Climatology and time series of surface meteorology in Ny-Ålesund , Svalbard

A consistent meteorological dataset of the Arctic site Ny-Ålesund (11.9 ◦ E, 78.9◦ N) spanning the 18 yr-period 1 August 1993 to 31 July 2011 is presented. Instrumentation and data handling of temperature, humidity, wind and pressure measurements are described in detail. Monthly mean values are shown for all years to illustrate the interannual variability of the di fferent parameters. Climatological mean values are given for temperature, humidity and pressure. From the climatological dataset, we also present the time series of annual mean temperature and humidity, revealing a temperature increase of +1.35 K per decade and an increase in water vapor mixing ratio of +0.22 g kg−1 per decade for the given time period, respectively. With the continuation of the presented measurements, the Ny-Ålesund high resolution time series will provide a reliable source to monitor Arctic change and retrieve trends in the future. The relevant data are provided in high temporal resolution as averages over 5 (1) min before (after) 14 July 1998, respectively, placed on the PANGAEA repository ( doi:10.1594/PANGAEA.793046). While 6 hourly synoptic observations in Ny-Ålesund by the Norwegian Meteorological Institute reach back to 1974 (Førland et al., 2011), the meteorological data presented here cover a shorter time period, but their high temporal resolution will be of value for atmospheric process studies on shorter time scales.

the presented measurements, the Ny-Ålesund high resolution time series will provide a reliable source to monitor Arctic change and retrieve trends in the future.
The relevant data are provided in high temporal resolution as averages over 5 [1] min before [after] 14 July 1998, respectively, placed on the PANGAEA repository (http://doi. pangaea.de/10.1594/PANGAEA.793046). 15 While synoptic observations by the Norwegian Meteorological Institute reach back to 1935 (Førland et al., 2011), the meteorological data presented here cover a shorter time period, but their high temporal resolution will be of value for atmospheric process studies on shorter time scales.

20
On the background of global warming, temperature increase in the Arctic has been found to be nearly twice as large as the global average (IPCC, 2007). A variety of processes have been proposed to contribute to this amplified Arctic warming. Most prominent is the ice-albedo feedback mechanism associated with the current retreat and thinning of Arctic sea ice and resulting changes in surface heat fluxes (Lindsay and ESSDD 5, 1057-1076 Climatology and time series of surface meteorology Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | cover and water vapour (Schweiger et al., 2008;Wu and Lee, 2012), and increased meridional heat transport to the northern high latitudes related to changes in cyclone behavior and atmospheric circulation patterns (Graversen et al., 2008;Zhang et al., 2008). As the Arctic sea ice cover decline is occurring faster than most of the climate models 5 predict (Stroeve et al., 2007), it is even more important to build on reliable atmospheric observation data from the Arctic region. At the Arctic site Ny-Ålesund (78.9 • N, 11.9 • E), the Alfred Wegener Institute for Polar and Marine Research (AWI) operates surface radiation measurements as part of the Baseline Surface Radiation Network (BSRN) since August 1992. Collocated me-10 teorological surface observations were initiated in August 1993, providing important auxiliary information to the radiation measurements (Kupfer, 2006). Considering Arctic climate change, these meteorological data meanwhile gain value as stand-alone climate dataset.
Here, we present a thorough description of the meteorological surface measure-15 ments in the period 1 August 1993 to 31 July 2011. These 18 yr of observational data form the basis for a Ny-Ålesund atmospheric surface climatology.

Ny-Ålesund surface meteorology dataset description
Ny-Ålesund is located on the west coast of Svalbard (Spitsbergen), one of the northernmost archipelagos in the Arctic. The village is an international center for various 20 modern Arctic research activities, situated at a fjord surrounded by glaciers, moraines, rivers, mountains, and a typical tundra system. With the special orography of surrounding mountains causing a tunnel effect along the Kongsfjord, Ny-Ålesund may not be a representative location for the Arctic in general, but surely indicates Arctic variability and change when looking at synoptic time scales. Its high latitudinal position implies 25 polar night condition between 24 October and 18 February, and polar day between 18 April and 24 August, respectively. During the summer period, the terrain is snow-free, with the melt season starting around May/June, the onset of melting depending on the individual synoptic situation of each year. Snow-fall can occur also in summer, but the snow cover usually remains stable only after October. During recent winters, there have been several heavy precipitation periods with rain rather than snow.

5
In the southern part of Ny-Ålesund, AWI operates a 10 m meteorological tower, mounted on a measurement field with soft tundra ground. The horizontal distance to the closest building is about 220 m, while a slightly elevated street -with no trafficis about 5 m to the other direction. To avoid effects of the growing village activities and to reduce potential effects of the close by street ridge, the meteorological tower has been shifted by 40 m in August 2011. In this climatology, we refer to data registered before this relocation to avoid potential effects on the dataset homogeneity. A detailed description of the different meteorological sensors and their time series is given below. A common feature of all parameters is the adjustment of the 5-min data registration to 1-min measurement intervals that was adapted since 14 July 1998 with a renewal of 15 the data acquisition hardware.
Generally, the meteorological tower is inspected at least once per day by the observatory personnel to check for disturbances of the measurement, e.g. icing of the radiation shields or mechanical damage from reindeer or birds.

20
In 2 m altitude above the solid surface, a PT-100 resistance thermometer is mounted to monitor surface air temperature. The instrument is a product of the company "Thies Clima", with a double radiation shield and a built-in ventilator providing an air flow of 3.5 m s −1 . The temperature range −30 • C...+80 • C is covered with an accuracy of ±0. due to malfunction, while on 23 November 1998, only the ventilator has been replaced. The instrument has been measuring in this configuration ever since. For all meteorological parameters in this study, monthly mean values are calculated from daily mean values which are retrieved using the hourly mean values. Missing values in the observational data (e.g. due to malfunction of the instrument) are handled 5 depending on their duration. If more than 50 % (30 min) of data are missing within 1 h, the hourly mean is considered a missing hourly mean value. If more than 21 % (5 h) of hourly mean values are missing within 1 day, the daily mean is considered a missing daily mean value. And respectively, if more than 12 % (∼ 4 days) of daily mean values are missing, the monthly mean is considered a missing monthly mean value.

10
Considering these numbers, no missing monthly mean values have to be taken into account for the surface air temperature climatology.
In Fig. 1a, the monthly mean surface air temperature is shown for all years included in this study. Obviously, the inter-annual variability is much larger in winter than in summer, e.g. the mean January temperature ranges from −17.0...−3.8 • C (∆T mean = 15 13.2 K), while the mean July temperature ranges from 4.6...6.9 • C (∆T mean = 2.3 K). This difference in variability is also represented in Fig. 1b, showing the monthly mean values averaged over the whole time period along with the standard deviation. Usually, the summer months with monthly mean surface air temperature above the frost point are snow cover free. Yet, the minimum temperature shown in Figure 1b indicates that 20 snowfall is possible also during the summer months, while the maximum temperature points to the possibility of rainfall during the winter months. The climatological monthly mean surface air temperatures are listed in Table 1.
On the same meteorological tower, temperature has been measured also in 10 m altitude with an identical sensor. A replacement of this instrument or part of it occurred

Surface atmospheric humidity (2 m)
Humidity has been measured by two parallel sensors, both installed in 2 m altitude above ground. While the first years, both humidity measurements were performed with a hair hygrometer, one of them (hereafter called RF1) was exchanged with a humicap sensor on 25 October 1999. After almost 3 yr of parallel humicap and hair hygrometer 5 measurements, the second humidity instrument (hereafter called RF2) also was exchanged with a humicap sensor on 28 July 2002. The hair hygrometer instruments were Thies Clima hygro-transmitters with a weather and radiation shield. The measuring element was hair with a specific treatment to obtain the desired hygrometric characteristics, explicitly suitable for temperatures below 10 freezing point. The measuring range was 10...100 % relative humidity, with a manufacturer declared accuracy of ± 2 % relative humidity in an ambient temperature of −35...+70 • C as measured by a ventilated PT-100.
Since 5 March 2002, the relative humidity with respect to water is calculated using the 2 m temperature measurement instead of the internal instrument temperature mea-20 surement, as similar measurements in Antarctica indicated radiative effects despite the radiation shield. The measured relative humidity at given internal temperature is used to retrieve the dew point temperature, a parameter that remains unaffected by radiative heating of the shelter. The obtained dew point temperature is combined with the external 2 m surface air temperature measurement to retrieve ambient relative humidity with 25 respect to water. Thus, by applying this method we gain independency of radiative effects on the humidity shelter, and in fact the difference in relative humidity between the 2 sensors has been largely reduced. In any case, we estimate the uncertainty of the humidity measurements to be larger than the accuracy declared by the manufacturer, in the order of 10 % relative humidity for the hair hygrometer and 5 % relative humidity for the humicap sensors, less than 5 % for the retrieval via dew point and external temperature. On 16 March 2002, the humicap sensor of RF1 has been replaced by a new one. On 7 July 2004, also the humicap sensor RF2 has been replaced, and both 5 instruments have been calibrated on site.
In our climatological study we refer only to RF1, as the fraction of missing data for RF2 is too high to produce representative mean values. Figure 2a   On 19 January 1996, the instrument was exchanged with a Paroscientific Digiquartz ® Pressure Sensor 740-16B, providing a barometric range accuracy of ±0.08 hPa declared by the manufacturer. No further changes have been applied to the station pressure measurement until the relocation of the meteorological tower and the pressure sensor in August 2011.

5
In Fig. 3a, the monthly mean surface air pressure values are shown for the analysed period. Due to the large amount of missing values, the months October 1993, July 1994, January 1996, and July 2005 are excluded from this climatological study. As the observed station-level pressure shows a very large month-to-month variation, the general annual evolution throughout the year is not necessarily apparent and therefor 10 explicitly indicated. In fact, average monthly mean values exhibit only small differences between summer and winter, as given in Table 1. More relevant may be the mean day-to-day absolute change in surface pressure shown in Fig. 3b, which indicates the dynamic of passing cyclonic systems. Pressure change from day-to-day is smaller during the summer months when pressure systems tend to be more stable. In spring and autumn, daily pressure changes are increased, whereas the maximum of day-today pressure variability occurs in winter. Regarding the monthly mean, this is either explained by a more frequent passage of frontal systems or by a steeper pressure gradient of the passing cyclonic systems. The enhanced cyclonic activity in winter is also reflected in the surface air temperature variability shown in Fig. 1.   20

Wind speed and direction (10 m)
For the detection of wind speed and direction, a Thies Clima combined wind sensor classic has been installed in 10 m altitude above ground. The sensor is measuring wind speed and direction of the horizontal air flow with an extremely low threshold velocity. Its measuring range is 0. are not considered in this climatological study as too many missing data affect the representativeness for these months, as also for September 1993.

5
As wind is a directional quantity, any averaging is little meaningful as opposing vectors will be cancelled out. Also, the large dynamic range of wind on a short time scale implies limitations to its representation in a climatological analysis. Turbulence and gusts are leveled out with averaging calculations, and their strength and frequency are not represented in the monthly or daily mean. We therefore consider hourly mean 10 values, retrieved by vectorial averaging of the 5-min [1-min] time resolved raw data, assuming that synoptic scale air flow does not significantly change below these time scales. Figure 4 gives an impression of the general wind situation in Ny-Ålesund, showing the two-dimensional frequency distribution of wind speed and wind direction in percent for each month.

15
The most prominent feature is the main wind direction from east/south-east (around 110 • ) which is present throughout the whole year, but with highest wind velocity in the cold period between autumn and spring. The high winter wind speed supports the assumption of enhanced cyclonic storm activity during this period of the year. Generally, strong wind with speed higher than 10 m s −1 almost exclusively arrives from 20 east/south-east. The orientation of the main wind direction is related to the local orography influencing the synoptic atmospheric flow, with the fjord oriented in west-east direction surrounded by mountains channeling the wind. In addition, katabatic flow from the Kongsvegen glacier about 10 km east of Ny-Ålesund affects the predominant flow at the site (Beine et al., 2001). 25 A second frequently observed wind direction is from south-west (around 225 • ), related to katabatic outflow from the Brøggerbreen glaciers channeled along the slopes of the Zeppelin mountain range south of Ny-Ålesund. This southwesterly flow has commonly a slower wind speed < 5 m s −1 , and together with strong temperature inversion is a main precondition for the formation of gravity waves during polar night conditions (Jocher et al., 2012). While all other wind directions occur likewise all year around, wind from the North (around 0 • /360 • ) is almost exclusively observed during the summer period, with velocities less than 5 m s −1 . This wind from the fjord in the North towards the mountain in 5 the South is likely to be a sea breeze formed by increasing temperature differences between the land and water, especially as it occurs only during the snow-free period of polar day. An identical instrument is measuring wind in 2 m altitude. The data are provided in the PANGAEA dataset, but not explicitly presented here due to their general similarity 10 to the 10-m-wind data.

Time series
The available 18 yr of homogeneous observational data reflect ongoing changes in the Arctic climate system. Indeed, during the analysed time period August 1993 to July 2011, surface air temperature has increased. Figure 5 shows the annual mean 15 temperature for the calendar years 1994 to 2010, as retrieved from the monthly mean temperatures presented above. Linear regression of this 17 complete years time series indicates a temperature increase of 1.35 ± 0.92 K per decade in Ny-Ålesund during recent years. Though covering a shorter time period, this increase is in accordance with the trend obtained using the synoptical observations done in Ny-Ålesund by the Nor-20 wegian Meteorological Institute reaching back to 1935 (Førland et al., 2011). With a systematic quality and homogeneity check of the continuing high resolution temperature data collection, the Ny-Ålesund time series will gain an additional source for the retrieval of Arctic temperature trends in the future, allowing also the analysis of short scale dynamics. 25 The observed increase in temperature is accompanied by an increase in absolute humidity. To avoid an effect of missing monthly mean pressure values, the annual mean H 2 O mixing ratio has been calculated using the average annual mean pressure value of 1008.15 hPa over the entire period. Figure 6 shows the annual mean H 2 O mixing ratio for 1994 to 2010, with the linear regression revealing an increase of 0.22 ± 0.14 g kg −1 per decade. These observed changes in sensible and latent heat are part of the complex changes in the Arctic climate system, related both to large scale advection by 5 atmospheric circulation systems and to local effects as enhanced evaporation above open water. The observed slight increase in relative humidity is not significant due to the large scatter of the annual mean values and the according high standard deviation. Stationlevel pressure and 10 m wind give no indication of change during the last two decades.

Summary
A homogeneous data set of high resolution meteorological parameter from Ny-Alesund, Svalbard, for the period August 1993 to July 2011 has been assembled for the use as climatological data. The quality-checked data of surface air temperature, humidity, wind speed and direction (all in 2 and/or 10 m altitude, respectively) as well as 15 surface atmospheric pressure are available online at http://doi.pangaea.de/10.1594/ PANGAEA.793046. The data provide an insight to the climatology of an Arctic landbased station at about 79 • N. With their high temporal resolution, they are suited as data base for atmospheric process studies on small time scales. Furthermore, the meteorological data contribute to the study of changes in the Arctic with focus on alpine/marine 20 landscape, and the continuation of their acquisition will enable to achieve trend analysis for this delicate region in the climate system. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Zhang, X., Sorteberg, A., Zhang, J., Gerdes, R., and Comiso, J. C.: Recent radical shifts of atmospheric circulations and rapid changes in       with the linear regression ± 1 standard deviation (black and grey lines, respectively) giving a 7 humidity increase of +0.22 g/kg per decade. 8 Fig. 6. Annual mean of the H 2 O mixing ratio in Ny-Ålesund from 1994 to 2010 (red dots), with the linear regression ±1 standard deviation (black and grey lines, respectively) giving a humidity increase of +0.22 g kg −1 per decade.