Monitoring phosphorus in the tributaries of a deep lake from the perspective of the receiving water body

Deep lakes under climate change are experiencing reduced deep mixing, with a consequent increase in nutrients concentration and oxygen deficit in the hypolimnion. From this perspective, greater efforts should be made to reduce the phosphorus load delivered to these lakes by their tributaries. An essential precondition for this action is a reliable estimate of the loads, which is challenging due to the marked temporal variability of the hydrologically-driven diffuse sources. In this article, we used a high-resolution phosphorus time series measured at the mouth of the main tributary of Lake Iseo and a machine learning algorithm to show the dominant role played by the acute, storm-dependent transport. The results emphasize the need to control loads strictly linked to precipitation and runoff in the drained watershed, representing 31% of the observed events but responsible for 64% of the overall load to the lake. We also proved evidence that the current obligations on nutrients monitoring miss the total phosphorus dynamics, leading to a systematic underestimation of the load conveyed by the inflows. Accordingly, we propose a sampling methodology where the timing and the methods for data interpolation are established according to the hydrological conditions in the drained watershed. Accounting for the economical and practical constraints imposed by the monitoring authorities, we propose to integrate, at the mouth of lake tributaries, a monthly manual sampling with an auto-sampler programmed to fill 2 bottles in correspondence of each high flow event. Our simulations showed that in this case the load estimation error was reduced below 1%, implying, on average, 13 field surveys/year and 21 laboratory analyses/year only. It is reasonable to expect comparable performances in hydrologically similar watersheds.