- Email: [email protected]
Whole truths vs. half truths – And a search for clarity in long-term water temperature records
Estuarine, Coastal and Shelf Science 157 (2015) A1eA6
Contents lists available at ScienceDirect
Estuarine, Coastal and Shelf Science journal homepage: www.elsevier.com/locate/ecss
Invited feature
Whole truths vs. half truths e And a search for clarity in long-term water temperature records R.W. Fulweiler a, *, A.J. Oczkowski b, K.M. Miller c, C.A. Oviatt d, M.E.Q. Pilson d a
Department of Earth and Environment, Department of Biology, Boston University, 685 Commonwealth Ave. Boston, MA 02215, USA United States Environmental Protection Agency, USA c CSC, Science, Engineering, and Mission Support, USA d Graduate School of Oceanography, University of Rhode Island, USA b
a r t i c l e i n f o
a b s t r a c t
Article history: Accepted 24 January 2015 Available online 10 February 2015
There is widespread acceptance among the scientific community that human activities are the primary cause of present day climate change. But, how a changing climate impacts ecosystems is still a source of confusion to the public. Some of this confusion is associated with a lack of clear communication among journalists and scientists, particularly when it comes to addressing variability and uncertainty in ecological datasets. Here we use long-term surface water temperature data sets from Narragansett Bay and a recent misunderstanding of long-term temperature data that occurred on the national stage as a case study. Specifically, we re-evaluate and update the record and examine the variability inherent in long-term data sets. We found that despite high year to year variations the surface waters of Narragansett Bay have increased between 1.4 C to 1.6 C total rise over the last fifty years. Winter warming has been especially high over this time period, increasing between 1.6 C to 2.0 C. Finally, we identify the need for scientists, politicians, and journalists to appropriately address data variability and we argue for increased communication among these groups. © 2015 Elsevier Ltd. All rights reserved.
Keywords: climate long-term changes water temperature temporal variations long-term records
Editor's note
1. Introduction
As the public and decision-makers increasingly become aware of local and global environmental change, it is necessary for the environmental research and education community to communicate and transfer information as accurately and transparently as is possible. It is both a matter of clear initial exposition of the actual factsdincluding education of the various stakeholder sectorsd, and prevention (and, unfortunately, too often) post-hoc correcting misunderstandings of what the facts are. In the Invited Feature Article in this issue, Fulweiler and colleagues review high quality temperature records for a well-studied coastal bay, and furnish a clear instance of information transfer on an issue that had become garbled in in the process of earlier communication: whether or not waters of Narragansett Bay had become warmer during recent decades.
As the global population hits 7 billion there can be little doubt that our presence has had a lasting impact on the biosphere. Human activities have modified over 50% of Earth's land surface (Hooke et al., 2012), we appropriate 24% of the Earth's net primary production (Haberl et al., 2004), and we have doubled the amount of biologically reactive nitrogen circulating through the biosphere (Fowler et al., 2013). These activities and many more have convinced some scientists that we are now in a geologic epic known as the Anthropocene (Crutzen, 2002). One of the most profound changes has been an increase in atmospheric levels of greenhouse gases and the subsequent warming of the planet. Importantly, there is a strong consensus, with 97% of climate researchers agreeing, that the majority of Earth's warming is driven by anthropogenic greenhouse gas emissions (Anderegg et al., 2010). Long-term trends in temperature are based on terrestrial and aquatic palaeoclimate data with time scales from decades to millions of years. According to the most recent Intergovernmental Panel on Climate Change (IPCC) report these proxy data show that temperature in the latter half of the 20th century are very likely higher than any period over the last 500 years and likely highest in the last 1300
* Corresponding author. E-mail address: [email protected] (R.W. Fulweiler). http://dx.doi.org/10.1016/j.ecss.2015.01.021 0272-7714/© 2015 Elsevier Ltd. All rights reserved.
A2
R.W. Fulweiler et al. / Estuarine, Coastal and Shelf Science 157 (2015) A1eA6
years (IPCC, 2013). Most recently, Marcott et al. (2013) reported that over the last decade global temperatures have yet to exceed peak interglacial values but that the last decade has been warmer than 75% of the Holocene temperature record. The instrumental record of air temperature is much shorter but exhibits a 0.76 C (0.57 Ce0.95 C) total increase between 1850e1899 and 2001e2005 (IPCC, 2013). Likewise, ocean water temperatures have increased. The ocean global volume mean temperature increase between the surface and 300 m depth was 0.31 C from the mid-1950s to the mid-1990s (Levitus et al., 2000). While more recent temperature trends have seen a “pause” in the rate of surface land and ocean temperature increase this apparent slowdown has been attributed to significant temperature increases in the deep ocean (Balmaseda et al., 2013; Cheng and Tung, 2014; Trenberth et al., 2014). Like the open ocean, coastal waters have also warmed. A unique one hundred and seventeen year temperature record from Great Harbor, Woods Hole, Massachusetts revealed warming between 1970 and 2002 at a rate of 0.04 C per year or a 1.2 C total rise in mean annual temperature (Nixon et al., 2004). Unfortunately, such consistent long-term records are rare and as we try to document temperature changes in other coastal systems we must rely on more variable data sets. Additionally, and this is true of all field observations, natural variability as well as larger scale climatic cycles can add noise to temperature record trends. Such variability can lead to confusion and even unwarranted controversy as the public seeks finite truths and scientists struggle with assessing and effectively communicating uncertainty associated with ecological measurements. When a discussion of the magnitude of warming of our local Narragansett Bay (Rhode Island) waters recently reached a national stage, we were motivated to take a closer look at the evidence for warming in the bay and to address the complexities associated with quantifying trends in long-term temperature datasets (Nixon et al., 2004; Emery, 2013). Briefly, in a speech to the Senate, Senator Whitehouse said, “Narragansett Bay waters are getting warmer e 4 degrees Fahrenheit warmer in the winter since the 1960s (4/9/2013).” He was referring to work we (RWF) published reporting a surface winter (December, January, February) warming of 2.2 C between 1960 and 2006 (Nixon et al., 2009). The temperature data referenced in Nixon et al. (2009) were collected as part of a long-term monitoring program designed to survey Phytoplankton populations in Narragansett Bay. At the time the Senator was referencing the most recently published study reporting the warming waters in the bay. However, seven years later there were more data available and the journalist running the “Truth-O-Meter” for PolitiFact accessed a different although similar temperature dataset collected as part of a complementary long-term monitoring program designed to survey fish. In plotting the temperature data from the Fish Survey, but not those from the Phytoplankton Survey, and running their own regression, PolitiFact ruled “The trend is certainly correct, but Whitehouse is too far off for the Truth-O-Meter to register True. It is, to pardon the pun, a matter of degree. Because the temperature rise is a little more than half of what he said, we rate his statement Half True (Emery, 2013).” Given the different datasets and different timescales used to make this determination, we felt that a more synthetic assessment was warranted. Here we use Narragansett Bay as a case study for discussing the complexities of ecological datasets and of communicating analytical results to broader audiences. First, we provide the most complete and up-to-date data on temperature increases in the bay and we discuss the perceived discrepancies that inspired us to compare three distinct, yet complementary, temperature datasets. We then touch on the wider implications of three decades of warming for the ecology of Narragansett Bay and other temperate marine systems.
2. Methods We analyzed and compared temperature data from Narragansett Bay, RI that were collected from three different sources. To our knowledge, these are the three longest temperature records for the bay. In two cases, water temperature was collected as part of the long-term weekly Phytoplankton and Fish Surveys conducted by graduate assistants from the Graduate School of Oceanography (GSO) at the University of Rhode Island in Narragansett, RI. The third temperature dataset was from a National Oceanic and Atmospheric Administration (NOAA) monitoring station in Newport, RI. We compiled each of these data sets to analyze for long-term temperature changes in this manuscript. 2.1. Temperature data collected during the Phytoplankton and Fish Surveys Through GSO assistantships, graduate students were responsible for gathering data in the morning Phytoplankton Survey and in the subsequent afternoon Fish Survey conducted from the same boat. While temperature measurements were made as part of both of these excursions, the temperature measurements were not the primary objective of the surveys. Two stations were monitored during these surveys: a station internal to the estuary and another at the opening of the bay. Because the latter is influenced by shelf waters and our focus here is on estuarine water temperatures we only used data from the internal estuary station (Fox Island, 41 34.2 N, 71 23.4 W). During the Phytoplankton Survey temperature was measured at surface and bottom depths first using a bucket and thermometer and then, after 2008, a YSI sonde (Model 6920 V2). Temperature measured with the YSI had a precision of within 0.15 C. Phytoplankton Survey data from 1999 to 2014 are available for download from http://www.gso.uri.edu/phytoplankton/(accessed June 2014). Earlier data from 1959 to 2007 were collected by Professor T. Smayda and are available at http://www.narrbay.org/d_projects/ plankton-tsv/plankton-tsv.htm (accessed June 2014). There were also years where some (1963, 1998, 2011) or all (1997, 2012) of the data were missing. Surface and bottom temperatures were measured at the same stations in the Fish Survey as in the Phytoplankton Survey. Fish Survey temperature data are available from 1959 to 2012 at http:// www.gso.uri.edu/fishtrawl/data.htm (accessed June 2014). More recent data are available upon request. Temperature measurements were made using a thermometer until 2006, after which a YSI sonde of the same model as that used in the Phytoplankton Survey. For both surveys, monthly average temperatures were calculated using the individual measurements, and overall winter means were calculated as the average of all individual measurements taken within the three winter months (December, January, February). Years for which some or all of the measurements were missing were excluded from the mean calculations and analyses. Due to the paucity of bottom temperature data, likely attributable to difficult field conditions, we only addressed surface temperature data. We were particularly concerned that the availability of bottom data might be skewed towards calmer seas associated with milder weather, thus biasing our results. 2.2. NOAA temperature data While the Phytoplankton and Fish Surveys are two of the longest temperature records for Narragansett Bay, measurements are made at a relatively low frequency (weekly). To add context, here we also analyzed additional estuarine temperature data from nearby
R.W. Fulweiler et al. / Estuarine, Coastal and Shelf Science 157 (2015) A1eA6
Newport, RI (41 30.30 N, 71 19.60 W). Measurements from 1955 to 1996 were collected by NOAA and their monthly averages were taken from (Oviatt, 2004). From 1996 to 2012, 15 min data were available online (http://tidesandcurrents.noaa.gov). The 1996e2012 data were first compiled into daily averages and then the daily averages were used to calculate the monthly and seasonal averages that were used in subsequent calculations of seasonal trends. NOAA data were not available from 28 Dec 2001 to 25 Apr 2002. Temperature values less than 3 C and greater than 35 C were excluded as were any negative values between April and November as they appear to be outliers, with values outside the observed temperature range of the bay. To be consistent with the Phytoplankton and Fish Survey datasets, only data from 1960 on were used in this analysis. 2.3. Statistical analyses All statistical analyses were performed using SAS Version 9.2 statistical software. Linear regression models were used for the analyses, to be consistent with the initial analyses performed and reported in Nixon et al. (2009). For each set of data, a linear regression model was fit, with the year as the independent variable and the mean temperature (for the specified month or season) as the dependent variable. For each model, the regression slope was interpreted as the estimated change in mean temperature (for the given month, winter season, or overall for the annual model) over a year. A 95% confidence interval was calculated about each model slope to quantify the variability associated with the estimated yearto-year changes, and to assess the statistical significance of those changes. For any model for which the confidence bounds around the slope do not contain 0, it can be concluded that the slope is significantly different from 0. In addition to the estimates of yearto-year change, an estimate of the increase over a 50-year period was made for each model by multiplying the slope by 50, and a 95% confidence interval was calculated about the estimated 50-year change by multiplying the upper and lower bounds for the slope confidence interval by 50. While generating linear regression models for each of the three sets of data allowed comparisons to Nixon et al. (2009) and the 2013 PolitiFact (Emery, 2013) evaluation, a single estimate of the temporal trend across the three sets would potentially provide a more useful, robust evaluation of the temperature changes. Therefore, for each of the annual, winter, and monthly periods, regression models were also fit using data for all three data sets combined. A multiple regression model was fit for each timespecific set, with additional model terms included to test whether the slope differed significantly among data sets (i.e., to test whether the magnitude of the temporal trend differed significantly between the Phytoplankton Survey, Fish Survey, and NOAA data). A statistically significant difference among slopes was observed for the July data only. For all other cases, the temporal trend could be described using a single estimate of the annual and 50-year increase. In addition to assessing changes in mean temperature, we were also interested in evaluating the frequency of unusually warm water temperatures. To do this, each summer temperature measurement was categorized as being above or below 23 C. A statistical model was then fit to the categorized data to test whether there was a significant increase in the probability that the number of summer days above 23 C was increasing over time. Because the data were now being evaluated on a categorized and not numeric result basis, a logistic regression rather than a linear regression was used for this analysis. All models were evaluated for any violations of the required assumptions for the linear regression model, including the temperature measurements not following a Normal or Gaussian
A3
distribution, decreasing or increasing variability across the time range, as well as evidence of a non-linear relationship between year and temperature (Weisberg, 1985). 3. Results and discussion 3.1. Comparing the Phytoplankton and Fish Survey temperature data For over five decades almost weekly measurements of surface water temperature were collected in Narragansett Bay as part of two separate efforts to quantify phytoplankton and fish species composition and abundance (Fig. 1). Importantly, neither of the surveys were specifically designed to collect water temperature and we are not aware of any efforts to inter-calibrate thermometers or standardize measurement techniques between them. Generally, each survey occurred on the same day with the Phytoplankton Survey conducted first, followed by the Fish Survey. But it was also not unusual, particularly during the winter months, for surveys to be conducted on adjacent days. In some years there were substantial differences between the two datasets. For example, in 1988 the annual mean was much higher in the Phytoplankton record (14.8 C) compared to the Fish record (10.8 C). On the other hand, in 1997 the annual mean for the
Fig. 1. Surface water temperature in Narragansett Bay measured during three different monitoring programs: Phytoplankton Survey (green circles), Fish Survey (blue circles), and NOAA buoy (grey shaded). A) Mean winter (December, January, and February) surface water temperature as measured during the Phytoplankton Survey (1960e2010), Fish Survey (1960e2011), and NOAA buoy (1960e2011). B) Mean annual surface water temperature as measured during the Phytoplankton Survey (1960e2011), Fish Survey (1960e2012), and NOAA buoy (1960e2012). Included years differ based on the availability of data, see text for details.
A4
R.W. Fulweiler et al. / Estuarine, Coastal and Shelf Science 157 (2015) A1eA6
Phytoplankton record (5.1 C) was much lower than the Fish record (11.5 C). In most cases the two datasets agree well and the overall means are virtually indistinguishable from each other (Phytoplankton: 11.52 C vs. Fish: 11.49 C). Some of the variability could be explained by the number of sampling events per year between the two data sets as surface water temperatures were collected less often during the Phytoplankton Survey compared to the Fish Survey. This was especially true in the years with the largest percent differences (1988 and 1997) where the Phytoplankton Survey only collected 32 and 17 measurements compared to the Fish Survey that collected 50 and 52 measurements, respectively. This difference in the measurement collection rate can have a large effect on the monthly means and models, because water temperatures can vary widely between the beginning and end of a month. This simple analysis highlights the importance of examining sampling effort when investigating long-term records of temperature or other environmental parameters (e.g. wind speed, dissolved oxygen concentrations, etc.).
temperature increases are significant (p < 0.05) for the majority of months except for two months in the Phytoplankton Survey temperature data (November, December), four months in Fish Survey temperature data (June, October, November, and December), and three months in the NOAA temperature data (June, July, November). Others have also observed seasonal differences in warming in the Chesapeake Bay (Preston 2004, Najjar et al., 2010) and Hudson River Estuary (Seekell and Pace, 2011). The seasonality associated
3.2. Narragansett Bay surface water temperature trends Nixon et al. (2009) reported that between 1960 and 2006, winter (D, J, F) surface water temperatures from the Phytoplankton Survey exhibited an overall mean warming of about 2.2 C. To check our previous work we re-calculated the winter surface warming from the Phytoplankton Survey temperature data over that time period with a linear regression model and found a significant increase of 0.047 C per year for a total increase of 2.4 C, slightly above the original Nixon et al. (2009) calculation. Using surface water temperatures from the Fish Survey and NOAA temperature data we found a similar overall 50-year increase (using data collected between 1960 and 2006) of 1.6 C and 2.2 C, respectively (Supplemental Table 1). Winter (D,J,F) surface water temperatures exhibit the highest inter-annual variability across all three data sets (Fig. 1a). Despite this variability a general trend of increasing winter and annual temperature is apparent (Fig. 1). Importantly, this increase is not the same in each month of the year. Instead the most pronounced monthly average temperature increase peaks in winter and decreases over the course of the year (Fig. 2). The monthly
Fig. 2. Mean month specific (±std. error) surface water temperature increase per year over the last five decades in Narragansett Bay highlights how the temperature increase is variable throughout the year. Data are model results based on three different data sets: the Phytoplankton Survey (green bars, 1960e2010), Fish Survey (blue bars, 1960e2012), and NOAA buoy (grey bars, 1960e2012). Included years differ based on the availability of data.
Fig. 3. Winter (December, January, February) days below 1 C (A) and above 5 C (B) and summer (June, July, August) days above 23 C (C) over the last five decades based on the Phytoplankton Survey data from mid-Narragansett Bay. Seasons with incomplete or absent data were excluded from the analyses and are indicated by gray bars. Absent seasons include the winters of 1997 and 2011 and the summers of 1997, 1998, 2011, and 2012.
R.W. Fulweiler et al. / Estuarine, Coastal and Shelf Science 157 (2015) A1eA6
with warming coastal waters makes it difficult to capture the level of associated ecosystem impact with a single annual value for temperature increase. This is particularly true in temperate estuaries where seasonally changing temperatures may favor some organisms over others (e.g., Gimenez, 2011). By averaging data to seasonal or annual timescales we can make general observations about long-term trends, but in doing so it is also possible to smooth over and obscure trends in short term (i.e. daily) events. For example, Narragansett Bay was historically characterized as having an annual temperature range of 1 Ce23 C (Kremer and Nixon, 1978). Using the long-term Phytoplankton record it appears that the annual temperature range in the bay is shifting with fewer winter days below 1 C and more winter days above 5 C (Fig. 3). The most pronounced change is in the summer (June, July, August) where we observe a significant (p < 0.0001) increase in the number of days above 23 C (Fig. 3). In the case of a dataset like the Phytoplankton or Fish Survey, where values are collected approximately weekly, changes in the frequency of extremely warm and/or cold measurements further confirm and strengthen the observation that temperature regime of the bay is clearly changing. 3.3. The bottom line: how much has the water temperature increased? Given our three long-term data sets, it is clear that even with the high inter-annual variability, the surface waters in Narragansett Bay have increased over the last five decades (Fig. 1). The magnitude of the overall increase depends on which dataset is used, the length of time analyzed, and how the data are handled (e.g. monthly, seasonal, or annual averages). Since 1960 the annual mean surface water temperature increase per year has been 0.032 C, 0.029 C, 0.027 C based on the temperature data from the Phytoplankton Survey, the Fish Survey, and the NOAA data sets, respectively (Table 1). If we combine all the water temperature data sets together the per year increase is 0.029 C. Thus the surface water temperature in Narragansett Bay has experienced a total rise in temperature of 1.4 C to 1.6 C over the last five decades (Table 1). Winter warming over the past half-century has been especially high. But exactly how high depends on the dataset (Fig. 1a). The estimated 50 year increase ranged from 1.6 C for the Fish
A5
Survey dataset to 2.0 C for the NOAA dataset (Table 1). These updated winter trends are not meaningfully different from one another when accounting for the model variability (as can be seen by the overlapping 95% confidence intervals in Table 1) nor from the value (2.2 C) that begot the “Half-truth” rating and inspired these analyses (Emery, 2013). After 2006 there were a series of colder winters that lessen the average winter temperature warming (Fig. 1a), explaining why the value we report here for the Phytoplankton Survey is slightly lower than that reported in Nixon et al. (2009). Summer temperatures also increase but to a lesser extent and with more variability among the data sets (NOAA: 0.8 C to Phytoplankton Survey: 1.6 C, when averaged across the three summer months: June, July, and August). Additionally, the calculated magnitude of these changes may be driven by unavailable data. For example, 2011 temperature data are missing from the Phytoplankton data set which includes a winter of high temperature according to the Fish Survey data set. As demonstrated in the Nixon et al. (2009) paper, there was substantial variability in the Phytoplankton dataset and this is true for all three datasets discussed here. Such variability is inherent in ecological data. One major driver of variability in temperature data are large scale climatic features such as the North Atlantic Oscillation (NAO). The NAO can modify temperatures in the Northeast United States coastal estuaries and North Atlantic waters (Hurrell and Dickson, 2004; Hurell, 2010) and can seemingly mask temperature trends. Generally temperatures are cooler than average during the negative phase and warmer than average during the positive phase along the east coast of the U.S. During cool negative oscillation periods such as the 1950s and 1960s and after the 2000s, winter temperatures have been colder and snowier than average. During the warm positive oscillation periods such as the 1910s to the 1930s and the 1970s to the 1990s, winter temperatures were warmer than average by up to 2 C with reduced snowfall (Oviatt, 2004). This variation produces a regional oscillation in the annual temperature with little overall change during the negative oscillation period and a net gain in the positive oscillation period. In the positive phase of NAO warm conditions result in the Atlantic Multidecadal Oscillation of warmer than average temperatures in North Atlantic waters ea term coined by Michael Mann (2012). This large pool of warm, salty water that results from the
Table 1 Regression model results for Annual and Winter (D, J, F) changes in temperature over approximately the last five decades for each of the available temperature datasets from Narragansett Bay, RI. Winter is defined as the mean of values from December of the previous year, January, and February. Annual means are for the calendar year. All values are C except for the R2 (%). Due to missing data the number of included years varies. See footnotes for specifics. Month
Model results R2
Intercept
Phytoplankton survey* Annual 37.3 52.8 Winter 24.4 71.9 y Fish survey Annual 19.9 45.6 Winter 39.6 60.7 NOAA newport, RIz Annual 39.4 42.6 Winter 26.0 74.7 All three data sets combinedx Annual 41.8 45.0 Winter 41.1 66.3 *
Estimated 50 year increase Intercept std. Error
Slope
Slope std. Error
Slope 95% lower lim.
Slope 95% upper lim.
Increase ( C)
95% lower lim.
95% upper lim.
12.3 19.4
0.032 0.038
0.006 0.010
0.020 0.018
0.045 0.057
1.6 1.9
1.0 0.9
2.3 2.9
10.2 18.6
0.029 0.032
0.005 0.009
0.019 0.013
0.039 0.051
1.4 1.6
0.9 0.7
1.9 2.5
9.5 19.0
0.027 0.040
0.005 0.010
0.018 0.021
0.037 0.059
1.4 2.0
0.9 1.0
1.9 3.0
5.5 10.9
0.029 0.036
0.003 0.006
0.023 0.025
0.034 0.046
1.4 1.8
1.2 1.2
1.7 2.3
Annual and winter temperature models include data from 1960 to 2010 and exclude 1963 and 1997e1998. Annual temperature model includes data from 1960 to 2012 and excludes 1963 and 1997e1998; winter model includes data from 1960-2011 and includes 1963 and 1997e1998. z Annual temperature model includes data from 1960 to 2012 and excludes 2002 and 2013; winter model includes data from 1960-2011 and excludes 2001. Additionally, we excluded temperatures below 3 for December, January, February and March and above 35 C from November to March; for all months we excluded values below 0 C. x Combined models were constructed using all of the data included as described in footnotes above. y
A6
R.W. Fulweiler et al. / Estuarine, Coastal and Shelf Science 157 (2015) A1eA6
Gulf Stream current carrying warm water northward for several decades, eventually sinks carrying this warmth to as deep as the thermocline at 1000 m during the ensuing negative oscillation phase which cools the surface waters by the following positive phase of the NAO (Cheng and Tung, 2014). The apparent slowdown in global warming has stimulated recent controversy but global warming nonetheless continues, as illustrated here using three separate temperature datasets from Narragansett Bay. 4. Conclusions While the majority of scientists are convinced that human activities are driving recent climate change much of the public remains uncertain. A thorough discussion of the hypotheses proposed to explain widespread public apathy over climate change is well beyond the scope of this paper and our expertise. We see the political discussion associated with Nixon et al. (2009) as a good example of the common disconnect between information produced by the scientific community and how it is perceived by the public, specifically with how best to address the variability of ecological datasets. As Emery (2013) rightly pointed out the amount of surface water temperature increase is a “matter of degree” but we would add a “plus or minus” to that statement. Importantly, the “plus or minus” was driven by comparing different data sets and different time periods. Journalists and scientists have similar goals of better understanding the world around us and then reporting what we learn in a clear, concise, and correct way. Journalists provide a valuable service by fact checking political comments and making scientific findings available to a broader audience. Nevertheless, with their position comes the responsibility of due diligence. As scientists our responsibility lies in making the time to communicate effectively with the public and with journalists and, as this case study illustrates, to do a better job at explaining and couching the inherent ecological variability associated with our averages and summaries. Science e like communication e is an iterative process and thus interpretation of results, most especially those from long-term data sets, will change over time. There is a continued need for open communication between those who do the science and those who have a voice with which to share it. After all, we are all seeking the whole truth. Acknowledgments We want to thank Dr. Theodore Smayda for starting the Phytoplankton Survey and Charles Fish who started the Fish Survey in the 1950s. We also wish to thank Dr. Tatiana Rynearson and Dr. Jeremy Collie who currently maintain the Phytoplankton and Fish Surveys, respectively, as well as those who have come before them and the Graduate School of Oceanography at the University of Rhode Island for funding the survey. These long-term data provide unique and valuable insights into changing coastal marine ecosystems. This report is contribution number ORD-009227 of the U.S. EPA's Office of Research and Development, National Health and Environmental Effects Research Laboratory, Atlantic Ecology Division. Although the information in this document has been funded by the U.S. Environmental Protection Agency, it does not necessarily reflect the views of the Agency and no official endorsement should be
inferred. Mention of trade names or commercial products does not constitute endorsement or recommendation for use. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.ecss.2015.01.021. References Anderegg, W.R., Prall, J.W., Harold, J., Schneider, S.H., 2010. Expert credibility in climate change. Proc. Natl. Acad. Sci. 107, 12107e12109. Balmaseda, M.A., Trenberth, K.E., Kallen, E., 2013. Distinctive climate signals in reanalysis of global ocean heat content. Geophys Res. Lett. 40, 1754e1759. Cheng, X., Tung, K., 2014. Varying planetary heat sink led to global warming slowdown and acceleration. Science 345, 897e903. Crutzen, P.J., 2002. Geology of mankind. Nature 415, 23. Emery, E., 2013. U.S. Sen. Sheldon Whitehouse: Narragansett Bay in R.I. has gotten 4 degrees warmer since the 1960. Providence J.. Fowler, D., Coyle, M., Skiba, U., Sutton, M.A., Cape, J.N., Reis, S., Sheppard, L.J., Jenkins, A., Grizzetti, B., Galloway, J.N., 2013. The global nitrogen cycle in the twenty-first century. Philos. Trans. R. Soc. B: Biol. Sci. 368, 20130164. Gimenez, L., 2011. Exploring mechanisms linking temperature increase and larval phenology: the importance of variance effects. J. Exp. Mar. Biol. Ecol. 400, 227e235. Haberl, H., Schulz, N.B., Plutzar, C., Erb, K.H., Krausmann, F., Loibl, W., Moser, D., Sauberer, N., Weisz, H., Zechmeister, H.G., 2004. Human appropriation of net primary production and species diversity in agricultural landscapes. Agric. Ecosyst. Environ. 102, 213e218. Hooke, R.L., Martín-Duque, J.F., Pedraza, J., 2012. Land transformation by humans: a review. GSA Today 22, 4e10. Hurell, J., 2010. North Atlantic oscillation. In: Turekian, K. (Ed.), Climate and Oceans - a Derivative of the Encyclopedia of Ocean Sciences, second ed. Elsevier, New York, NY, pp. 33e40. Hurrell, J.W., Dickson, R., 2004. Climate variability over the North Atlantic. In: Stensenth, N.C., Ottersen, G., Hurrell, J.W., Belgrano, A. (Eds.), Marine Ecosystems and Climate Variation: the North Atlantic: a Comparative Perspective. Oxford University Press, New York, NY, pp. 15e32. IPCC, 2013. Summary for policymakers. In: Stocker, T.F., Qin, D., Plattner, G.-K., Tignor, M., Allen, S.K., Boschung, J., Nauels, A., Xia, Y., Bex, V., Midgley, P.M. (Eds.), Climate Change 2013: the Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge, United Kingdom and New York, N.Y., USA. Kremer, J.N., Nixon, S.W., 1978. Coastal Marine Ecosystem: Simulation and Analysis, Coastal Marine Ecosystem: Simulation and Analysis. Springer-Verlag. Levitus, S., Antonov, J.I., Boyer, T.P., Stephens, C., 2000. Warming of the world ocean. Science 287, 2225e2229. Mann, M.E., 2012. The Hockey Stick and the Climate Wars: Dispatches from the Front Lines. Columbia University Press, New York, NY. Marcott, S.A., Shakun, J.D., Clark, P.U., Mix, A.C., 2013. A reconstruction of regional and global temperature for the past 11,300 Years. Science 339, 1198e1201. Najjar, R.G., et al., 2010. Potential climate-change impacts on the Chesapeake Bay. Estuar. Coast. Shelf Sci. 86, 1e20. Nixon, S.W., Fulweiler, R.W., Buckley, B.A., Granger, S.L., Nowicki, B.L., Henry, K.M., 2009. The impact of changing climate on phenology, productivity, and benthicepelagic coupling in Narragansett Bay. Estuar. Coast. Shelf Sci. 82, 1e18. Nixon, S.X., Granger, S., Buckley, B.A., Lamont, M., Rowell, B., 2004. A one hundred and seventeen year coastal water temperature record from Woods Hole, Massachusetts. Estuaries 27, 397e404. Oviatt, C.A., 2004. The changing ecology of temperate coastal waters during a warming trend. Estuaries 27, 895e904. Preston, B.L., 2004. Observed winter warming of the Chesapeake Bay estuary (1949e2002): implications for ecosystem management. Environ. Manag. 34 (1), 125e139. Seekell, D.A., Pace, M.L., 2011. Climate change drives warming in the Hudson river estuary, New York (USA). J. Environ. Monit. 13, 2321e2327. Trenberth, K.E., Fasullo, J.T., Branstator, G., Phillips, A.S., 2014. Seasonal aspects of the recent pause in surface warming. Nat. Clim. Change 4, 911e916. Weisberg, S., 1985. Applied Linear Regression, second ed. Wiley & Sons, New York, N.Y.
Contents lists available at ScienceDirect
Estuarine, Coastal and Shelf Science journal homepage: www.elsevier.com/locate/ecss
Invited feature
Whole truths vs. half truths e And a search for clarity in long-term water temperature records R.W. Fulweiler a, *, A.J. Oczkowski b, K.M. Miller c, C.A. Oviatt d, M.E.Q. Pilson d a
Department of Earth and Environment, Department of Biology, Boston University, 685 Commonwealth Ave. Boston, MA 02215, USA United States Environmental Protection Agency, USA c CSC, Science, Engineering, and Mission Support, USA d Graduate School of Oceanography, University of Rhode Island, USA b
a r t i c l e i n f o
a b s t r a c t
Article history: Accepted 24 January 2015 Available online 10 February 2015
There is widespread acceptance among the scientific community that human activities are the primary cause of present day climate change. But, how a changing climate impacts ecosystems is still a source of confusion to the public. Some of this confusion is associated with a lack of clear communication among journalists and scientists, particularly when it comes to addressing variability and uncertainty in ecological datasets. Here we use long-term surface water temperature data sets from Narragansett Bay and a recent misunderstanding of long-term temperature data that occurred on the national stage as a case study. Specifically, we re-evaluate and update the record and examine the variability inherent in long-term data sets. We found that despite high year to year variations the surface waters of Narragansett Bay have increased between 1.4 C to 1.6 C total rise over the last fifty years. Winter warming has been especially high over this time period, increasing between 1.6 C to 2.0 C. Finally, we identify the need for scientists, politicians, and journalists to appropriately address data variability and we argue for increased communication among these groups. © 2015 Elsevier Ltd. All rights reserved.
Keywords: climate long-term changes water temperature temporal variations long-term records
Editor's note
1. Introduction
As the public and decision-makers increasingly become aware of local and global environmental change, it is necessary for the environmental research and education community to communicate and transfer information as accurately and transparently as is possible. It is both a matter of clear initial exposition of the actual factsdincluding education of the various stakeholder sectorsd, and prevention (and, unfortunately, too often) post-hoc correcting misunderstandings of what the facts are. In the Invited Feature Article in this issue, Fulweiler and colleagues review high quality temperature records for a well-studied coastal bay, and furnish a clear instance of information transfer on an issue that had become garbled in in the process of earlier communication: whether or not waters of Narragansett Bay had become warmer during recent decades.
As the global population hits 7 billion there can be little doubt that our presence has had a lasting impact on the biosphere. Human activities have modified over 50% of Earth's land surface (Hooke et al., 2012), we appropriate 24% of the Earth's net primary production (Haberl et al., 2004), and we have doubled the amount of biologically reactive nitrogen circulating through the biosphere (Fowler et al., 2013). These activities and many more have convinced some scientists that we are now in a geologic epic known as the Anthropocene (Crutzen, 2002). One of the most profound changes has been an increase in atmospheric levels of greenhouse gases and the subsequent warming of the planet. Importantly, there is a strong consensus, with 97% of climate researchers agreeing, that the majority of Earth's warming is driven by anthropogenic greenhouse gas emissions (Anderegg et al., 2010). Long-term trends in temperature are based on terrestrial and aquatic palaeoclimate data with time scales from decades to millions of years. According to the most recent Intergovernmental Panel on Climate Change (IPCC) report these proxy data show that temperature in the latter half of the 20th century are very likely higher than any period over the last 500 years and likely highest in the last 1300
* Corresponding author. E-mail address: [email protected] (R.W. Fulweiler). http://dx.doi.org/10.1016/j.ecss.2015.01.021 0272-7714/© 2015 Elsevier Ltd. All rights reserved.
A2
R.W. Fulweiler et al. / Estuarine, Coastal and Shelf Science 157 (2015) A1eA6
years (IPCC, 2013). Most recently, Marcott et al. (2013) reported that over the last decade global temperatures have yet to exceed peak interglacial values but that the last decade has been warmer than 75% of the Holocene temperature record. The instrumental record of air temperature is much shorter but exhibits a 0.76 C (0.57 Ce0.95 C) total increase between 1850e1899 and 2001e2005 (IPCC, 2013). Likewise, ocean water temperatures have increased. The ocean global volume mean temperature increase between the surface and 300 m depth was 0.31 C from the mid-1950s to the mid-1990s (Levitus et al., 2000). While more recent temperature trends have seen a “pause” in the rate of surface land and ocean temperature increase this apparent slowdown has been attributed to significant temperature increases in the deep ocean (Balmaseda et al., 2013; Cheng and Tung, 2014; Trenberth et al., 2014). Like the open ocean, coastal waters have also warmed. A unique one hundred and seventeen year temperature record from Great Harbor, Woods Hole, Massachusetts revealed warming between 1970 and 2002 at a rate of 0.04 C per year or a 1.2 C total rise in mean annual temperature (Nixon et al., 2004). Unfortunately, such consistent long-term records are rare and as we try to document temperature changes in other coastal systems we must rely on more variable data sets. Additionally, and this is true of all field observations, natural variability as well as larger scale climatic cycles can add noise to temperature record trends. Such variability can lead to confusion and even unwarranted controversy as the public seeks finite truths and scientists struggle with assessing and effectively communicating uncertainty associated with ecological measurements. When a discussion of the magnitude of warming of our local Narragansett Bay (Rhode Island) waters recently reached a national stage, we were motivated to take a closer look at the evidence for warming in the bay and to address the complexities associated with quantifying trends in long-term temperature datasets (Nixon et al., 2004; Emery, 2013). Briefly, in a speech to the Senate, Senator Whitehouse said, “Narragansett Bay waters are getting warmer e 4 degrees Fahrenheit warmer in the winter since the 1960s (4/9/2013).” He was referring to work we (RWF) published reporting a surface winter (December, January, February) warming of 2.2 C between 1960 and 2006 (Nixon et al., 2009). The temperature data referenced in Nixon et al. (2009) were collected as part of a long-term monitoring program designed to survey Phytoplankton populations in Narragansett Bay. At the time the Senator was referencing the most recently published study reporting the warming waters in the bay. However, seven years later there were more data available and the journalist running the “Truth-O-Meter” for PolitiFact accessed a different although similar temperature dataset collected as part of a complementary long-term monitoring program designed to survey fish. In plotting the temperature data from the Fish Survey, but not those from the Phytoplankton Survey, and running their own regression, PolitiFact ruled “The trend is certainly correct, but Whitehouse is too far off for the Truth-O-Meter to register True. It is, to pardon the pun, a matter of degree. Because the temperature rise is a little more than half of what he said, we rate his statement Half True (Emery, 2013).” Given the different datasets and different timescales used to make this determination, we felt that a more synthetic assessment was warranted. Here we use Narragansett Bay as a case study for discussing the complexities of ecological datasets and of communicating analytical results to broader audiences. First, we provide the most complete and up-to-date data on temperature increases in the bay and we discuss the perceived discrepancies that inspired us to compare three distinct, yet complementary, temperature datasets. We then touch on the wider implications of three decades of warming for the ecology of Narragansett Bay and other temperate marine systems.
2. Methods We analyzed and compared temperature data from Narragansett Bay, RI that were collected from three different sources. To our knowledge, these are the three longest temperature records for the bay. In two cases, water temperature was collected as part of the long-term weekly Phytoplankton and Fish Surveys conducted by graduate assistants from the Graduate School of Oceanography (GSO) at the University of Rhode Island in Narragansett, RI. The third temperature dataset was from a National Oceanic and Atmospheric Administration (NOAA) monitoring station in Newport, RI. We compiled each of these data sets to analyze for long-term temperature changes in this manuscript. 2.1. Temperature data collected during the Phytoplankton and Fish Surveys Through GSO assistantships, graduate students were responsible for gathering data in the morning Phytoplankton Survey and in the subsequent afternoon Fish Survey conducted from the same boat. While temperature measurements were made as part of both of these excursions, the temperature measurements were not the primary objective of the surveys. Two stations were monitored during these surveys: a station internal to the estuary and another at the opening of the bay. Because the latter is influenced by shelf waters and our focus here is on estuarine water temperatures we only used data from the internal estuary station (Fox Island, 41 34.2 N, 71 23.4 W). During the Phytoplankton Survey temperature was measured at surface and bottom depths first using a bucket and thermometer and then, after 2008, a YSI sonde (Model 6920 V2). Temperature measured with the YSI had a precision of within 0.15 C. Phytoplankton Survey data from 1999 to 2014 are available for download from http://www.gso.uri.edu/phytoplankton/(accessed June 2014). Earlier data from 1959 to 2007 were collected by Professor T. Smayda and are available at http://www.narrbay.org/d_projects/ plankton-tsv/plankton-tsv.htm (accessed June 2014). There were also years where some (1963, 1998, 2011) or all (1997, 2012) of the data were missing. Surface and bottom temperatures were measured at the same stations in the Fish Survey as in the Phytoplankton Survey. Fish Survey temperature data are available from 1959 to 2012 at http:// www.gso.uri.edu/fishtrawl/data.htm (accessed June 2014). More recent data are available upon request. Temperature measurements were made using a thermometer until 2006, after which a YSI sonde of the same model as that used in the Phytoplankton Survey. For both surveys, monthly average temperatures were calculated using the individual measurements, and overall winter means were calculated as the average of all individual measurements taken within the three winter months (December, January, February). Years for which some or all of the measurements were missing were excluded from the mean calculations and analyses. Due to the paucity of bottom temperature data, likely attributable to difficult field conditions, we only addressed surface temperature data. We were particularly concerned that the availability of bottom data might be skewed towards calmer seas associated with milder weather, thus biasing our results. 2.2. NOAA temperature data While the Phytoplankton and Fish Surveys are two of the longest temperature records for Narragansett Bay, measurements are made at a relatively low frequency (weekly). To add context, here we also analyzed additional estuarine temperature data from nearby
R.W. Fulweiler et al. / Estuarine, Coastal and Shelf Science 157 (2015) A1eA6
Newport, RI (41 30.30 N, 71 19.60 W). Measurements from 1955 to 1996 were collected by NOAA and their monthly averages were taken from (Oviatt, 2004). From 1996 to 2012, 15 min data were available online (http://tidesandcurrents.noaa.gov). The 1996e2012 data were first compiled into daily averages and then the daily averages were used to calculate the monthly and seasonal averages that were used in subsequent calculations of seasonal trends. NOAA data were not available from 28 Dec 2001 to 25 Apr 2002. Temperature values less than 3 C and greater than 35 C were excluded as were any negative values between April and November as they appear to be outliers, with values outside the observed temperature range of the bay. To be consistent with the Phytoplankton and Fish Survey datasets, only data from 1960 on were used in this analysis. 2.3. Statistical analyses All statistical analyses were performed using SAS Version 9.2 statistical software. Linear regression models were used for the analyses, to be consistent with the initial analyses performed and reported in Nixon et al. (2009). For each set of data, a linear regression model was fit, with the year as the independent variable and the mean temperature (for the specified month or season) as the dependent variable. For each model, the regression slope was interpreted as the estimated change in mean temperature (for the given month, winter season, or overall for the annual model) over a year. A 95% confidence interval was calculated about each model slope to quantify the variability associated with the estimated yearto-year changes, and to assess the statistical significance of those changes. For any model for which the confidence bounds around the slope do not contain 0, it can be concluded that the slope is significantly different from 0. In addition to the estimates of yearto-year change, an estimate of the increase over a 50-year period was made for each model by multiplying the slope by 50, and a 95% confidence interval was calculated about the estimated 50-year change by multiplying the upper and lower bounds for the slope confidence interval by 50. While generating linear regression models for each of the three sets of data allowed comparisons to Nixon et al. (2009) and the 2013 PolitiFact (Emery, 2013) evaluation, a single estimate of the temporal trend across the three sets would potentially provide a more useful, robust evaluation of the temperature changes. Therefore, for each of the annual, winter, and monthly periods, regression models were also fit using data for all three data sets combined. A multiple regression model was fit for each timespecific set, with additional model terms included to test whether the slope differed significantly among data sets (i.e., to test whether the magnitude of the temporal trend differed significantly between the Phytoplankton Survey, Fish Survey, and NOAA data). A statistically significant difference among slopes was observed for the July data only. For all other cases, the temporal trend could be described using a single estimate of the annual and 50-year increase. In addition to assessing changes in mean temperature, we were also interested in evaluating the frequency of unusually warm water temperatures. To do this, each summer temperature measurement was categorized as being above or below 23 C. A statistical model was then fit to the categorized data to test whether there was a significant increase in the probability that the number of summer days above 23 C was increasing over time. Because the data were now being evaluated on a categorized and not numeric result basis, a logistic regression rather than a linear regression was used for this analysis. All models were evaluated for any violations of the required assumptions for the linear regression model, including the temperature measurements not following a Normal or Gaussian
A3
distribution, decreasing or increasing variability across the time range, as well as evidence of a non-linear relationship between year and temperature (Weisberg, 1985). 3. Results and discussion 3.1. Comparing the Phytoplankton and Fish Survey temperature data For over five decades almost weekly measurements of surface water temperature were collected in Narragansett Bay as part of two separate efforts to quantify phytoplankton and fish species composition and abundance (Fig. 1). Importantly, neither of the surveys were specifically designed to collect water temperature and we are not aware of any efforts to inter-calibrate thermometers or standardize measurement techniques between them. Generally, each survey occurred on the same day with the Phytoplankton Survey conducted first, followed by the Fish Survey. But it was also not unusual, particularly during the winter months, for surveys to be conducted on adjacent days. In some years there were substantial differences between the two datasets. For example, in 1988 the annual mean was much higher in the Phytoplankton record (14.8 C) compared to the Fish record (10.8 C). On the other hand, in 1997 the annual mean for the
Fig. 1. Surface water temperature in Narragansett Bay measured during three different monitoring programs: Phytoplankton Survey (green circles), Fish Survey (blue circles), and NOAA buoy (grey shaded). A) Mean winter (December, January, and February) surface water temperature as measured during the Phytoplankton Survey (1960e2010), Fish Survey (1960e2011), and NOAA buoy (1960e2011). B) Mean annual surface water temperature as measured during the Phytoplankton Survey (1960e2011), Fish Survey (1960e2012), and NOAA buoy (1960e2012). Included years differ based on the availability of data, see text for details.
A4
R.W. Fulweiler et al. / Estuarine, Coastal and Shelf Science 157 (2015) A1eA6
Phytoplankton record (5.1 C) was much lower than the Fish record (11.5 C). In most cases the two datasets agree well and the overall means are virtually indistinguishable from each other (Phytoplankton: 11.52 C vs. Fish: 11.49 C). Some of the variability could be explained by the number of sampling events per year between the two data sets as surface water temperatures were collected less often during the Phytoplankton Survey compared to the Fish Survey. This was especially true in the years with the largest percent differences (1988 and 1997) where the Phytoplankton Survey only collected 32 and 17 measurements compared to the Fish Survey that collected 50 and 52 measurements, respectively. This difference in the measurement collection rate can have a large effect on the monthly means and models, because water temperatures can vary widely between the beginning and end of a month. This simple analysis highlights the importance of examining sampling effort when investigating long-term records of temperature or other environmental parameters (e.g. wind speed, dissolved oxygen concentrations, etc.).
temperature increases are significant (p < 0.05) for the majority of months except for two months in the Phytoplankton Survey temperature data (November, December), four months in Fish Survey temperature data (June, October, November, and December), and three months in the NOAA temperature data (June, July, November). Others have also observed seasonal differences in warming in the Chesapeake Bay (Preston 2004, Najjar et al., 2010) and Hudson River Estuary (Seekell and Pace, 2011). The seasonality associated
3.2. Narragansett Bay surface water temperature trends Nixon et al. (2009) reported that between 1960 and 2006, winter (D, J, F) surface water temperatures from the Phytoplankton Survey exhibited an overall mean warming of about 2.2 C. To check our previous work we re-calculated the winter surface warming from the Phytoplankton Survey temperature data over that time period with a linear regression model and found a significant increase of 0.047 C per year for a total increase of 2.4 C, slightly above the original Nixon et al. (2009) calculation. Using surface water temperatures from the Fish Survey and NOAA temperature data we found a similar overall 50-year increase (using data collected between 1960 and 2006) of 1.6 C and 2.2 C, respectively (Supplemental Table 1). Winter (D,J,F) surface water temperatures exhibit the highest inter-annual variability across all three data sets (Fig. 1a). Despite this variability a general trend of increasing winter and annual temperature is apparent (Fig. 1). Importantly, this increase is not the same in each month of the year. Instead the most pronounced monthly average temperature increase peaks in winter and decreases over the course of the year (Fig. 2). The monthly
Fig. 2. Mean month specific (±std. error) surface water temperature increase per year over the last five decades in Narragansett Bay highlights how the temperature increase is variable throughout the year. Data are model results based on three different data sets: the Phytoplankton Survey (green bars, 1960e2010), Fish Survey (blue bars, 1960e2012), and NOAA buoy (grey bars, 1960e2012). Included years differ based on the availability of data.
Fig. 3. Winter (December, January, February) days below 1 C (A) and above 5 C (B) and summer (June, July, August) days above 23 C (C) over the last five decades based on the Phytoplankton Survey data from mid-Narragansett Bay. Seasons with incomplete or absent data were excluded from the analyses and are indicated by gray bars. Absent seasons include the winters of 1997 and 2011 and the summers of 1997, 1998, 2011, and 2012.
R.W. Fulweiler et al. / Estuarine, Coastal and Shelf Science 157 (2015) A1eA6
with warming coastal waters makes it difficult to capture the level of associated ecosystem impact with a single annual value for temperature increase. This is particularly true in temperate estuaries where seasonally changing temperatures may favor some organisms over others (e.g., Gimenez, 2011). By averaging data to seasonal or annual timescales we can make general observations about long-term trends, but in doing so it is also possible to smooth over and obscure trends in short term (i.e. daily) events. For example, Narragansett Bay was historically characterized as having an annual temperature range of 1 Ce23 C (Kremer and Nixon, 1978). Using the long-term Phytoplankton record it appears that the annual temperature range in the bay is shifting with fewer winter days below 1 C and more winter days above 5 C (Fig. 3). The most pronounced change is in the summer (June, July, August) where we observe a significant (p < 0.0001) increase in the number of days above 23 C (Fig. 3). In the case of a dataset like the Phytoplankton or Fish Survey, where values are collected approximately weekly, changes in the frequency of extremely warm and/or cold measurements further confirm and strengthen the observation that temperature regime of the bay is clearly changing. 3.3. The bottom line: how much has the water temperature increased? Given our three long-term data sets, it is clear that even with the high inter-annual variability, the surface waters in Narragansett Bay have increased over the last five decades (Fig. 1). The magnitude of the overall increase depends on which dataset is used, the length of time analyzed, and how the data are handled (e.g. monthly, seasonal, or annual averages). Since 1960 the annual mean surface water temperature increase per year has been 0.032 C, 0.029 C, 0.027 C based on the temperature data from the Phytoplankton Survey, the Fish Survey, and the NOAA data sets, respectively (Table 1). If we combine all the water temperature data sets together the per year increase is 0.029 C. Thus the surface water temperature in Narragansett Bay has experienced a total rise in temperature of 1.4 C to 1.6 C over the last five decades (Table 1). Winter warming over the past half-century has been especially high. But exactly how high depends on the dataset (Fig. 1a). The estimated 50 year increase ranged from 1.6 C for the Fish
A5
Survey dataset to 2.0 C for the NOAA dataset (Table 1). These updated winter trends are not meaningfully different from one another when accounting for the model variability (as can be seen by the overlapping 95% confidence intervals in Table 1) nor from the value (2.2 C) that begot the “Half-truth” rating and inspired these analyses (Emery, 2013). After 2006 there were a series of colder winters that lessen the average winter temperature warming (Fig. 1a), explaining why the value we report here for the Phytoplankton Survey is slightly lower than that reported in Nixon et al. (2009). Summer temperatures also increase but to a lesser extent and with more variability among the data sets (NOAA: 0.8 C to Phytoplankton Survey: 1.6 C, when averaged across the three summer months: June, July, and August). Additionally, the calculated magnitude of these changes may be driven by unavailable data. For example, 2011 temperature data are missing from the Phytoplankton data set which includes a winter of high temperature according to the Fish Survey data set. As demonstrated in the Nixon et al. (2009) paper, there was substantial variability in the Phytoplankton dataset and this is true for all three datasets discussed here. Such variability is inherent in ecological data. One major driver of variability in temperature data are large scale climatic features such as the North Atlantic Oscillation (NAO). The NAO can modify temperatures in the Northeast United States coastal estuaries and North Atlantic waters (Hurrell and Dickson, 2004; Hurell, 2010) and can seemingly mask temperature trends. Generally temperatures are cooler than average during the negative phase and warmer than average during the positive phase along the east coast of the U.S. During cool negative oscillation periods such as the 1950s and 1960s and after the 2000s, winter temperatures have been colder and snowier than average. During the warm positive oscillation periods such as the 1910s to the 1930s and the 1970s to the 1990s, winter temperatures were warmer than average by up to 2 C with reduced snowfall (Oviatt, 2004). This variation produces a regional oscillation in the annual temperature with little overall change during the negative oscillation period and a net gain in the positive oscillation period. In the positive phase of NAO warm conditions result in the Atlantic Multidecadal Oscillation of warmer than average temperatures in North Atlantic waters ea term coined by Michael Mann (2012). This large pool of warm, salty water that results from the
Table 1 Regression model results for Annual and Winter (D, J, F) changes in temperature over approximately the last five decades for each of the available temperature datasets from Narragansett Bay, RI. Winter is defined as the mean of values from December of the previous year, January, and February. Annual means are for the calendar year. All values are C except for the R2 (%). Due to missing data the number of included years varies. See footnotes for specifics. Month
Model results R2
Intercept
Phytoplankton survey* Annual 37.3 52.8 Winter 24.4 71.9 y Fish survey Annual 19.9 45.6 Winter 39.6 60.7 NOAA newport, RIz Annual 39.4 42.6 Winter 26.0 74.7 All three data sets combinedx Annual 41.8 45.0 Winter 41.1 66.3 *
Estimated 50 year increase Intercept std. Error
Slope
Slope std. Error
Slope 95% lower lim.
Slope 95% upper lim.
Increase ( C)
95% lower lim.
95% upper lim.
12.3 19.4
0.032 0.038
0.006 0.010
0.020 0.018
0.045 0.057
1.6 1.9
1.0 0.9
2.3 2.9
10.2 18.6
0.029 0.032
0.005 0.009
0.019 0.013
0.039 0.051
1.4 1.6
0.9 0.7
1.9 2.5
9.5 19.0
0.027 0.040
0.005 0.010
0.018 0.021
0.037 0.059
1.4 2.0
0.9 1.0
1.9 3.0
5.5 10.9
0.029 0.036
0.003 0.006
0.023 0.025
0.034 0.046
1.4 1.8
1.2 1.2
1.7 2.3
Annual and winter temperature models include data from 1960 to 2010 and exclude 1963 and 1997e1998. Annual temperature model includes data from 1960 to 2012 and excludes 1963 and 1997e1998; winter model includes data from 1960-2011 and includes 1963 and 1997e1998. z Annual temperature model includes data from 1960 to 2012 and excludes 2002 and 2013; winter model includes data from 1960-2011 and excludes 2001. Additionally, we excluded temperatures below 3 for December, January, February and March and above 35 C from November to March; for all months we excluded values below 0 C. x Combined models were constructed using all of the data included as described in footnotes above. y
A6
R.W. Fulweiler et al. / Estuarine, Coastal and Shelf Science 157 (2015) A1eA6
Gulf Stream current carrying warm water northward for several decades, eventually sinks carrying this warmth to as deep as the thermocline at 1000 m during the ensuing negative oscillation phase which cools the surface waters by the following positive phase of the NAO (Cheng and Tung, 2014). The apparent slowdown in global warming has stimulated recent controversy but global warming nonetheless continues, as illustrated here using three separate temperature datasets from Narragansett Bay. 4. Conclusions While the majority of scientists are convinced that human activities are driving recent climate change much of the public remains uncertain. A thorough discussion of the hypotheses proposed to explain widespread public apathy over climate change is well beyond the scope of this paper and our expertise. We see the political discussion associated with Nixon et al. (2009) as a good example of the common disconnect between information produced by the scientific community and how it is perceived by the public, specifically with how best to address the variability of ecological datasets. As Emery (2013) rightly pointed out the amount of surface water temperature increase is a “matter of degree” but we would add a “plus or minus” to that statement. Importantly, the “plus or minus” was driven by comparing different data sets and different time periods. Journalists and scientists have similar goals of better understanding the world around us and then reporting what we learn in a clear, concise, and correct way. Journalists provide a valuable service by fact checking political comments and making scientific findings available to a broader audience. Nevertheless, with their position comes the responsibility of due diligence. As scientists our responsibility lies in making the time to communicate effectively with the public and with journalists and, as this case study illustrates, to do a better job at explaining and couching the inherent ecological variability associated with our averages and summaries. Science e like communication e is an iterative process and thus interpretation of results, most especially those from long-term data sets, will change over time. There is a continued need for open communication between those who do the science and those who have a voice with which to share it. After all, we are all seeking the whole truth. Acknowledgments We want to thank Dr. Theodore Smayda for starting the Phytoplankton Survey and Charles Fish who started the Fish Survey in the 1950s. We also wish to thank Dr. Tatiana Rynearson and Dr. Jeremy Collie who currently maintain the Phytoplankton and Fish Surveys, respectively, as well as those who have come before them and the Graduate School of Oceanography at the University of Rhode Island for funding the survey. These long-term data provide unique and valuable insights into changing coastal marine ecosystems. This report is contribution number ORD-009227 of the U.S. EPA's Office of Research and Development, National Health and Environmental Effects Research Laboratory, Atlantic Ecology Division. Although the information in this document has been funded by the U.S. Environmental Protection Agency, it does not necessarily reflect the views of the Agency and no official endorsement should be
inferred. Mention of trade names or commercial products does not constitute endorsement or recommendation for use. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.ecss.2015.01.021. References Anderegg, W.R., Prall, J.W., Harold, J., Schneider, S.H., 2010. Expert credibility in climate change. Proc. Natl. Acad. Sci. 107, 12107e12109. Balmaseda, M.A., Trenberth, K.E., Kallen, E., 2013. Distinctive climate signals in reanalysis of global ocean heat content. Geophys Res. Lett. 40, 1754e1759. Cheng, X., Tung, K., 2014. Varying planetary heat sink led to global warming slowdown and acceleration. Science 345, 897e903. Crutzen, P.J., 2002. Geology of mankind. Nature 415, 23. Emery, E., 2013. U.S. Sen. Sheldon Whitehouse: Narragansett Bay in R.I. has gotten 4 degrees warmer since the 1960. Providence J.. Fowler, D., Coyle, M., Skiba, U., Sutton, M.A., Cape, J.N., Reis, S., Sheppard, L.J., Jenkins, A., Grizzetti, B., Galloway, J.N., 2013. The global nitrogen cycle in the twenty-first century. Philos. Trans. R. Soc. B: Biol. Sci. 368, 20130164. Gimenez, L., 2011. Exploring mechanisms linking temperature increase and larval phenology: the importance of variance effects. J. Exp. Mar. Biol. Ecol. 400, 227e235. Haberl, H., Schulz, N.B., Plutzar, C., Erb, K.H., Krausmann, F., Loibl, W., Moser, D., Sauberer, N., Weisz, H., Zechmeister, H.G., 2004. Human appropriation of net primary production and species diversity in agricultural landscapes. Agric. Ecosyst. Environ. 102, 213e218. Hooke, R.L., Martín-Duque, J.F., Pedraza, J., 2012. Land transformation by humans: a review. GSA Today 22, 4e10. Hurell, J., 2010. North Atlantic oscillation. In: Turekian, K. (Ed.), Climate and Oceans - a Derivative of the Encyclopedia of Ocean Sciences, second ed. Elsevier, New York, NY, pp. 33e40. Hurrell, J.W., Dickson, R., 2004. Climate variability over the North Atlantic. In: Stensenth, N.C., Ottersen, G., Hurrell, J.W., Belgrano, A. (Eds.), Marine Ecosystems and Climate Variation: the North Atlantic: a Comparative Perspective. Oxford University Press, New York, NY, pp. 15e32. IPCC, 2013. Summary for policymakers. In: Stocker, T.F., Qin, D., Plattner, G.-K., Tignor, M., Allen, S.K., Boschung, J., Nauels, A., Xia, Y., Bex, V., Midgley, P.M. (Eds.), Climate Change 2013: the Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge, United Kingdom and New York, N.Y., USA. Kremer, J.N., Nixon, S.W., 1978. Coastal Marine Ecosystem: Simulation and Analysis, Coastal Marine Ecosystem: Simulation and Analysis. Springer-Verlag. Levitus, S., Antonov, J.I., Boyer, T.P., Stephens, C., 2000. Warming of the world ocean. Science 287, 2225e2229. Mann, M.E., 2012. The Hockey Stick and the Climate Wars: Dispatches from the Front Lines. Columbia University Press, New York, NY. Marcott, S.A., Shakun, J.D., Clark, P.U., Mix, A.C., 2013. A reconstruction of regional and global temperature for the past 11,300 Years. Science 339, 1198e1201. Najjar, R.G., et al., 2010. Potential climate-change impacts on the Chesapeake Bay. Estuar. Coast. Shelf Sci. 86, 1e20. Nixon, S.W., Fulweiler, R.W., Buckley, B.A., Granger, S.L., Nowicki, B.L., Henry, K.M., 2009. The impact of changing climate on phenology, productivity, and benthicepelagic coupling in Narragansett Bay. Estuar. Coast. Shelf Sci. 82, 1e18. Nixon, S.X., Granger, S., Buckley, B.A., Lamont, M., Rowell, B., 2004. A one hundred and seventeen year coastal water temperature record from Woods Hole, Massachusetts. Estuaries 27, 397e404. Oviatt, C.A., 2004. The changing ecology of temperate coastal waters during a warming trend. Estuaries 27, 895e904. Preston, B.L., 2004. Observed winter warming of the Chesapeake Bay estuary (1949e2002): implications for ecosystem management. Environ. Manag. 34 (1), 125e139. Seekell, D.A., Pace, M.L., 2011. Climate change drives warming in the Hudson river estuary, New York (USA). J. Environ. Monit. 13, 2321e2327. Trenberth, K.E., Fasullo, J.T., Branstator, G., Phillips, A.S., 2014. Seasonal aspects of the recent pause in surface warming. Nat. Clim. Change 4, 911e916. Weisberg, S., 1985. Applied Linear Regression, second ed. Wiley & Sons, New York, N.Y.