Student Intern Research Projects NWS-PSU The impact of climate change and lake ice out dates in the Eastern United States By Frank Schneider1 The Pennsylvania State University and Richard H. Grumm National Weather Service in State College Abstract: An examination was made of lake ice out data over the eastern United States. Ice out data was obtained from Minnesota to Maine. Every lake examined showed a general trend toward earlier ice out dates. In addition to ice out dates, a few lakes had both ice-in and ice-out data allowing the examination of the changes in total ice days. Similar to ice out dates, the length of time lakes are iced over shows a trend toward a later total time of ice cover. The few lakes examined showed a trend toward a later beginning of the ice season. National Weather Service Cooperative Observing site data in close proximity to several of the lakes were examined. These data showed a basic trend toward warmer late winter and early season mean temperatures similar to the trend in earlier ice out dates. In general warm March and April weather was associated with the earlier ice out dates. Conversely cold late winter and early springs were associated with the later ice out dates, which often extended into late April and May during the coldest years. 1 Lead author and primary researcher as part of an undergraduate internship at the National Weather Service in State College Autumn semester 2012. Student Intern Research Projects NWS-PSU 1. Introduction Climate change is normally approached from the perspective of changes in the temperatures regional domains, often using single observing locations, or large geographic regions of the globe. The issue is often presented related to warmer winters, warmer summers, changes in precipitation patterns, and tropical storm intensities. The local effects of climate change such as the changes in the growing season length and changes in the foliage are more local and tangible to people. This study uses lake ice out data in an attempt to bring the concept of climate change to a more local level with potential impacts that are tangible to the general public. This paper examines lake ice data and its potential relationship with climate change. Lake ice out data is far more readily available than lake ice in dates. Thus, it is easier to examine the change in ice out dates relative to ice in and the total length of time a lake may be covered by ice. Thus, this study focuses on the change of ice out dates over time which can then be related back to climate change. This study of lake ice data was conducted to assess climate change through the observations on the changes of ice out dates. This approach brings the concept of climate change down to a very local and understandable level. The lake ice-out data were compared to nearby temperature data. Nearby, National Weather Service (NWS) Cooperative Observing Program (COOP) sites were used to see if there were any noticeable relationship between temperature and ice out dates. The 8 lakes used in this study spanned from New England to Minnesota (Figure 1) Lake Ice-in Data Ice-Out Data Missing Data Periods Sunapee n/a 1869-2012 none Winnipesaukee,NH n/a 1887-2012 none Spofford, NH n/a 1927-2012 none Minnetonka, MN n/a 1855-2010 1861-1876 and 1879-1886 Brant, NY 1908-2011 1908-2011 1910 and 1911 Medota, WI 1908 -2012 1908 -2012 1854 and 1855 Moosehead, ME n/a 1848-2008 none Ponkapoag, MA n/a 1886-2008 none Table 1. A listing of the lakes used in this study, the length of the record for ice in and ice out dates when available. Missing data are noted as these data are often tracked by historical societies, volunteers or other volunteer organizations. 2. Methods and Data These data were obtained from a number of websites which tracked lake iceout dates. The Lake Sunapee data was obtained from the Sunapee Tower Office. The Lake Student Intern Research Projects NWS-PSU Winnipesaukee data was gathered from Wikipedia. The Lake Spofford data was retrieved from the Chesterfield, NH historical society. Lake Minnetonka’s data in Minnesota was obtained from the Freshwater Society. Lake Brant’s data was retrieved from The Brant Lake Association. Data on Moosehead Lake was obtained from the New England Water Science Center in Maine. The observers were Kennebec Water Power Company and Maine Department of Inland Fisheries and Wildlife. Data on Lake Ponkapoag was also obtained from New England Water Science Center in Maine, and the observer was Blue Hills Observatory.The National Weather Service (NWS) Cooperative Observing Program (COOP) data were used to compare the ice out data to monthly mean temperature data. These COOP data were obtained from the NWS XMACIS program or xm Applied Climate Information System. The goal was to determine the relationships between changes in lake ice to temperatures. All the obtained data was analyzed and plotted in MS_Excel. The lake ice-out and if available ice-in dates were in the format of year, month and day. These dates were converted to the Julian date before an analysis was conducted to avoid issues related to leap year. Thus all calculations are based on the day of the year not the calendar date. The mean ice-out dates for each lake were computed and plotted to examine trends in ice-out dates. Excel’s built in functions were used to produce regression analysis and trend lines. All graphics using ice out data were produced using MS-Excel. A full suite of Excel statistical analysis features were used to examine the mean, standard deviation, skewness and Kurtosis of the ice out data. 3. Results All the analyzed lakes showed a similar trend of earlier Ice out dates. There were some lakes with more significant changes than others. Two Lakes Brant in New York and Mendota in Wisconsin had both ice-in dates and ice-out dates, allowing an analysis of the total days they were frozen over. As previously stated, both lakes showed trends of decreasing total number of frozen-over. From 1908 to 2012, Lake Mendota had a decrease in the number of frozen-over days of about 19 days. Earlier in the twentieth century the lake was frozen over for 106 days which has decreased to around 87 frozen-over days (Figure 2). Over the same period, Lake Brant had a decrease of about 6 days from 89 to about 83 days (Figure 3). It’s unclear as to why there was a relatively smaller change in frozen-over days at Lake Brant relative to Lake Mendota. The more northern latitude and interior continental location could play a role in the larger changes at Lake Mendota relative to Lake Brant. Perhaps climate change and warming is occurring more rapidly in central North America relative to the coastal regions. The data from COOP sites that were within 15 miles of the lakes show that Mendota (Figure4) had an average increase in temperatures for the month of March of around 8 degrees, while Lake Brant (Figure 5) shows an average temperature increase of about 2 Student Intern Research Projects NWS-PSU degrees with data reaching a decade past that of Mendota. These results suggest that warming during the month of March is occurring faster in Wisconsin than New York. The temperature scatter plot shows more variability of the yearly mean monthly temperatures in Wisconsin relative to central New York (Figs. 4 & 5). The lake frozen-over data supported these temperature data and show a similar trend. Lake Mendota had a higher standard deviation and a much steeper regression line suggesting highly variable data. All data showed a negative trend, with an average of earlier ice out dates since the data collection began. Three lakes, Moosehead (Figure 6), Mendota (Figure 7) and Sunapee (Figure 8) had the longest and most consistent data records when compared with the other lakes that were analyzed. The trend analysis could be used to estimate how early the ice out dates will be in 10 or 60 years, assuming a steady increase in monthly temperatures. For example Moosehead Lake (Figure 6) in Maine has expected ice out dates for 2020, 2040, 2070, of 122, 121, 119 days respectively, the ice season in the spring could be about 3 days shorter by 2070. The data has some extreme variability in it and is reflected in the R^2 value, showing that the trend line may not be a good predictor for ice out dates at any specific year. The trends provide the only useful information over a period of time. The data from all lakes suggested that 1945 was an interesting year with a consistently earlier ice-out date at each respective lake. At 2 lakes 1945 had the earliest ice out date at Sunapee and Moosehead lakes. The COOP data showed very high monthly average temperature values for the March 1945 (COOP Data). This year there was actually an instense heat wave in north america caused by a strong ridge that built over the middle part of the country. (Figure 9) (Grumm 2011) If you look at the ice out charts (ice out charts), you can see that the ice outs for the lakes are occurring, on average, earlier each year. These trends are consistent with all the lakes we have looked at so far. We could use this data and trend lines to assess what the ice out dates might be in the future, but as stated earlier with the example using Moosehead, the r^2 values are relatively low. Meaning our trend lines don’t fit the data well, and give a fairly inaccurate prediction. The following tables show a statistical summary of the Lakes that were analyzed. The r values are all relatively low, suggesting a weak correlation. Also from the table you can see that all the data is skewed to the left, towards earlier ice out dates. It should be noted that Brant,NY had some years were there was no ice over, so use with caution. Student Intern Research Projects NWS-PSU Lake R² Regression Equation Trend Sunapee,NH 0.1538 y = -0.0961x + 299.99 Earlier Ice out Winnipesaukee,NH 0.0868 y = -0.0799x + 264.53 Earlier Ice out Spofford, NH 0.0745 y = -0.1296x + 357.52 Earlier Ice out Minnetonka, MN 0.0203 y = -0.0339x + 169.57 Earlier Ice out Brant, NY 0.0024 y = -0.0165x + 135.7 Earlier Ice out Medota, WI 0.0864 y = -0.1705x + 107.4 Earlier Ice out Moosehead, ME 0.1249 y = -0.0591x + 241.71 Earlier Ice out Ponkapoag, MA 0.1763 y = -0.1696x + 406.01 Earlier Ice out Table 2: Lake ice out data analysis showing the lake name and state, the r value, the regression equation and the general trend of these data. 4. COOP Data With an early look at the coop sites near some of the lakes. These data show a correlation between rising temperatures and earlier Ice out dates. This trend was well defined for Lake Mendota, Lake Sunapee and Moosehead Lake (Figure 10) (Figure 11) (Figure 12). It should be noted that the COOP data for Moosehead Lake (Figure 12) had a significant amount of data missing. There was no data for 1933 and from 1949 to 1961. Despite the missing data, the trend lines showed an unmistakable warming trend. Lake Julian Date Average Day of Year Standard Deviation Skewness Kurtosis Sunapee,NH 113.47 23-Apr 10.19 -0.37 0.09 Winnipesaukee,NH 109.41 19-Apr 9.44 -0.39 0.33 Spofford, NH 101.41 11-Apr 10.43 -0.19 -0.10 Minnetonka, MN 103.56 13-Apr 9.54 -0.27 0.75 Brant, NY 103.36 13-Apr 8.03 -0.01 -0.44 Medota, WI 93.80 3-Apr 11.86 -0.23 0.04 Moosehead, ME 127.81 3-May 7.77 -0.18 0.22 Ponkapoag, MA 75.88 17-Mar 14.34 -0.71 0.07 Table 3: Statistical summary of all lakes. Data include the Lake name and State, the average ice-out date, the corresponding day of the year, the standard deviation in days, the skewness and kurtosis of these data. Student Intern Research Projects NWS-PSU 5. Conclusion The lake ice-out data provides evidence of local climate change over the past 100 years. Warmer springs are reducing the length of the ice season. For two lakes which had both ice-in and ice-out data, a similar trend was observed in the autumn with later ice-in dates and thus a slow reduction in the total number of days with ice cover. The impacts of shorter ice seasons could lead to warmer lakes which could lead to changes to the flora and fauna of these lakes, an area of worthy further study. These data are in agreement with the IPCC studies related to CO2 emissions, at least locally in the central and eastern United States, the climate has been warming. If the role of CO2 is the cause of this warming, the ice out dates at all the lakes presented in this study will continue to shift to an earlier date. Other studies have shown that the impact of warming is stronger at higher latitudes. Recent changes in the Arctic Sea ice show similar, if not more dramatic trends with rapidly decreasing ice in the Arctic during the spring and summer and longer period of time to re-establish ice cover in the arctic. Both sea ice and lake ice data show that at least in eastern United States and in the Arctic, climate change is real and trending toward warmer conditions less conducive to maintaining ice on the lakes of the eastern United States as long as it was in the first half of the 20th century. The ice-out dates are trending towards earlier ice-out as the regional climate warms. These data seem to imply a stronger effect in the upper Midwest relative to New England. The COOP data support the ice-out data showing that the current warming trend is occurring in both the upper Midwest and the northeast, with a sharper increase in the Upper-Midwest. The focus here was on ice-out data during the spring months. However, the yearly temperature trends are similar showing a gradual warming trend over the course of a year. This implies that there will be less snowfall and more wintertime precipitation in the form of rain. There is clearly a trend toward shorter and less severe winter seasons which will decrease winter activities such as ice fishing, ice skating on lakes, and snowmobiling. A recent article (NY Times 2012) suggested that climate change is decreasing the snow season and will lead to a demise of many ski resorts in the eastern and southern United States. Conversely, warm season activities, such as cycling, swimming, fishing, and hiking will increase with the lengthening warm season. A similar trend is already lengthening the growing season. One issue with a lengthening growing season is the impact of late freezes such as the freeze which followed the record warm conditions of March 2012 (Grumm 2012). Investments in warm season recreation will likely provide less risk than investments in winter related activities. Pools, which traditionally are open from Memorial Day to Labor Day, may have to rethink the length of the swimming season as the climate warms. Student Intern Research Projects NWS-PSU In the autumn, the growing season will be longer, and foliage will occur at a later date and will continue to do so as this trend persists. This could impact the leaf-tourist industry in many parts of the United States and Canada. The data show a lot of variability and therefore we cannot predict in detail the changes ahead, but the trends are clear and we can tell that the Earth is warming and climate change is occurring. 6. Acknowledgements. Thank you to the Pennsylvania State University for their wealth of COOP data that supports my observations. Thank you to Rich Grumm for guiding me and aiding me with this project. Most ice out data was retrieved via webpages. The Spofford Lake data was retrieved from the Cheshire County, NH historical society in August 2012. Student Intern Research Projects NWS-PSU Figure 1: A map of the all the locations of the lakes used in this study. A total of 8 lakes were used. (Return to text) Student Intern Research Projects NWS-PSU Mendota, Days Frozen Over y = -0.1705x + 432.47 R² = 0.0864 140 120 100 Ice over 80 days 60 Series1 Linear (Series1) 40 20 1900 1920 1940 1960 1980 2000 Year Figure2: Scatter plot of the number of days Lake Mendota was frozen over. (Return to text) 140 Brant, Days Frozen Over y = -0.0162x + 118.94 R² = 0.0005 120 100 Days Frozen 80 Over 60 Series1 Linear (Series1) 40 20 0 1900 1920 1940 1960 1980 Year Figure 3: Number of Days that Lake Brant was frozen over (1908 to 1994) (Return to text) 2000 Student Intern Research Projects NWS-PSU Average Temperatures in March, Mendota 50 y = 0.1091x - 183.14 R² = 0.1772 45 40 35 Series1 30 Linear (Series1) 25 20 15 1940 1950 1960 1970 1980 1990 2000 2010 2020 Figure 4: Average temperatures in March for Lake Mendota. (Return to text) Average Temperature for March, Brant 45 y = 0.0029x + 27.373 R² = 0.0002 40 Temperature 35 30 Series1 25 Linear (Series1) 20 1940 1950 1960 1970 1980 1990 2000 2010 2020 Year Figure 5: Average temperature for Lake Brant in the month of march, ranging from 1943 to 2011. (Return to text) Student Intern Research Projects NWS-PSU 160 Moosehead Ice-Out Day y = -0.0591x + 241.71 R² = 0.1249 140 Julian Day Julian Day 120 Linear (Julian Day) 100 1850 1870 1890 1910 1930 Year 1950 1970 1990 2010 Figure 6: Julian Day that Lake Moosehead iced-out (1850 to 2008) (Return to text) 135 Mendota Ice-Out Days 125 y = -0.0973x + 281.92 R² = 0.1362 115 105 Julain Day 95 Series1 Linear (Series1) 85 75 65 55 1850 1900 1950 Year Figure 7: Ice- out dates for Lake Mendota (Return to text) 2000 Student Intern Research Projects NWS-PSU Sunapee Ice-Out Day y = -0.0961x + 299.99 R² = 0.1538 140 130 120 JulianDay Linear (JulianDay) Julian Days 110 100 90 80 1860 1880 1900 1920 1940 1960 Year Figure 8: Julian Day that Lake Sunapee iced-out. (1869 to 2012) (Return to text) 1980 2000 2020 Student Intern Research Projects NWS-PSU Figure 9: As in Figure 1 except for the composite pattern and standardized anomalies from the 20th Century re-analysis data for the period of 0000 UTC 1 March through 1800 UTC 31 March 1945. (Return to text) Student Intern Research Projects NWS-PSU Average Monthly Temperatures, Mendota 60 y = 0.0108x + 45.211 R² = 0.019 50 Temperature (F) 40 y = 0.0382x + 29.273 R² = 0.0951 30 20 February March 10 y = 0.0292x + 18.881 R² = 0.0444 April Linear (February) 1869 1875 1881 1887 1893 1899 1905 1911 1917 1923 1929 1935 1941 1947 1953 1959 1965 1971 1977 1983 1989 1995 2001 2007 0 Year Linear (March) Linear (April) Figure 10: The average monthly air temperatures for the months of February, March and April for Lake Mendota (Return to text.) Student Intern Research Projects NWS-PSU Average Monthly Temperatures, Moosehead 45 40 y = 0.0399x + 34.579 R² = 0.077 Temperature (F) 35 30 y = 0.0301x + 21.817 R² = 0.0223 25 20 15 February 10 March 5 y = 0.0724x + 9.4616 R² = 0.0931 1932 1936 1939 1942 1945 1948 1964 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 2012 0 Year April Linear (February) Linear (March) Linear (April) Figure 12: Average monthly air temperatures for February, March, and April for Lake Moosehead (Return to text) Student Intern Research Projects NWS-PSU Average Monthly Temperatures, Sunapee Temperature (F) 70 60 y = 0.0933x + 51.186 R² = 0.151 50 y = 0.0642x + 39.04 R² = 0.0787 40 February 30 y = 0.0591x + 29.962 R² = 0.0493 20 March April Linear (February) 10 Linear (March) Year Linear (April) Figure 11: Average monthly air temperatures for the moths of February, March and April for Lake Sunapee (Return to text)