A Transformer model for predicting near-surface temperature and relative humidity

<p>It is a well-known fact that numerical weather prediction (NWP) models exhibit systematic errors, especially for near-surface variables. Reasons for this are, among other, the inability of these models to successfully handle sub-grid phenomena and shortcomings in the physical formulation of the model dynamics. Even though high-resolution regional NWP models usually have a spatial resolution of a few kilometers (or even finer) they generally exhibit local biases due to unresolved topography and obstacles. In order to obtain more local and site-specific forecasts post-processing methods can be used. Here, we have implemented a Transformer Neural Network model for post-processing 48-hour forecasts of 2 m temperature and relative humidity. The observational data used in this study consist of observations of 2 m air temperature and relative humidity from a network of private weather stations (PWSs). All in all, data from more than 1,000 locations are used. Forecast data from the Global Forecast System (GFS) model – such as temperature, relative humidity, wind speed and direction, radiation fluxes and upper level model fields – are also used as input to the model. The model is trained on 1.5 years of observational and forecast data and the performance is evaluated using an independent validation dataset of PWSs. We find that the Transformer post-processing model reduces the bias and standard deviation compared to the raw NWP forecast for a majority of stations. Furthermore, the model is validated on completely independent data from the Danish Meteorological Institute’s (DMI’s) observational network, where good results were obtained. Overall, the Transformer model produces forecasts that better match the locally observed weather.</p>

Prototype products for climate change adaption in German coastal regions

<p>The “German Strategy for Adaption to Climate Change” (DAS) is the political framework to climate change adaption in Germany. The newly established DAS basic service “Climate and Water” will provide monitoring and projection data to evaluate requirements for climate change adaption. Various products covering the German water bodies (coastal and inland) and its response to climate change will be generated and frequently updated by a cooperation of four German federal agencies. The products will be tailored to a variety of stakeholders needs.</p><p>Within this framework, the Federal Maritime and Hydrographic Agency (BSH) will provide products based on an ensemble of climate projections for the German coast which will be created in cooperation with different research institutes and authorities, e.g. the Danish Meteorological Institute (DMI).</p><p>Data from a DMI climate projection run based on the HIROMB-BOOS model (HBM) with a meteorological forcing from DMI-HIRHAM5 (one of the RCMs in EURO-CORDEX ensemble) and for the RCP 8.5 scenarios has been analysed in view of different oceanographic parameters such as sea level, sea surface temperature, salinity, currents and ice. This data set includes the historical periods 1981-2010 and the RCP 8.5 periods 2041-2070 and 2071-2100. Therefore, it provides an expedient basis to develop prototype products regarding climate change adaption at the German coasts for customers of the DAS basic service “Climate and Water”. The initial prototype products are presented and discussed in regards to the sufficiency to evaluate requirements for climate change adaption.</p>

Inter-comparison of snow depth over Arctic sea ice from reanalysis reconstructions and satellite retrieval

In this study, we compare eight recently developed snow depth products over Arctic sea ice, which use satellite observations, modeling, or a combination of satellite and modeling approaches. These products are further compared against various ground-truth observations, including those from ice mass balance observations and airborne measurements. Large mean snow depth discrepancies are observed over the Atlantic and Canadian Arctic sectors. The differences between climatology and the snow products early in winter could be in part a result of the delaying in Arctic ice formation that reduces early snow accumulation, leading to shallower snowpacks at the start of the freeze-up season. These differences persist through spring despite overall more winter snow accumulation in the reanalysis-based products than in the climatologies. Among the products evaluated, the University of Washington (UW) snow depth product produces the deepest spring (March–April) snowpacks, while the snow product from the Danish Meteorological Institute (DMI) provides the shallowest spring snow depths. Most snow products show significant correlation with snow depths retrieved from Operational IceBridge (OIB) while correlations are quite low against buoy measurements, with no correlation and very low variability from University of Bremen and DMI products. Inconsistencies in reconstructed snow depth among the products, as well as differences between these products and in situ and airborne observations, can be partially attributed to differences in effective footprint and spatial–temporal coverage, as well as insufficient observations for validation/bias adjustments. Our results highlight the need for more targeted Arctic surveys over different spatial and temporal scales to allow for a more systematic comparison and fusion of airborne, in situ and remote sensing observations.

Variations of surface waters characteristics in the Eastern Arctic Ocean in 2010-2019 at different scales: in situ and satellite study

<p>The last 10 years of Arctic Ocean observations showed the dramatic changes and new records of the sea ice minimum. The largest variations were observed in the Eastern Arctic: the Kara, the Laptev, the East-Siberian seas. This region is a key area of the important freshwater input from the great Siberian rivers (Ob’, Yenisei, Lena). This remote area remains one of the less studied in the Arctic Ocean, although several regular expeditions (such as NABOS or Transdrift) together with special expeditions following the Northern Route, such as Tara-2013 expedition, or recent Transarktika-2019 expedition help to monitor the changes of surface waters in recent years.</p><p>The use of new satellite-derived datasets, (e.g., SST blended product from Danish Meteorological Institute or REMSS, SSS SMOS from LOCEAN University of Sorbonne) fill the gaps and help to better understand the complex dynamics of surface waters in the Eastern Arctic ocean.</p><p>In this work, we discuss the surface waters variations using in situ and satellite data at different scales.  Synoptic scales are studied with continuous and point in situ measurements (thermosalinographs and CTD data). The recent scientific results of Transarktika-2019 expedition are presented. In the summer season of 2019 (July-October) Transarktika expedition did oceanographic measurements following the Northern Route twice, from Vladivistok to Murmansk and back to Vladivostok. The seasonal variations are analyzed over the period of 10 years, comparing with climatological data. The difference between the climatological values of SST or SSS can reach 5 or more units in some areas of the Eastern Arctic. The results of interannual variations analysis using satellite data, suggest the salinification (“Atlantification”) of the southern areas and freshening of the northern parts of the Eastern Arctic.</p><p>The development of SSS SMOS Arctic product was supported by the French CNES-TOSCA SMOS-OCEAN project. Anastasiia Tarasenko, Nikita Kusse-Tiuz, Mikhail Makhotin and Vladimir Ivanov acknowledge financial support from the Ministry of Science and Higher Education of the Russian Federation, project RFMEFI61619X0108</p>

Inter-comparison of snow depth over sea ice from multiple methods

<p>In this study, we compare eight recently developed snow depth products that use satellite observations, modeling or a combination of satellite and modeling approaches. These products are further compared against various ground-truth observations, including those from ice mass balance buoys (IMBs), snow buoys, snow depth derived from NASA's Operation IceBridge (OIB) flights, as well as snow depth climatology from historical observations.</p><p>Large snow depth differences between data sets are observed over the Atlantic and Canadian Arctic sectors. Among the products evaluated, the University of Washington snow depth product (UW) produces the overall deepest Spring snow packs, while the snow product from the Danish Meteorological Institute (DMI) provide the shallowest Spring snow depths. There is no significant trend for mean snow depth among all snow products since the 2000s, however, those in regional varies larhely. Two products, SnowModel-LG and the NASA Eulerian Snow on Sea Ice Model: NESOSIM, also provide estimates of snow density. Arctic-wide, these density products show the expected seasonal evolution with varying inter-annual variability, and no significant trend since the 2000s. Compared to climatology, snow density from SnowModel-LG is generally denser, whereas that from NESOSIM is less. Both SnowModel-LG and NESOSIM densities have a larger seasonal change than climatology.</p><p>Inconsistencies in the reconstructed snow parameters among the products, as well as differences and with in-situ and airborne observations can in part be attributed to differences in effective footprint and spatial/temporal coverage, as well as insufficient observations for validation/bias adjustments. Our results highlight the need for more targeted Arctic surveys over different spatial and temporal scales to allow for a more systematic comparison and fusion of airborne, in-situ and remote sensing observations.</p>

Inter-comparison of snow depth over sea ice from multiple methods

In this study, we compare eight recently developed snow depth products that use satellite observations, modeling or a combination of satellite and modeling approaches. These products are further compared against various ground-truth observations, including those from ice mass balance buoys (IMBs), snow buoys, snow depth derived from NASA's Operation IceBridge (OIB) flights, as well as snow depth climatology from historical observations. Large snow depth discrepancies between the different snow depth data sets are observed over the Atlantic and Canadian Arctic sectors. Among the products evaluated, the University of Washington snow depth product (UW) produces the overall deepest spring (March-April) snow packs, while the snow product from the Danish Meteorological Institute (DMI) provide the shallowest spring snow depths. There is no significant trend in the mean snow depth among all snow products since the 2000s, despite the great differences in regional snow depth. Two products, SnowModel-LG and the NASA Eulerian Snow on Sea Ice Model (NESOSIM), also provide estimates of snow density. Arctic-wide, these density products show the expected seasonal evolution with varying inter-annual variability, and no significant trend since the 2000s. The snow density in SnowModel-LG is generally higher than climatology, whereas NESOSIM density is generally lower. Both SnowModel-LG and NESOSIM densities have a larger seasonal change than climatology. Inconsistencies in the reconstructed snow parameters among the products, as well as differences between in-situ and airborne observations can in part be attributed to differences in effective footprint and spatial/temporal coverage, as well as insufficient observations for validation/bias adjustments. Our results highlight the need for more targeted Arctic surveys over different spatial and temporal scales to allow for a more systematic comparison and fusion of airborne, in-situ and remote sensing observations.

Observationally constrained reconstruction of 19th to mid-20th century sea-ice extent off eastern Greenland

Arctic sea ice has a significant impact on the global radiation budget, oceanic and atmospheric circulation and the stability of the Greenland ice sheet (Vaughan et al. 2013). Prior to the era of aircraft and satellite, information on sea-ice extent relied on observations from ships and people living at the coast. This information is a valuable contribution to better understand the history of sea ice. However, the information exists in a range of formats, e.g., sea-ice extent before the late 1800s is typically reported in the literature as an annual index from a single geographical point or as hand-drawn maps. This makes it difficult to assess and compare data across time and space. The combination of digitised historical maps and single-point data makes the information more accessible and provides a record that can help understand the dynamics and processes of the climate and its interactions with the cryosphere (Chapman & Walsh 1993). In this study, maps of sea-ice extent by Koch (1945) were digitised. We use these maps in combination with sea-ice charts from the Danish Meteorological Institute (DMI) and Koch’s sea-ice index from 1820 to 1939, to map estimated sea-ice extent between Iceland and Greenland going back to 1821. This information has not been included in even the most recent databases of Arctic sea ice (Walsh et al. 2015, 2017). Furthermore, we extract time series of sea-ice extent at a number of locations and investigate the relationship between them. Our observation area is along eastern Greenland, between the southern tip of Greenland at 59°46´N northwards to 77°21´N.

The impact of climate on risk of mania

IntroductionBipolar disorder varies with season: admissions for depression peak in winter and mania peak in summer. Sunlight presumably increases the risk of mania through suppression of melatonin. If so, we expect admissions for mania to vary in accordance with climate variations.ObjectivesTo investigate how climate and climate changes affects admissions for mania.AimsTo identify which climate variables – sunshine, ultraviolet radiation, rain and snow cover – affect admissions for mania.To examine whether year-to-year weather variation as well as long-term climate changes reflects the variation in number of admissions for mania.MethodsThis register-based nationwide cohort study covers all patients admitted for mania (ICD-10 code F31 or F30.0–F30.2) between 1995 and 2012 in Denmark. Climate data, obtained from the Danish Meteorological Institute, were merged with admission data and correlated using an Unobserved Component Model regression model.Preliminary resultsIn total, 8893 patients were admitted 24,313 times between 1995 and 2012: 6573 first-admissions and 17,740 readmissions. Linear regression shows significant association between admissions per day and hours of sunshine (P < 0.01) and ultraviolet radiation (UV) dose (P < 0.01). Average days with snow cover and rain were not significantly correlated with admissions. Analyses on year-to-year variation and long-term change are not yet available.Preliminary conclusionsAdmissions for mania are correlated with sunshine and UV, but not rain and snow cover. If more patients are admitted during very sunny summers compared with less sunny summers this implies a relation with light itself and not just season.Disclosure of interestThe authors have not supplied their declaration of competing interest.

Exposure to sunshine early in life prevented development of type 1 diabetes in Danish boys

We aimed to assess the association between exposure to sunshine during gestation and the risk of type 1 diabetes (T1D) in Danish children.The study population included 331,623 individuals born in Denmark from 1983 to 1988; 886 (0.26%) developed T1D by the age of 15 years. The values of sunshine hours were obtained from the Danish Meteorological Institute. Gestational exposure to sunshine was calculated by summing recorded monthly sunshine hours during the full 9 months prior to the month of birth. The linear variable then was split into two categories separated by the median value.Cox regression models showed that more sunshine during the third gestational trimester was associated with lower hazards (HR) of T1D at age 5–9 years in males: HR (95% CI): 0.60 (0.43–0.84), p=0.003. Our results should be considered in the context of evidence-based recommendations to the public about skin protection from the sun.