Data Driven Identification of Environmental Hot Moments through Regimes Shift

Temporal variability observed in watershed hydrological and biogeochemical attributes in an environment can be attributed to non-linear and intricately coupled hydrological and biogeochemical processes. Regime shifts in these processes - that occur as a result of unique combination of hydrological events and biogeochemical conditions - can result in the formation of hot moments or short bursts of time that exhibit disproportionately higher reaction rates in a longer intervening time period. These HMs exhibit disproportionate influence on ecosystem dynamics as these time periods are associated with elevated environmental health risks or significant potential for contaminant removal. Thus identification of environmental HMs is vital for understanding the ecosystem dynamics and risk prevention and assessment.

The objective of our study is to develop a predictive understanding of the development and occurrence of HMs through data-driven methodology. Hydrological and biogeochemical data acquired at East River Watershed in Colorado, were used for HMs identification and prediction. 3 years of hydrological and biogeochemical attributes, such as discharge, calcium, nitrate, have been used to facilitate our understanding of the watershed dynamics at East River Watershed. Principle Component Analysis (PCA) and time series analysis have been performed to identify the dominant hydrological and geochemical processes and the corresponding temporal scales. Hidden Markov Models (HMMs) have been applied to distinguish the temporal variability of these dominant processes and identify regime shifts. Through understanding the regime shifts, triggering events (arising regime) and sharp transitions (switching regimes) can be identified through data driven models. These triggering events and sharp transitions are usually the target time point and duration that contribute to HMs development and occurrence. Our data-driven method provides a unique perspective for detection of environmental HMs and understanding ecosystem dynamics in heterogeneous and temporally variable watershed environments