Near-term forecasts of environmental outcomes can inform real-time decision making. Data assimilation modeling techniques can be used for forecasts to leverage real-time data streams, where the difference between model predictions and observations can be used to adjust the model to make better predictions tomorrow. In this use case, we developed a process-guided deep learning and data assimilation approach to make 7-day forecasts of daily maximum water temperature in the Delaware River Basin. Our modeling system produced forecasts of daily maximum stream temperature with an average root mean squared error (RMSE) from 1.2 to 1.6°C for 1-day lead time across all sites. The data assimilation algorithm successfully adjusted the process-guided deep learning model states and marginally improved forecast performance when compared to forecasts produced using the process-guided deep learning model alone (7-13% lower RMSE with the data assimilation algorithm). Our model characterized forecast uncertainty relatively well as 57-80% of observations were within 90% forecast confidence intervals across all sites and lead times, and the uncertainty associated with our forecasts allow managers to anticipate probability of exceedances of ecologically relevant thresholds and aid in decisions about releasing reservoir water downstream. The flexibility of deep learning models to be applied to various prediction problems shows promise for using these types of models to forecast many other important environmental variables and aid in decision making.
+Four first-order (Hadean, Archean, Proterozoic and Phanerozoic eon) and nine second-order (Paleoarchean, Mesoarchean, Neoarchean, Paleoproterozoic, Mesoproterozoic, Neoproterozoic, Paleozoic, Mesozoic and Cenozoic era) units continue to provide intuitive subdivision of geological time. Major transitions in Earth’s tectonic, biological and environmental history occurred at approximately 2.5-2.3, 1.8-1.6, 1.0-0.8 and 0.7-0.5 Ga, and so future rock-based subdivision of pre-Cryogenian time, eventually by use of global stratotypes (GSSPs), will likely require only modest deviation from current chronometric boundaries (GSSAs) at 2.5, 1.6 and 1.0 Ga, respectively. Here we argue that removal of GSSAs could be expedited by establishing event-based concepts and provisional, approximate ages for eon-, era- and period-level subdivisions as soon as practicable, in line with ratification of an Ediacaran GSSP in 2004 and chronostratigraphic definition of the Cryogenian Period at c. 720 Ma in 2012. We also outline the geological basis behind current chronometric divisions, explore how they might differ in any future rock-based scheme, identify where major issues might arise during the transition, and outline where some immediate changes to the present scheme could be easily updated/formalised, as a framework for future GSSP development. In line with these aims, we note that the currently recommended four-fold Archean subdivision has not been formally ratified and agree with previous workers that it could be simplified to an informal three-fold subdivision, pending more detailed analysis. Although the ages of period boundaries would inevitably change in a more closely rock-based or chronostratigraphic scheme, we support retention of all currently ratified period names. Existing period names, borrowed from the Greek, were chosen to delimit natural phenomena of global reach. Any new global nomenclature ought to follow this lead for consistency, and so we discourage the use of supercontinent names (e.g. Rodinian, Columbian) and regional phenomena, however exceptional. In this regard, we tentatively suggest that a new period (e.g. the ‘Kratian’), could precede the Tonian as the first period of the Neoproterozoic Era and we concur with previous authors that the existing Siderian Period (named for banded iron formations) would fit better as a chronostratigraphically defined period of the terminal Archean. Indeed, all pre-Cryogenian subdivisions will need more conceptual grounding in any future chronostratigraphic scheme. We conclude that improved rock-based division of the Proterozoic Eon would likely comprise a three-fold, period-level subdivision of the Paleoproterozoic Era (Oxygenian Rhyacian, Orosirian), a four-fold subdivision of the Mesoproterozoic Era (Statherian, Calymmian, Ectasian, Stenian) and potentially four-fold subdivision of the Neoproterozoic Era (pre-Tonian ‘Kratian’, Tonian, Cryogenian and Ediacaran). Future refinements towards an improved rock-based pre-Cryogenian geological time scale could be propoosed by new international bodies to cover the 1) pre-Ediacaran Neoproterozoic, 2) Mesoproterozoic, 3) Paleoproterozoic and 4) Archean (and Hadean) as few experts and disciplines can speak to the entire pre-Cryogenian rock record.
+