A predictive geospatial approach for modelling phosphorus concentrations in rivers at the landscape scale

•Spatiality of phosphorus in rivers explored, accounting for environmental conditions.•Hydrological transportation and connectivity more influential than pollutant source.•Geospatial models better explain phosphorus variance than simple regression models.•Geospatial models can optimise management and monitoring for river water quality.

SummaryEnrichment by phosphorus (P) constitutes a significant pressure on river systems, and is one of the main causes of freshwater pollution globally. Catchment environmental conditions influence the timing and magnitude of P release and transfer to water bodies, and therefore can potentially provide a basis for identifying water bodies vulnerable to impairment by P and/or resistant to restoration efforts. The current research involved construction of a geospatial database, comprising monthly values for flow-weighted concentrations of molybdate reactive phosphorus (fwMRP) sampled in rivers from 2006 to 2008 together with spatially-expressed environmental data relating to 18 different variables for 54 catchments in the Republic of Ireland. A regression–kriging modelling methodology within a landscape-scale, geospatial approach was tested. Environmental conditions relating to hydrological transportation and connectivity (slope, degree of surface saturation, soil water content) were found to exert greater influence over concentrations of P in rivers than direct proxies of sources of P (e.g. human population level or land use). Geospatial models provided greater explanation of P variance than regression models (an improvement in predictive capability of up to 8.5%). Data for fwMRP were segregated sub-annually into two periods, one focused on summer and the other on winter months. A geospatial model for the period including winter months was found to have a better predictive capability than the one that centred upon the summer, with the latter routinely overestimating fwMRP when compared with observed (test) data. Geospatial models potentially provide a means of optimising monitoring regimes for river water quality, and can also be used as a screening tool to focus management and remediation measures where they are likely to prove most effective.