In the light of change: a mixed methods investigation of climate perceptions and the instrumental record in northern Sweden

The initial study phase was qualitative with recorded informal in-depth interviews with 14 reindeer herders examining their lived experiences of changes in the environment. Sami reindeer herders were chosen through purposive sampling as local northern Sweden population representatives since they were considered the most knowledgeable in the context of climate and weather variability. The study received the consent of the Sami Parliament and approval of the Research Ethics Committee at Umea University (dnr 09_193 § 48/09). All interviews were performed by the first author who is non-Sami, brought up in northern Sweden alongside the Sami reindeer herding society, but without personal involvement with the lifestyle. The study aimed at maximum variation sampling and the interviewees were 3 women and 11 men, 16–75 years of age; geographically dispersed (Fig. 1) with an average of 39 years of reindeer herding experience. A thematic interview guide was used and respondents were asked open-ended questions about experiences of changes to the environment, what they thought about these changes and the future for reindeer herding. The number of interviews proceeded until the data appeared saturated. All interviews were performed during spring 2010, mainly in respondents’ own homes.

The interviews were transcribed verbatim and analyzed using qualitative content analysis. This analysis involves identifying meaningful units to code and sort the data into subcategories, followed by the sorting of ideas generated by the data thereby moving the analyses to a higher interpretive level. Subcategories are merged into categories and one or several core themes emerge (Graneheim & Lundman 2004; Krippendorff 2004). Peer review processes were used during coding procedures as well as during analyses for triangulation purposes and to increase trustworthiness and validity.

In the present mixed methods study, only 13 interviews were used in the analyses, making the mean and median age of the participants 55 and 56 years, respectively. The discarded interview (one woman) did not match the current geographical study area.

The variables for the quantitative study were chosen based on reported changes that have a great impact on the reindeer herders’ daily life and work, implicating a possibly even higher sensitivity to these specific changes. A number of quantitative measurements of climate variability in northern Sweden over the past three decades were selected to visualize the regional trends and to compare instrumental records to respondents’ perceptions. Four specific observations (O1–O4) from the reindeer herding Sami identified in the interview study were adopted for the subsequent quantitative investigation:

Daily measurements of maximum and minimum temperature and snow depth were obtained for 10 stations in northern Sweden and northern Norway (Fig. 1). Stations were selected based on data availability and proximity to the residences of the interview study subjects. Since the period of record for each station varies, data from the period 1978–2007 were used, consistent with the relevant time scale of the interviews (Table 1). These quality-controlled meteorological data were downloaded from the US National Climatic Data Center, accessible via http://www.ncdc.noaa.gov. Separate quantitative analyses were conducted to test each of the Sami observations listed above (O1–O4). The quantitative methods are detailed below. Statistical analyses and graphics were completed in MATLAB version 7.10.0 (The MathWorks, Inc., USA) and R version 2.13.2.

A 21-day-centered moving average smoother to the daily snow depth data was applied and each “snow season” was identified as any period of at least 50 consecutive days in which the smoothed snow depth time series was greater than 25 mm. The moving average was employed to exclude early and late season snowfalls that may precede or follow days with no snow cover, instead focusing on the period of the year in which snow depth consistently was above 25 mm. The snow depth was selected as an indicator for the minimum conditions in which the reindeer herders are able to use skis and snowmobiles when herding. The start date, end date, and duration of the snow season were calculated for each year, and the yearly time series tested for linear trends in each variable. This analysis spanned 28 full winters beginning with winter 1978/1979 and ending in winter 2006/2007.

To determine how wintertime temperatures have changed, linear trends for all percentiles of minimum and maximum temperature were calculated on a station-by-station basis. A percentile-based approach was applied because various physical processes and social activities (and thus, perceptions) may be sensitive to different portions of the temperature distribution (Robeson 2004). Percentiles of minimum and maximum temperature were calculated on an annual basis for each station, excluding any years with gaps of missing data of 2 weeks or more. Least-squares linear regression was then used to examine the trend in each percentile of temperature at each station. In total, 2000 regression slopes and intercepts were calculated (100 per station × 10 stations, maximum and minimum temperature). Trends with p values less than 0.05 were considered significant.

The “cold wave” definition adopted in some temperature health research was employed (Barnett et al. 2011), where a period is defined as one of several days in which temperatures are consistently below a certain value. Daily mean temperature (T_MEAN) was calculated as the average of the minimum and maximum daily temperatures. Cold periods were identified as those in which T_MEAN did not exceed the annual 25th percentile of T_MEAN for at least 10 consecutive days. The total number of cold period days within each year of record was then calculated and a quasi-Poisson regression was fitted to the time series of yearly counts. Sensitivity analysis was performed to test other percentile and duration thresholds.

Three different approaches for assessing if the frequency of “rapid fluctuations” in temperature had changed over the time period were employed. First, trends in the percentiles of daily changes in maximum and minimum temperatures were examined using the regression method presented for O2, but using the variables ΔT_MAX and ΔT_MIN, calculated as T_dayN − T_dayN-1. To facilitate the possibility that the perceived changes in temperature fluctuation might occur over longer time scales, the temperature range for overlapping 5-day periods was also examined. The range for each period was calculated as the highest maximum temperature minus the lowest minimum temperature. The median 5-day temperature range for each month over the period of record along with the 75th, 90th, and 99th percentiles were calculated and linear regression was used to test for significant trends for each percentile and month. Finally, because the condition of the ground and standing water as frozen (or not) is easily observable by—and important to—the herders, trends in the number of days with daily maximum temperatures above freezing and daily maximum temperatures below zero were also investigated. Trends in the frequency of these “near freezing” days might be indicative of conditions that the herders associate with variation or change in the weather and/or unpredictability. At each station, the number of days each year with T_MAX above 1, 2, and 3 °C were tabulated coincident with days with T_MIN below − 1, − 2, and − 3 °C, respectively, and we assessed trends over time with quasi-Poisson regression.

In general, the qualitative study concluded that climate change was just one more stressor that, added to multiple other stressors, were pushing the Sami reindeer herders toward the limit of their abilities to adapt to an evolving physical, social, cultural, ecological, and economic environment. Some respondents even indicated uncertainty about the viability of the reindeer herding lifestyle beyond the present generation. The already present stressors included tourism, hydropower, wind power, bad economic viability, constantly decreasing grazing lands, forestry, the building of new roads, and the very important factor of predator pressure and governmental predator policy,¹ among others.

With respect to environmental changes, respondents identified changes in several different ways and they reported extensive alterations to the weather and climate since the 1970s, which they also perceived to have accelerated during the first decade of the 2000s. The changes included long and warm autumns, later onset of freezing conditions, shorter snow cover season, early and sudden onset of spring and more unstable, unpredictable and variable weather; each of these perceptions is described in more detail below.

Spring was said to now come early and very abruptly, weeks earlier in many places than in the past. The perspective also existed that spring might then come to a halt for many weeks, making the winter pastures perishable and unfitting when the summer pastures are still inappropriate due to rough conditions and heavy snow. The period between winter and summer grazing has always been a difficult time and today that difficulty is exaggerated by the extensive forestry that has reduced old intact forests with arboreal beard lichen (Usnea of the family Parmeliaceae) to a minimum, beard lichen constituting the classic reindeer emergency forage.

In terms of weather conditions, the interviewees experienced and emphasized an increased variability in the weather, making it more unpredictable and unstable. Changes in the weather result in the need to move the herd and more frequent changes hence markedly increase the workload on the herders. Rapid and extreme fluctuations in temperatures and extreme weather were perceived to have become more common and several herders emphasized the increased variability and instability but they also talked about increased weather intensity.

The herders had heard and learned about climate change effects from the media as well as being informed by the authorities about projected changes. Some reindeer herders even perceived projections of future climate change as an even heavier burden than the already tangible effects they had themselves noticed. As pastoralists, the reindeer herders are experts on adaptation and handling change but now articulated that they were coming under severe stress. The herders did not recognize themselves anymore and believed that the old rules of traditional knowledge no longer hold true, as the lived experiences of the older generations become less relevant for today’s herders facing what they perceive as a different weather regime.

In summary, this study illustrates how the winters of northern Sweden have undergone substantial changes during the last three decades and how these specific changes affect reindeer herding. Climate change effects in northern Sweden were investigated through Sami reindeer herder observations combined with 30 years of regional meteorological data. Three out of four tested changes observed for many years by the reindeer herders were consistent with the instrumental record. The contradictory result regarding observation 4—rapid fluctuations in temperature are becoming more common—could indicate that the two methods employed here did not manage to measure the same component of climate change. The concept of “weather variability” is broader than some of the more specific language used in the interviews by the Sami people about temperature patterns and timing and cold spells. Accordingly, there was the least confidence that the quantitative metric matched the perceived trends for this variable out of the four that were tested.

The weather data provided support for the herder notion of a shorter snow cover season and data for two stations exhibited an incredible two full months decrease in snow cover over the investigated 30-year period of time. The snow season grew shorter by more than a full day every year for several stations yet the trend was not significant for all stations examined or all test examined. One station even exhibits a later snow season end date (ARJ). Some herders report a strong spring snow melt delay in the high mountains in a certain area while the overall herder impression was a much earlier spring onset and snow melt. The herders who report a later high mountain snow melt have winter grazing areas in the coastal region corresponding to a weather station with a strong signal of shorter snow cover (LUL). This illustrates that the combination of several changes in different directions of different local or regional areas can work together, further increasing the effect on reindeer herding. If the winter pastures become destroyed a month earlier than usual at the same time as the summer grazing lands stay unfit longer than before, then the already difficult period in the spring including the calving season gets prolonged several weeks with potentially detrimental consequences for the herders and the reindeer.

All herders perceived the winters to be warmer and summers to have changed the least and there was also a more rapid increase in wintertime compared to summertime temperatures in the weather data. There was an overall temperature distribution shift to the right on a horizontal temperature scale and a compression of the distribution curve was also found, reflective of stronger trends in wintertime temperatures than other seasons. Similarly, to the high degree of consistency across the interview subjects in their perceptions of warming winters, nearly every statistical test at every station pointed toward substantial temperature increases in the cold season.

This variable also showed a strong concordance between the two methodologies. One station experienced a 50% reduction of the days in extended cold periods over the study period, yet not all stations showed statistically significant results. The decrease in extended cold periods concurs with the pronounced decline in the coldest percentiles of minimum temperatures. This decrease in cold periods could of course also be perceived as just winter warming; however, the herders’ reports of the means of change comply with decreased cold periods and again the consequences for the herders are problematic.

There was no evidence in support of increased temperature variability in the weather data. On the contrary, a weak signal of the opposite was observed; a decrease in temperature variability, using two of the statistical approaches. Temperature or weather variability more generally may represent an aspect of climate that is perceived based on a combination of multiple various factors interacting with one another, but difficult to measure with the significant-change-in-one-parameter statistical analyses approach. The interacting factors could involve physical and psychological dimensions that were unable to be measured holistically. Tests of the number of days with temperatures fluctuating around zero revealed positive trends at most stations, although significant trends were only observed at one or two depending on the test parameters. It is possible that a change in the climate more related to key biophysical thresholds like the freezing point is what is being perceived by herders when they report increasing fluctuations because the systems on which they live and depend are highly sensitive to those thresholds. Our analysis supports this possibility, additional probing would be required to reach a strong conclusion.

We undertook a mixed methods analysis to learn if and how community perceptions of climatic changes could complement quantitative assessments of trends in temperature and precipitation. Our objective was to use a methodological approach encompassing the complexity, subjectivity, and context-dependent high sensitivity usually associated with qualitative methods; along with the scale, consistency, generalizability, and validity more generally associated with quantitative methods (Agrawal 1995). We sought to create such insights in the context of the northern Sweden Sami reindeer herding populations’ experience to produce a clearer, more societally relevant understanding of how the climate is changing, may change in the future, and what the impacts to people and environment may be in a distinctive and rapidly changing location.

Our analysis revealed statistically significant trends for a majority of the variables tested related to climate change in northern Sweden. The results support the hypothesis that meteorological data support local perceptions. This is an important finding because there is a limited pool of evidence comparing local perceptions to meteorological records regarding climate change, as well as traditional to western knowledge, and we find a fair degree of correspondence between the two. Knowledge of climate change and its impacts can extend beyond the spatial domains covered by meteorological sensors. In places where meteorological sensors and people are collocated, mixed methods research can reveal a richer understanding of the specific mechanisms by which climate change affects society. We suggest that this is especially true for perceptions derived from indigenous peoples. In places where no meteorological sensors exist, the traditional knowledge of local people may provide equally useful information. Other scholars have noted that indigenous peoples are particularly sensitive to environmental changes and may even notice important changes before reported in the formal scientific literature (Greene et al. 2001; Krupnik & Jolly 2002; Riedlinger & Berkes 2001). The lived experience of the indigenous peoples and the inherited knowledge gained over generations from a subsistence lifestyle can provide an excellent early indicator of environmental change. Berkes and Berkes concluded that “Indigenous knowledge appears to bring some unique advantages in dealing with multiple variables and complexity” and described Indigenous peoples’ holistic way of integrating multiple complex parameters as a resemblance of a fuzzy logic system (Berkes & Berkes 2009). The greater magnitude of the reported perceived changes in the qualitative data as compared to the effect sizes described in the quantitative data may indicate such increased sensitivity among the Swedish reindeer herding Sami as well.

We also found that the extent to which local perceptions and meteorological observations aligned was dependent on the specific climatic phenomenon we examined. This is an important finding for future mixed methods research regarding climate change and its impacts. Future scholars could work to develop appropriate quantitative indicators of concepts like “climate variability” and “unpredictability” that were widely perceived as important to study participants but not apparent in the meteorological record as we examined it. Similarly, future research could also strive to explore stressors beyond climate change that may be leading to certain perceptions about how climate is affecting livelihoods.

Our analysis revealed dramatic changes in some climatological variables that mirrored perceptions of stress and susceptibility among interviewees. For example, we found a shortening of the snow cover season by almost two full months over a 30-year period. Similarly, we observed a decline in the number of days in extended cold periods approaching 50% at some stations. These modes of change are significant for reindeer herding. Interviewees described how the decrease in extended cold periods result in the need for maintained preparedness to move the herd and thereby increase the workload. An absence of snow also increase the workload and later formation of lake and river ice hamper and delay the annual migration in the spring and fall. The affected freeze-thaw cycle reduces accessible grazing and improves conditions for predators through the subsequent dispersion of the herds, thereby further increasing the workload as well as the reindeer herding trades’ susceptibility to predator pressure (Furberg et al. 2011). The timing and variance of weather and climate patterns are of particular importance. In polar regions, studies have shown that the observed decrease in snow cover extent and duration is mostly derived from higher temperatures in the spring months resulting in earlier and more rapid snow melt (Derksen & Brown 2011), which has differential societal impacts than if large-scale changes were observed in the fall.

Warming in the Sami region is occurring at an exceptionally rapid rate compared to other places around the globe and the warming is mainly concentrated during the winters (Derksen & Brown 2011; IPCC 2013). As shown in this study, warmer winters have vast effects on the reindeer herding community but will also affect other activities, trades, and industries as well, apart from reindeer herding. Winter sport tourism is a major activity in northern Sweden, skiing both downhill and cross-country as well as snowmobile driving or dog sledding. In a study of the future for Swedish downhill skiing published 2007 using climate change scenarios for 2070 to 2100, researchers estimated the loss of expenditure of a projected 2 months shortening of the snow cover season to be larger than the year 2006 total ski-ticket turnover of the whole Swedish ski industry (Moen & Fredman 2007). The current study has shown that several stations already exhibit a snow cover season shortening of this magnitude, as compared to the 1970s.

Nearly half of the interview subjects commented on their perceived increase in temperature variability. This perception of a shift toward greater temperature variability has been reported by many Arctic as well as non-Arctic Indigenous populations worldwide. In a study of residents of Finland and the Russian peninsula, the interview subjects used terms like “rapid,” “no stability,” “fluctuates,” “drastic,” and “apocalypse” to describe changes they had seen in temperature variability. As an example, here is a quote from reindeer herder for life, now retired, Heikki Hirvasvuopio, in Snowchange (Helander-Renvall & Mustonen 2004):

In light of the current study’s apparent contradiction between a perceived increase in variability and an observed decline in meteorological data variability, future work might consider metrics other than those investigated in this analysis as a means of capturing “relevant” variability to which people are sensitive (Walsh et al. 2005). We cannot be fully confident that the metrics derived to explore the notion of variability in this study are consistent with those reported by interview subjects; scientific and community interpretations of this concept may be quite different. Indigenous peoples in many locations across the world also emphatically report increased weather unpredictability (Berner et al. 2005; Fox 1998; Jolly et al. 2002; McDonald et al. 1997). This particular phenomenon merits further investigation since it has massive impacts on the subsistence lifestyles and it might represent the same—or another, not yet monitored—climate change effect as the increased temperature variability.

In this study, we have provided a new analysis of temperature trends in the region by examining individual percentiles and visualizing those trends in the context of temperature distribution shifts. In the Intergovernmental Panel on Climate Change Fourth Assessment Report (IPCC AR4), there is some discussion of how future temperature distributions might compare to those of today. Some global climate models suggest a shift in the mean value toward higher temperatures and a growing tail at the high end of the distribution, such that record-setting and extreme heat days will increase in frequency at a greater rate than overall temperature increases (Christoph et al. 2004). IPCC AR5 revisited this discussion and reported that, to date, it is unclear if there are substantive changes occurring to the global temperature distribution beyond a shift in the mean. This study did not find any shifts indicative that an increase in the variance of the temperature distribution has occurred in this study region over the past three decades. Instead, the results show a shift toward generally higher temperatures accompanied by a slight compression of the temperature distribution. The entire distribution is shifting to the right, but the left portion of the distribution (associated with low winter temperatures) is moving at a greater rate (see Supplemental Fig. 1). This pattern is different from the asymmetrical trends and implied increased variability reported for 100 stations across Europe since 1975 (Klein Tank & Können 2003), but our study is focusing on a much smaller geographic region.

Some of the specific shifts in temperature distribution that have occurred in northern Sweden are particularly challenging for reindeer herding and the effects of these specific changes become visible through the qualitative data. To facilitate effective adaptation measures, the need for more detailed information together with an interpretation of the expected effects by local experts, in this case, the reindeer herders, is key for success. More detailed weather data monitoring with higher spatial resolution is essential to be able to handle and adapt to current and future change. Most of the planning for adaptation has so far had a national focus; however, people experience climate change impacts at the local scale and the localized effects of climate change vary greatly from place to place. This geographical variability underlines the need for local action both in terms of vulnerability assessments and in terms of the development of adaptation strategies (Measham et al. 2011). Local information on indigenous peoples’ experiences, perceptions, and knowledge is necessary to capture, as a basis to select the right adaptation strategies to increase the overall resilience of local communities. This study contributes to the limited research body that claims policy makers should more explicitly consider indigenous peoples’ reports on perceptions of climate change as a valued resource when developing adaptation policies, plans, and strategies. Given the changes already reported in the scientific literature, we encourage policy makers in the Arctic reindeer herding regions to implement and support national adaptation policies including action plans that offer a sustainable future for reindeer herding. Examples of societal adaptation measures of great importance for the reindeer herders could be secured grazing areas with preserved old forests along migration paths and the provision of safe modes of passage across water reservoirs. If the changing characteristics of winters favor predators, as reported by our interviewees, then taking this fact into account when deciding upon predator population sizes could be another powerful policy measure to support sustainable reindeer herding. The herders’ experience and expertise in adaptation may also help and guide adaptation in other areas of the society.

The impacts of climate change and peoples’ perceptions of climate change might have important public health consequences through causal pathways even more complex than those described elsewhere in this manuscript. For example, public media have covered an increased number of suicides among young male Sami. The researcher AO Eikeland has reported about suicide and mental illness as an Arctic problem (Eikeland 1999). Scholars have reported several underlying causes behind the mental health problems and climate change is one of them (Cunsolo Willox et al. 2015). Our present study clearly shows how climate change poses a threat to the reindeer herding identity and lifestyle. JP Stoor et al. have hypothesized that suicide for Sami in Sweden might have become a way to preserve one’s Sami identity when the identity is being threatened, for example, if one must give up reindeer herding (Stoor et al. 2015). In this context, climate changes then pose direct threats to life. The Lancet Commission on health and climate change published 2015, reported on mental health impacts and anxiety-related responses from climate change through different pathways and about solastalgia (Albrecht et al. 2007), a sense of loss experienced when peoples’ land had been damaged and they had lost resources (Watts et al. 2015). There are a few existing studies on climate change and mental health in the Arctic but there is a need for further research efforts in this context (Cunsolo Willox et al. 2012).

The development of scientific evidence by its nature takes time, whereas peoples’ perceptions of climate change in relation to indigenous knowledge may provide an early signal of change. It has been reported that in small island communities, hydro-meteorological hazards have been precisely predicted by local people, based on their observations of changes in the environment, celestial bodies, and behavior of animals, in relation to the time of the year they happen. These observations were also considered being indicators of other hazards (Hiwasaki et al. 2014). Humans living close to nature observe changes before data analysis reveals statistically significant trends. Research methodology needs to be continuously developed to show complex and elusive phenomena, and mixed methods may capture what exists “between” methods that neither of the methods would have captured if used on their own.

Limitations of this study include classical objections to mixed methods work such as the possibility of not investigating the same phenomena, for example, weather variability, with the two different methodologies. Additionally, from a statistical perspective, this study at first glance committed serious violations with respect to the multiple comparison problem, increasing the probability of making a type I error (Saville 1990). The focus in this study, however, is not on any individual regression slope or statistical measure. Instead, we focused on the broad patterns and effect sizes and signs along with statistical tests. Although we performed many tests without any adjustment for the p value, the multiple comparison problem should not confound the results.

To conclude, this study of climate change effects in northern Sweden illustrates the complementary nature of mixed methods research by demonstrating some of the major changes this region is experiencing through two different lenses and how these changes affect reindeer herding. Increasing temperatures, especially wintertime, along with a decrease in extended cold periods and the snow cover season, are among a suite of social, environmental, and economic stressors that continue to impact the lifestyle the Sami reindeer herders—largely to their detriment. This analysis has demonstrated that, in many cases, the perception of those who most closely interact with the environment provides an accurate representation of climate change as measured by a network of meteorological stations. As such, the local perception of climate change can and should be important in shaping policy and the voice of advocates and policymakers.