Spatial prediction of water quality variables along a main river channel, in presence of pollution hotspots.

In order to treat and evaluate the available data of water quality and fully exploit monitoring results (e.g. characterize regional patterns, optimize monitoring networks, infer conditions at unmonitored locations, etc.), it is crucial to develop improved and efficient methodologies. Accordingly, estimation of water quality along fluvial ecosystems is a frequent task in environment studies. In this work, a particular case of this problem is examined, namely, the estimation of water quality along a main stem of a large basin (where most anthropic activity takes place), from observational data measured along this river channel. We adapted topological kriging to this case, where each watershed contains all the watersheds of the upstream observed data ("nested support effect"). Data analysis was additionally extended by taking into account the upstream distance to the closest contamination hotspot as an external drift. We propose choosing the best estimation method by cross-validation. The methodological approach in spatial variability modeling may be used for optimizing the water quality monitoring of a given watercourse. The methodology presented is applied to 28 water quality variables measured along the Santiago River in Western Mexico.