Several model inter-comparison projects (MIPs) have been carried out recently by the climate, hydrological, agricultural and other modelling communities to quantify modelling uncertainties and improve modelling systems. Here we focus on MIP design for large-scale water quality models. Water quality MIPs can be useful to improve our understanding of pollution problems and facilitate the development of harmonized estimates of current and future water quality. This can provide new opportunities for assessing robustness in estimates of water quality hotspots and trends, improve understanding of processes, pollution sources, water quality model uncertainties, and to identify priorities for water quality data collection and monitoring. Water quality MIP design should harmonize relevant model input datasets, use consistent spatial/temporal domains and resolutions, and similar output variables to improve understanding of water quality modelling uncertainties and provide harmonized water quality data that suit the needs of decision makers and other users.

Cryptosporidium is a leading cause of diarrhoea and infant mortality worldwide. A better understanding of the sources, fate and transport of Cryptosporidium via rivers is important for effective management of waterborne transmission, especially in the developing world. We present GloWPa-Crypto C1, the first global, spatially explicit model that computes Cryptosporidium concentrations in rivers, implemented on a 0.5 × 0.5° grid and monthly time step. To this end, we first modelled Cryptosporidium inputs to rivers from human faeces and animal manure. Next, we use modelled hydrology from a grid-based macroscale hydrological model (the Variable Infiltration Capacity model). Oocyst transport through the river network is modelled using a routing model, accounting for temperature- and solar radiation-dependent decay and sedimentation along the way. Monthly average oocyst concentrations are predicted to range from 10⁻⁶ to 10² oocysts L⁻¹ in most places. Critical regions (‘hotspots’) with high concentrations include densely populated areas in India, China, Pakistan and Bangladesh, Nigeria, Algeria and South Africa, Mexico, Venezuela and some coastal areas of Brazil, several countries in Western and Eastern Europe (incl. The UK, Belgium and Macedonia), and the Middle East. Point sources (human faeces) appears to be a more dominant source of pollution than diffuse sources (mainly animal manure) in most world regions. Validation shows that GloWPa-Crypto medians are mostly within the range of observed concentrations. The model generally produces concentrations that are 1.5–2 log10 higher than the observations. This is likely predominantly due to the absence of recovery efficiency of the observations, which are therefore likely too low. Goodness of fit statistics are reasonable. Sensitivity analysis showed that the model is most sensitive to changes in input oocyst loads. GloWPa-Crypto C1 paves the way for many new opportunities at the global scale, including scenario analysis to investigate the impact of global change and management options on oocysts concentrations in rivers, and risk analysis to investigate human health risk.