Reconstructing Historical Changes in Watersheds from Environmental Records: An Information Theory Approach

Abstract:

A 20% of the world’s population is living in watersheds that suffer from water shortage. This situation has complex causes associated with historical changes in watersheds. However, disentangling the role of key drivers of water availability like climate change or land use practices is challenging. Part of the difficulty resides in that historical analysis is basically a process of empirical reconstruction from available environmental records (e.g. sediment cores or long-term hydrologic time series). We developed a mathematical approach, based on information theory, for historical reconstructions in watersheds. We analyze spectral entropies calculated directly or indirectly for sediment cores or long-term hydrologic time series respectively. Spectral entropy measures changes in Shannon’s information of natural patterns (e.g. particle size distributions in lake bottoms or streamflow regimes) as they respond to different drivers. We illustrate the application of our approach with two case studies: a reconstruction of a time series of historical changes from a sediment core, and the detection of hydrologic alterations in watersheds associated to climate and forestry activities. In the first case we calculated spectral entropies from 700 sediment layers encompassing 1500 years of history in Loon Lake (Southern Oregon). In the second case, we calculated annual spectral entropies from daily discharge for the last 45 years in two experimental watersheds in the H. J. Andrews LTER site (Oregon Cascades). In Loon Lake our approach separated, without supervision, earthquakes from landslides and floods. It can also help to improve age models for sedimentary layers. At H. J. Andrews’s sites our approach was able to identify hydrological alterations following a complete clear cut in 1975. It is also helpful to identify potential long-term impacts of these forestry activities, enhanced by climate change. Our results suggest that spectral entropy is central for translating between historical structural changes in natural patterns, and their timing and relevance in watershed history. Therefore, a more robust reconstruction of watershed history and a better identification of drivers for pressing environmental issues, seems possible under the framework of Shannon’s information theory.