The main aim of historical climatologists in recent decades has been to reconstruct the climate on different spatial scales – from global to regional and local – for the last centuries. For this purpose, documentary evidence is used from before the beginning of regular meteorological observations (Pfister et al., 1999; Brázdil et al., 2005), which is usually taken as the year 1850, except for Africa and the Arctic (for which 1890 is the usual start point) (Brönnimann et al., 2019). One type of very useful information about past weather and climate is an early, irregular, isolated series of meteorological measurements. Recently, an initiative was created to inventory all available data of this type (including existing databases) for the whole world, including the Arctic, and then make it available in digital form (Brönnimann et al., 2019; Lundstad et al., 2023). This activity has found recognition and interest among many scientists and has been called “data rescue activity”. For the Arctic as a whole and for its different regions (Canadian Arctic, Greenland, Svalbard and Novaya Zemlya), a summary of existing meteorological measurements was conducted by Przybylak and Vizi (2005), Vinther et al. (2006), Przybylak et al. (2010, 2016, 2018), and Przybylak and Wyszyński (2017). Queries conducted in many European and Canadian archives and libraries allowed the collection of 118 series of average monthly air temperature values from the years 1801–1920 (for details, see Table I in Przybylak et al., 2010), with the greatest amount dating from after the 1st International Polar Year 1882/83. The majority of the gathered series (77.1 %) cover periods of less than two years; however, series of one year or less dominate (58.5 %) (Przybylak et al., 2010).

Meteorological data from prior to 1850 are rare and mainly available for Europe and eastern North America (see Fig. 2a in Brönnimann et al., 2019). According to the map presented in this article (Fig. 2a), such a series for Greenland exists only for the south-western coast. Detailed descriptions of the observations available for this time are provided in papers by Vinther et al. (2006) and, recently, for the late 18th century by Demarée et al. (2020), Demarée and Ogilvie (2021) and Przybylak et al. (2024). According to these publications, the oldest series of observations of about one year or more is that for Neu-Herrnhut (now Nuuk) for the period September 1767–July 1768. All known meteorological observations made here in the late 18th century were usually performed by the Moravian missionaries (for more information about their activity, see, e.g., Lüdecke, 2005, Demarée and Ogilvie, 2008, or Demarée et al., 2020). They are usually available in the form of tables with meteorological data. Some have survived and are available in archives in Germany (Moravian Archives Herrnhut), the UK (The Library and Archives of the Royal Society and Moravian Church Archive and Library in London) and the USA (Moravian Archives Bethlehem) (see Demarée and Ogilvie, 2008 for more details). Others have been included in texts of reports, usually prepared for each month of the year, and most often sent annually by ship to Europe (diaries handwritten in English or Old German). The latter sources, which contain only sparse and irregular measurement data relating to only a few parameters, have been assessed in terms of their utility in application to climate studies (particularly for the study of weather extremes in SW Greenland) by Kodzik (2019) and Borm et al. (2021). Any new data series from the early instrumental period, especially for the Arctic (including Greenland), is very precious for calibrating proxy-based climate reconstructions from so-called natural archives (ice cores, tree rings, lacustrine sediments, etc.). Vinther et al. (2006) document this fact very clearly, showing a strong correlation (0.6–0.7) between reconstructed SW Greenland winter temperatures and ice-core winter-season proxy. Although they did extensive work rescuing a large amount of early instrumental data for SW Greenland, they found neither the reportedly oldest meteorological data series for Nuuk, relating to September 1767–July 1768 (see their Fig. 2, where the series starts at 1784), nor the series analyzed by us for 1806–1813. This latter, entirely new series contains complete data from six years of observations and is the longest series available for the Arctic (including Greenland) for the period before 1840. The Arctic is defined as the region after Atlas Arktiki (Treshnikov, 1985; see also Fig. 1.1 in Przybylak, 2016). For the 1840s and 1850s, a continuous series of data is available for Illulisat (Jacobshavn) (Vinther et al., 2006) and probably also for Lichtenau (1843–1851) and Neu-Herrnhut (1843–1860) according to Table 2 (Lüdecke, 2005). To check the completeness of the last two series, we need to have access to the so-called Lamont collections (kept at the Chair of Ecoclimatology at the Technical University of Munich).

From this brief review of the state of knowledge about early meteorological observations available for Greenland, it is clear that many exist. Despite many years of searches for such observations by many Danish and UK scientists (Vinther et al., 2006) and by Polish researchers and climatologists (Przybylak et al., 2010), no one has managed to obtain information about the existence of a long-term continuous series of daily and sub-daily meteorological observations at Nuuk for the period between 1806 and 1813. While searching the Royal Society archives in London for early instrumental meteorological observations by the Moravian Brethren, we happened upon this series. Therefore, the main goal of this article is to present this newly discovered series of observations to a broader audience of scientists and to present the first results of climate analysis. This analysis is limited, however, to a description of conditions and changes in air temperature in Nuuk at that time (Descriptions of other parameters will follow in future publications.). The secondary goal is to compare air temperatures in the study period against earlier temperature observations from the late 18th century (Przybylak et al., 2024) and with later records, including modern observations from 1991–2020.

The meteorological observations analyzed here were made at the beginning of the 19th century in the territory of present-day Nuuk (Danish name Godthåb, ϕ=64°11′0.49′′ N, λ=51°43′17.65′′ W; see also Fig. 1), the capital and largest city of Greenland. Currently, they are available in the manuscript MA/154 in the Library and Archives of the Royal Society in London. The title of the manuscript is: Meteorological observations at Godthåb [Nuuk], Greenland, by Charles Lewis Giesecke. More details about this source are available at https://catalogues.royalsociety.org/CalmView/Record.aspx?src=CalmView.Catalog&id=MA%2f154&pos=1 (last access: 1 August 2025). According to the title page of the manuscript and the Royal Society catalogue describing the source, the scientist responsible for making them was the German mineralogist Dr. Charles Lewis Giesecke (born Johann Georg Metzler [1761–1833]) (Fig. 2). However, given the numerous biographical accounts of his stay in Greenland (e.g., Monaghan, 1993; Jørgensen, 1996; Wyse Jackson, 1996; Whittaker, 2001), there is some doubt as to whether he conducted the meteorological observations alone; he may have had help from the local Inuit community and Moravian missionaries.

Dr. Charles Lewis Giesecke arrived at Friederichshaab in Greenland on 31 May 1806 (Whittaker, 2001). The primary purpose of his research visit and stay in SW Greenland, which was planned to last for one to two years, was to conduct a geological reconnaissance, particularly in search of new minerals. We know that the meteorological observations he began on 1 November 1806, continued regularly until April 1807. From then until August of the same year, the observations were intermittent and irregular. Therefore, he likely carried out most of his geological research from June to October 1806 and then from May to July 1807. There is also evidence that he conducted studies of minerals in the summers of 1808 and 1809 (Whittaker, 2001); however, these trips were likely shorter. During this time, he organized research expeditions along the south-western coast of Greenland, from Upernavik in the north to Cape Farewell in the south (Whittaker, 2001). To reach these places, he mostly used boats, but also sledges and travelled on foot. Due to the onset of the Napoleonic War in 1807, much of Europe was occupied by French troops. Thus, Giesecke could not return to Denmark and decided to stay in Greenland until the political situation changed and allowed him to return (Whittaker, 2001). This fact likely led to the initiation of systematic meteorological observations, which were conducted over six years (August 1807–July 1813), resulting in the almost-complete series of observations made three times a day (morning, midday and evening). The gaps occur in September 1807 and cover 18th–19th for all measurement times, plus 22nd–30th for midday measurements and 21st–30th for evening. The gaps were filled using data from the same days of the year, but from the years 1808–1812. These data were used to calculate the differences between noon/evening observations and morning observations – for which data from September 1807 are available. The resulting five-year average differences were added to the morning observation data from the period of 20–30 September 1807. On the other hand, missing data from 18 and 19 September were reconstructed using average values calculated from ten temperature values measured at the same observation time on adjacent days. The data used in this study were taken from the manuscript MA/154 found in the Archives of the Royal Society in London (Fig. 3). The original manuscript is available in Copenhagen Diamond Library (Mineralogisches Reisejournal über Grönland gehalten 1806–1813. (I Nr. 323 a: Tägliche Wetter-Beobachtungen in Godthaab von 1 November 1806–16 August 1813: 323 b: Nogle Ord om og til Grønlands Opkomst Hans Kgl. Majest. allerunderdanigst tilegenet; ogsaa paa Tysk.) 9 Bd. Karl Ludwig Giesecke 1761–1833 mineralog (DBL) 1806–1813. As mentioned above, it is possible that, for these measurements, especially during his exploratory travels, he involved Inuits and (especially) the Moravian missionaries. These latter were already present in the area and had extensive experience in conducting meteorological observations (see e.g., Demarée and Ogilvie, 2008; Przybylak et al., 2024).

As Fig. 3 shows, Giesecke conducted measurements and observations of the following meteorological variables: atmospheric pressure (morning and evening), air temperature (morning, midday and evening) and wind direction (morning and evening). The precise times of observations are not given. In addition to the meteorological data available in the meteorological registers, he provided brief descriptions of the weather conditions for each day. At the end of each month, he also included a short summary of the weather conditions (at the bottom of the table of meteorological data). Unfortunately, the register does not provide information about the units used for the measurements or details about the thermometer's exposure. It is assumed that the thermometer was placed on the north-facing wall outside Giesecke's building. However, in the Arctic, where there is polar day during summer months, such placement does not eliminate the influence of solar radiation unless the thermometer is adequately shielded. The detailed location of the place where observations were probably done (whether it was near the sea or at some distance from the sea, its elevation, etc.) is also unknown.

In the article, we present an analysis limited to air temperature – the most important variable in climate analysis in the polar regions – influencing most strongly both environmental changes and human life in these regions. We first verified all transcribed data against the source data. In the next step, these data were quality-controlled first by (i) visual inspection and then (ii) by the method recommended by WMO (Aguilar et al., 2003). All sub-daily series of temperature (morning, midday and evening) were analyzed, looking for data exceeding 3–4 SD. We found only a few such cases, all of which we analyzed using air pressure and wind direction (atmospheric circulation) in terms of the probability of occurrence (determining the cause and consistency with the course of the aforementioned meteorological elements). For this purpose, we also used the descriptions written by Giesecke, who regulary indicated high/low values and large day-to-day fluctuations (and explained them, giving the reasons). We did not find any erroneous temperature values. We also calculated the differences between observations taken at noon and in the morning, at noon and in the evening, and in the evening and in the morning. We also checked several values exceeding 3–4 SD, and all of them were deemed correct. We performed a similar test for day-to-day changes and again did not find any changes that aroused doubts. All these methods demonstrated that the data are of good quality and homogeneous. Our comparison with present values (see results) also support this favourable assessment, as did comparison with parallel observations available for Nuuk for the years 1811–1812.

As we wrote earlier, the unit in which air temperature measurements are presented in the meteorological register is not given. To determine which unit was likely used in Giesecke's measurements, we compared his results with parallel temperature observations available for another location in Godthåb during 1811–1812 (Vinther et al., 2006). Based on these, we concluded that temperatures were measured in degrees Celsius. Although the exact times of air temperature measurements are unknown, we calculated mean daily air temperature (MDAT) using a weighted average: (Tmorning + Tmidday + 2 × Tevening)/4. To assess the probable biases in the calculation of MDATs, we used hourly air temperature data from the Nuuk 4250 station from 2010 to 2020 (Drost Jensen, 2022). We calculated MDATs according to eight different formulas: 1MDAT1=T1+T2+T3,…,T24242MDAT2=T6+T12+T1833MDAT3=T7+T13+T1934MDAT4=T8+T14+T2035MDAT5=T8+T14+2⋅T2046MDAT6=T8+T14+2⋅T2147MDAT7=T6+T12+2⋅T2048MDAT8=T6+T12+2⋅T214 where MDAT1 is calculated from 24 hourly measurements per day (the “real daily average”).

In the next step, we compared the MDAT1 results against those obtained using all other formulas (MDAT2–MDAT8, which rely on only three measurement times per day) (see Table 1). Analysis shows that, in the cold half-year (October–March), the maximum bias reaches 0.1 °C, whereas in summer the bias is highest but does not exceed 0.5 °C. The formulas MDAT6, MDAT7, and MDAT8 are closer to MDAT1, with a bias typically of 0.0 °C, except in June, July, and August, when the bias reaches a maximum of 0.2 °C.

None of the possible biases in air temperature measurements that are associated with the type and accuracy of the thermometer used or its exposure can be estimated due to the lack of relevant information. However, the exposure bias should be minimal or zero for the polar night due to the absence of solar radiation at such times. However, we must add here that scientists at that time (approximately the late 18th century) were familiar with methods of meteorological measurement (see e.g., Middleton, 1969; Frisinger, 1977; Camuffo et al., 2017; Pfister et al., 2019). Knowledge about the appropriate placement and protection of thermometers was common. During the period under study, the temperature in Nuuk never dropped below 38.8 °C (the freezing point of mercury). This fact means that the freezing of mercury cannot have been a cause of measurement errors. We still cannot rule out the existence of minor errors that may have been committed by the person copying the original data from the source mentioned earlier, held at the archive in Copenhagen.

The MDAT values, which are available at 10.18150/IGYNGV (Przybylak et al., 2025), were used to calculate standard (monthly, seasonal and annual means) and less-typical climate statistics (indices) such as: (1) day-to-day temperature variability (DDTV); (2) frequency of occurrence of MDAT in 1-degree intervals, including the calculations of skewness (γ1) and kurtosis (γ2) of analyzed sets of air temperature using formulas recommended by Von Storch and Zwiers (1999); (3) thermal seasons after Baranowski (1968) proposition for polar regions (for details, see Przybylak et al., 2024); (4) thermal roses of wind (relation between wind direction and temperature); and (5) annual air temperature range (ATR) and thermal climate continentality (K) using the following formula proposed by Ewert (1972): 9K=ATR-3.81⋅sinφ+0.138.39⋅sinφ+7.47⋅100% where ATR is the annual air temperature range calculated as the difference between the mean temperatures of the warmest and coldest months, and ϕ is the geographical latitude.

This index considers the dependence of ATR not only on geographical latitude, but also on the percentage of land in a given latitudinal band. The K index changes worldwide from -1.5 % for areas with extreme oceanic climate to more than 140 % for the extremely continental part of Eastern Siberia (Ewert, 1997). For more details about the methods used to calculate the statistics mentioned above, see Przybylak et al. (2014, 2024).

The obtained results for the study period were compared against earlier temperature observations from the late 18th century (Przybylak et al., 2024) and with later records (Vinther et al., 2006), including modern observations from 1991–2020 (monthly data after Cappelen and Drost Jensen, 2021, and hourly data, based on which daily averages were calculated, after Drost Jensen, 2022). However, due to the lack of complete hourly data for the 4250 Nuuk (Fig. 1) station and the inability to calculate daily averages from 24 measurements for the whole period 1991–2020, the gaps in the daily averages were filled using an interpolation method (Nordli et al., 2020) based on data taken from the neighbouring 4254 Mittarfik Nuuk station (for more details see Przybylak et al., 2024, p. 1455).

Comparison of the same years 1807–1813 was only possible with data taken from reanalyses: the NOAA/CIRES/DOE 20th Century Reanalysis (V3) (20CRv3) available for 1806–2015 (https://www.psl.noaa.gov/data/gridded/data.20thC_ReanV3.html, last access: 20 August 2025) (see also Slivinski et al., 2019, 2021) and the Modern Era Reanalysis for 1421–2008 (ModE-RA; Valler et al., 2024) available via ClimeApp (Warren et al., 2024, https://mode-ra.unibe.ch/climeapp/, last access: 21 August 2025).

This newly discovered series of meteorological measurements in Nuuk (SW Greenland), encompassing a complete series of observations from August 1807 to July 1813, is very unique and therefore very valuable for the study of climate in this part of the Arctic. Analyses of the history of early instrumental meteorological observations, conducted by Przybylak et al. (2010) for the entire Arctic and by Vinther et al. (2006) for Greenland, lead to the conclusion that the newly discovered series is the longest which exists for the Arctic prior to the 1840s. Its usefulness, apart from providing detailed insight into the climatic conditions in SW Greenland at the beginning of the 19th century, is much broader. The data can be used for the calibration of, among others: (i) climate reconstructions based on various proxy data; (ii) climate reconstructions based on models; and (iii) climate reconstructions from the NOAA/CIRES/DOE 20th Century Reanalysis (V3) (20CRv3) available for 1806–2015 (see also Slivinski et al. 2019, 2021). Slivinski et al. (2021) documented that the 20CRv3 is a good tool for reconstructing even extreme weather and climate events, and we therefore plan to use this reanalysis to study (in a separate paper) the reason for the described great warming from 25 to 26 February 1812.

Better knowledge about the climate in Greenland at the beginning of the 19th century can help explain the reasons for very cold spells that have occurred on this island in the past millennium. We documented in the Results section that the temperature in the study period was considerably lower than in the contemporary period (1991–2020). We also checked how this period compares thermally with other available instrumental data (the newest and the oldest). For this purpose, we calculated 30-year average temperatures (August 1870–July 1900, August 1900–July 1930, ..., August 1990–July 2020), starting from the beginning of the available continuous series of meteorological measurements in Nuuk since 1866 (Przybylak, 2000; Vinther et al., 2006) (Fig. 13). Moreover, in order to have periods that are more comparable in length with the study period, we also chose, from the entire mentioned series (1870–2020), two six-year periods representing the periods that were the warmest (1926–1932) and the coldest (1881–1887) and simultaneously described the range of thermal changes in 1870–2020 (Fig. 14a). For Nuuk, for the period before 1870, a complete five-year series of temperature data (1816–1820) is also available that was gathered by Vinther et al. (2006). It is worth adding that this series includes the year 1816 (called the “year without summer”), which was one of the coldest years in all the Northern Hemisphere (due to the Tambora eruption, see Raible et al., 2016). We used this 1816–1820 series for comparison against the thermal state of the period 1807–1813. In addition, we also added a curve to Fig. 14a that represents the mean yearly cycle of temperature based on data available for the period 1784–1792 (after Przybylak et al. 2024). How the available datasets from reanalyses (20CRv3 and ModE-RA) reconstruct the surface air temperature is shown in Figs. 5 and 14b. Figures 13 and 14 clearly indicate that the period under study was extremely cold. It was significantly colder than: (i) all 30-year periods; (ii) the coldest six-year period in the series 1871–2020; (iii) the late 18th century; and (iv) reconstructed temperature for the period 1807–1813 by 20CRv3. The period 1807–1813 was also colder (but only slightly) than the five-year period 1816–1820. Only the ModE-RA data are significantly colder than the measured data from October to February, while for the remaining months they are similar (Fig. 14b). Vinther et al. (2006) concluded that the decade 1810 was the coldest in the entire record available for Nuuk, i.e. since 1784. They have at their disposal data from the mentioned five-period as well as for the years 1811 and 1812. The data come, however, from another locality than the place of meteorological observations made by Giesecke. A comparison of data for the latter two years (1811 and 1812) presented by Vinther et al. (2006) and in this study is shown in Fig. S2. There is a good coherence between monthly means from June to December, but in the rest of the year, the differences are quite large. The reasons for these are difficult to explain, but they may be connected mainly with the localities of measurement places being different. Probably, the data presented by Vinther et al. (2006) comes from a site located near the coast, while Giesecke's site is assumed to have been located at some distance from the coast and probably also at a higher altitude. Brohan et al. (2010) studied Arctic marine data from the seas surrounding the south part of Greenland in the period 1810–1825 and found that this period was also “significantly cooler than that of today”.

The question arises as to whether and to what extent this cold period in SW Greenland at the beginning of the 19th century is correctly reflected in air temperature reconstructions based on proxy data and computer climate simulations. Unfortunately, in the area of Nuuk, there does not exist any series of proxy data of value for the reconstruction of air temperature; the closest according to Kaufman et al. (2009) and McKay and Kaufmann (2014) is located in Lake Braya Sø, lying about 313 km to the north (ϕ=67°00′ N, λ=50°42′ W). Lake sediments allow for reconstructing summer temperature only, according to Table 1 in McKay and Kaufman (2014), but the reconstruction is not shown. Further north in the Ilulissat region, there are other lakes for which proxy data are available, and it is possible to reconstruct July temperature based on chironomids present in sediments (Axford et al., 2013), which were not included in the paper written by McKay and Kaufman (2014). The temperature reconstructions presented in Axford et al. (2013) are not of sufficient resolution to be used for comparing short-term temperature changes in the last millennium. Most of the proxy data for Greenland (and the most well-known) come from the Greenland Ice Sheet. According to the information given in Table 1 (McKay and Kaufman, 2014), ten series of proxy data exist in a PAGES Arctic 2k database published in 2013 by PAGES 2k Consortium. Two more such data series can be found in the publication of Jones and Mann (2004), who used data from Fisher et al. (1996) and Vinther et al. (2003). All of them use ice-core data (δ18O) and allow for the reconstruction of annual mean temperatures. Recently, two series of annual air temperatures for central (Kobashi et al., 2010) and central-northern (Hörhold et al., 2023) Greenland for the last millennium were reconstructed using ice-core data, i.e. isotope ratios of nitrogen (¹⁵N/¹⁴N) and argon (⁴⁰Ar/³⁶Ar) in air bubbles and δ18O, respectively. According to Kobashi et al. (2010), the beginning of the 19th century saw one of the coldest temperatures in the last millennium; lower temperatures were noted only in two short periods in the 18th century and in the first half of the 17th century. The reconstruction presented in Fig. 1 of Hörhold et al. (2023) shows a picture for central-northern Greenland that differs only slightly. It shows four periods of similar size cooling in the last millennium, including one in the first half of the 19th century. It also shows that the mean annual air temperature at the beginning of the 19th century was about 3 °C colder than the present time and about 2 °C colder than medieval times. Using instrumental data from Nuuk from the periods 1807–1813 and 1991–2020, we found a 3.3 °C increase in annual air temperature from historical to present times. This indicates a very good agreement between the measured and temperatures reconstructed using ice-core data. As we mentioned earlier, such a good correlation (r=0.67 and r=0.60 for the periods 1785–1872 and 1873–1970, respectively) between the extended SW Greenland temperature series and the ice-core was also found by Vinther et al. (2006). Hörhold et al. (2023) found that the Arctic 2k reconstruction after McKay and Kaufman (2014), which represents large parts of the higher Arctic circumpolar region, shows a low correlation over Greenland. Nevertheless, the greatest cooling in the last millennium, according to the Arctic 2k reconstruction, evidently occurred at the beginning of the 19th century. This is particularly clear for the revised PAGES 2k Arctic reconstruction (see Fig. 2 in McKay and Kaufman, 2014). According to that reconstruction, the mean annual Arctic air temperature at the beginning of the 19th century was about 2 °C colder than at present. Arctic-wide summer-weighted annual temperature reconstructed for the last 400 years by Overpeck et al. (1997, see their Fig. 3) also shows that a great cooling started at the beginning of the 19th century but peaked in the 1840s. The temperature at this time was about 1.5 °C colder than in the late 20th century. Also, the spatially resolved two-millennium summer (June–August) temperature reconstructed for the Arctic and sub-Arctic domain (north of 60° N) calculated by Miller et al. (2010) reveals a great cooling in the first half of the 19th century, being one of the three coolest episodes in the two millennia analyzed. In comparison to the present, the temperature in the first half of the 19th century was about 2 °C colder (see their Fig. 2). This reconstruction is very close to the reconstruction of mean annual air temperatures presented by McKay and Kaufman (2014); the latter, however, shows a cooling that was greater and occurred earlier than in the case of reconstructed summer temperatures (see Fig. 3 in Miller et al. 2018). A two-millennium summer (June–August) temperature reconstruction over the Arctic and sub-Arctic domain (north of 60° N) is also presented by Werner et al. (2018). They used a set of 44 annually dated temperature-sensitive proxy archives of various types from the PAGES 2k database, revised in 2017. Also, in this reconstruction, a cooling at the beginning of the 19th century is evident that is the greatest of the entire 2000 years. Werner et al. (2018) compared their reconstruction with the other six reconstructions published in recent years and found their reconstruction to be very close to that presented in the work of McKay and Kaufman (2014). The summer cooling, however, according to this reconstruction (see Fig. 3 in Werner et al., 2018), was about 0.5 °C smaller than McKay and Kaufman's reconstruction. Other reconstructions presented by Werner et al. (2018) in their Fig. 3 (i.e., Kaufman et al., 2009; Shi et al., 2012; Hanhijärvi et al., 2013; Tingley and Huybers, 2013) also confirm the occurrence of significant cooling, except the last-cited paper. In conclusion, we can state that the cooling in the first half of the 19th century was common throughout the Arctic, including Greenland.

Let us now check how climate models reconstruct this period in Greenland and the Arctic. According to the annual mean surface temperature reconstructed for the Arctic using an ensemble of five simulations performed with a global three-dimensional, atmosphere-sea-ice-ocean model driven by both natural (solar and volcanic) and anthropogenic (increase in greenhouse gas concentrations and tropospheric aerosols) forcings, the last millennium featured four periods of great cooling. These were in the mid-11th century, the second halves of the 15th and the 17th centuries, and the beginning of the 19th century (see Fig. 1c in Goose and Renssen, 2003). The last two cooling periods simulated by the model occurred in the years 1670–1700 and 1800–1830, when surface temperatures were about 0.5 °C colder than the mean in the Arctic from the period 1000–1850. Another modelling work, presented by Crespin et al. (2013), using the three-dimensional Earth system model of intermediate complexity LOVECLIM forced by changes in solar irradiance, volcanic activity, land use, greenhouse gas concentrations and orbital parameters, also shows a very cold period in the first half of the 19th century. This was the coldest period in the entire last millennium. The temperature difference in comparison to present times reached about 1.8 °C. Figure 1.4 in Crespin (2014) shows that simulations of annual mean Arctic temperature in five different GCM models (CCSM4 [USA], GISS-E2-R [USA], IPSL-CM5A-LR [France], MPIESM-P [Germany], and Bcc-csm1-1 [China]) confirm the existence of cold period in the first half of the 19th century, but this was not always found to be the coldest period in the past millennium.

According to investigations conducted by Crespin et al. (2013) and Crespin (2014), the reasons for this great cold period at the beginning of the 19th century were mainly intense volcanic activity, in particular in the period 1810–1840, and a smaller-scale decrease in solar irradiance connected with the Dalton minimum occurring in 1790–1830 (most intensely in 1797–1827; Silverman and Hayakawa, 2021) as well as changes in land use. Volcanic eruptions were the main reason for the cooling that occurred in the 1810s (in particular, in 1811 and 1817–1818), according to Vinther et al. (2006). They connected these two cold spans with eruptions of an unidentified volcano around 1809 and the very well-known eruption of the Tambora volcano in 1815, respectively. Our results, presented in Fig. 5, confirm the existing finding of Vinther et al. (2006), i.e. the existence of very cold temperatures in the first years of the 1810s. According to our data, a cooling occurred after 1809 that was initially small in winter 1809/10 and then significantly greater in the following three winters, 1810/11–1812/13. This cooling is also observed in data from palaeoreanalysis (ModE-RA), but not in the data from the 20th Century Reanalysis (20CRv3). In summer, some cooling is evident only in 1811. As a result, the coldest expedition year after a volcanic eruption in 1809 occurred in 1810/11. Data series presented in Fig. 5 (except 20CRv3) also confirm Schneider et al.'s (2017) finding that, after the 1809 event, temperatures remain low until the Mt. Tambora eruption of 1815. Changes in atmospheric circulation that occurred after this eruption, from a negative to a positive NAO indices in wintertime (see Luterbacher et al., 2002a, b), caused a change in wind directions in SW Greenland, (i.e., domination of winds from the northern sector, see Fig. 6). Positive phase of the NAO cause a significant decrease (3–6 °C) of winter 1810/11 air temperature in the western part of Greenland in contemporary series of data, particularly in its SW part (see Fig. 12 in Przybylak 2000). A similar finding was recently presented by Faust et al. (2025; see their Fig. 4) for south-west Greenland, based on temperature reconstruction using a high-resolution sedimentary record for the late Holocene. Box (2002), investigating air temperature variability in Greenland from 1873 to 2001, found that the greatest anomalies during this period were also linked to large volcanic eruptions, sea-ice extent, and the NAO. Brohan et al. (2010), analyzing marine data extracted from the reports of the noted whaling captain William Scoresby Jr (Greenland Sea) and from the records of a series of Royal Navy expeditions to the Arctic preserved in the UK National Archives, concluded that the cold early-19th-century climate was linked with low solar activity and followed a series of large volcanic eruptions.

Summarizing the obtained results, we can again underline that the first two decades of the 19th century in the study region, in all of Greenland and on average in the Arctic, were, according to reconstructions based on proxy data (Miller et al., 2010; Werner et al., 2018), one of the coldest periods in the past two millennia (and possibly the coldest). It is also probable that such a large scale of cooling and its durability are one of the main reasons that almost all available reconstructions of air temperatures using different proxy data or using climate models for this purpose are almost fully consistent with the available meteorological observations for this period. For example, such good correlations between instrumental temperature data and different reconstructions are not observed for analyses encompassing the earlier period (1784–1792) for SW Greenland (for details, see Przybylak et al., 2024). The late 18th century in SW Greenland and many places in the Arctic was warm, but the warming in this time was not so spectacular as the cooling that occurred in the Arctic at the beginning of the 19th century. Also, a good agreement exists concerning the reasons for the cooling of the early 19th century in Greenland and the Arctic. The two main reasons underlined by most authors studying this issue are mainly intense volcanic activity and, to a lesser degree, low solar activity connected with the Dalton minimum. In addition, land use changes in mid-latitudes are also mentioned by some researchers. Changes in atmospheric circulations described using the NAO must also be taken into account in the case of SW Greenland's climate.

There are many advantages of instrumental observations in comparison to the possible reconstruction of climate using proxy data and simulations by climate models. Reconstructions are less precise, have smaller resolution and are limited most often to (i) only some parts of the year (mainly to summer) and (ii) one main climate variable (air temperature). For this reason, searching for new meteorological data series in archives and libraries is extremely important and necessary. It seems that it should still be intensified despite the fact that, for the last approximately 30–40 years, this type of activity has been carried out intensely worldwide. The newly discovered unique series of meteorological observations for SW Greenland for the period 1807–1813 that is presented in this article proves that new data of this type can still be found for various parts of the globe, including the Arctic, where they are many times more valuable than comparable data for moderate latitudes (for which the network of stations is significantly denser). Unfortunately, fewer scientists are involved in this type of data rescue activity for the Arctic at present than in previous times.