (azimuth) for asymmetric effects (e.g., lateral slopes) guaranteeing its real length but its planar projection. Influences are evaluated proportionally. In addition, for segments above a defined length, a speed flag is added to indicate reduced data density and reliability.

An interactive map allows you to review the selected settings in relation to the potentially influencing features: Segment colors reflect its cumulative scores, flags as rings in relation to its cause, and geo-layers toggleable. Mouse-over tooltips provide instant score breakdowns for iterative parameter tuning.

The pipeline enables targeted filtering of uncertain segments, application of region- or forest-type-specific correction factors, and integrative comparison of land-use groups (point vs. segment scale). Initially tested on a pilot transect in the Harz Mountains (~ 8 km), ~60% were marked as having substantial impacts, demonstrating its necessity as well as its robustness and practical applicability. Fully transferable across Germany, it paves the way for consistent, large-scale Rail-CRNS SWC mapping. Future steps include machine-learning-based weight optimization.