�鶹��Ƶ

As the global climate changes and temperatures continue to increase, lakes with seasonal ice coverage are experiencing reductions in ice duration. Decreased ice duration can have impacts on the chemical, biological, and physical processes in the lakes as well as downstream ecosystems. Records of ice-on and ice-off timing (“ice phenology”) on lakes are imperative for understanding the potential impacts on subsequent biogeochemical and ecological conditions. However, records of ice phenology are non-existent for many lakes. This gap could be filled with models that use meteorological or hydrological variables to reconstruct (“hindcast”) past ice phenology. To date, two modeling approaches have been used to determine the presence of ice on lakes: leveraging meteorological variables recorded in or near the watershed, or using water temperature sensors on buoys in the center of the lake. In contrast, this study tested the degree to which hydrological variables recorded at the outlet of The Loch, a subalpine lake in Rocky Mountain National Park, could be used to reconstruct an ice-clearance record.��
To calibrate a model that determines ice presence, I used 10 years of streamflow, water temperature, and conductivity data from a Parshall flume at the lake outlet, alongside weekly photos of the lake that confirmed whether ice was on or off the lake. From there, I built logistic regression model versions that hindcasted the presence of ice on The Loch. I then applied the model version with the highest accuracy to the entire hydrological dataset for The Loch, which spans back to 1984, and created a time series of hindcasted ice-off dates. The variables used in the final model version (R2 = 0.88) were stream temperature and cumulative discharge. In most years, the probability of The Loch’s being ice free increased rapidly after cumulative discharge exceeded 14.16 m3/s, and stream temperature exceeded 5°C. This method for reconstructing ice clearance on lakes may be useful for other systems with infrequent in situ observations, and future work can evaluate its efficacy where both long-term ice measurements and lake outflow data exist. This method also takes advantage of automated and relatively inexpensive pressure and temperature sensors placed in the outflow of lakes , which could allow more lakes to be monitored.