Reconstructed Complete Data Series of Air Temperature, Snow Depth, and Ground Temperature Unveil Opportunities for Understanding Evolving Arctic Climate Dynamics

This blog describes the ideas and work behind the publication: Daily station-level records of air temperature, snow depth, and ground temperature in the Northern Hemisphere.
Published in Earth & Environment
Reconstructed Complete Data Series of Air Temperature, Snow Depth, and Ground Temperature Unveil Opportunities for Understanding Evolving Arctic Climate Dynamics

Share this post

Choose a social network to share with, or copy the shortened URL to share elsewhere

This is a representation of how your post may appear on social media. The actual post will vary between social networks

Interrelationships between atmospheric, terrestrial, and cryospheric temperature regimes in the Arctic climate system

The Arctic region is experiencing rapid climate warming at rates double the global average (Rantanen et al., 2022). This warming is intricately linked to changes in key climate variables such as air temperature (Ta), ground temperature (Tg), and snow depth (Sd). These variables are interconnected through complex feedback mechanisms and energy exchanges within the Arctic system.

Ta is perhaps the most direct indicator of warming trends. As atmospheric greenhouse gas concentrations rise, more heat is trapped, leading to increasing air temperatures. However, the impact of this warming on the Arctic extends beyond just the air. Tg, which reflects the thermal state of the soil and permafrost, is also being affected. Warmer air temperatures lead to heat transfer into the ground, causing permafrost thaw and potential landscape changes. Sd plays a crucial role in this process as well. Snow acts as an insulator, regulating the exchange of heat between the atmosphere and the ground. Decreasing snow cover, a consequence of warmer temperatures, can lead to more heat being absorbed by the ground, further exacerbating permafrost thaw. Conversely, deeper snow can insulate the ground and slow the rate of permafrost degradation. This feedback loop between air temperature, ground temperature, and snow depth is a key driver of the Arctic warming and its associated impacts.

The changing dynamics of these variables have far-reaching consequences for the Arctic ecosystem, including the potential release of greenhouse gases from thawing permafrost, shifts in vegetation spatial distributions, and impacts on wildlife habitats. By unraveling the intricate connections between atmospheric, terrestrial, and cryospheric temperature regimes, scientists can develop more accurate climate models, inform adaptation and mitigation strategies, and contribute to the broader understanding of the Arctic's role in the global climate system.

The Paucity of Complete In-Situ Observational Datasets in the Arctic Region

The urgent need to investigate the cumulative and compound effects of multiple drivers of change in the Arctic system has been recently highlighted, necessitating a convergence science approach to understanding the impacts (Ivanov et al., 2024). However, a major challenge in the Arctic research stems from the discontinuous and incomplete nature of observational data records (Fig. 1). Across the vast and remote Arctic region, meteorological and environmental monitoring stations are sparsely distributed and difficult to maintain, leading to substantial spatial and temporal gaps in data coverage. While there is an abundance of datasets providing data on Ta, Sd, and/or Tg, their temporal characteristics can be highly inconsistent (e.g., a station can have time-varying temporal resolutions) and contain numerous missing data points. Such datasets also mix multiple resolutions, such as sub-daily, daily, weekly, monthly, and yearly.

Moreover, although some reanalysis data products offer contiguous series of climate variables at a high temporal resolution, the spatial resolution provided by such models can be coarse and unable to capture local-scale variations that can be represented by in-situ observations. Remote sensing-based data products on air temperature and snow depth have been increasingly used in research over the past decade. However, they are often fraught with substantial uncertainties that require complex post-processing and validation against ground-based data to be used with confidence. Additionally, some datasets were created by integrating various data sources (e.g., satellite images, reanalysis data, and ground-based data), but only for specific areas or countries, particularly for air temperature and snow depth. These discontinuities and missing values, and significant uncertainties limit our ability to accurately characterize climate patterns in the Arctic change, trends, and processes. These factors also underscore the need for a uniform, standardized daily dataset of these variables on a large scale.

Figure 1. Locations and regional contributions of in situ observational records used in this study.

Leveraging AI/ML as a Time machine: Reconstructing Historical-Missing Data

The field of artificial intelligence and machine learning (AI/ML) has revolutionized various domains, and its impact on data reconstruction is nothing short of remarkable. These advanced techniques hold the potential to serve as a "time machine", enabling researchers and historians to unlock the secrets of the past by reconstructing historical data that may have been lost, incomplete, or inaccessible through traditional means.

In this study, we harness the power of long short-term memory (LSTM) networks combined with reanalysis data (ERA5-Land) to reconstruct daily data for Ta, Sd, and Tg. Specifically, the geographic scope encompasses the broader Northern Hemisphere north of 30°N latitude (below which there are no stations that have measured ground temperature in the Northern Hemisphere), including the data-sparse Arctic region, with the goal of enabling diverse applications of the resultant dataset. Specifically, we reconstruct data for 27,768 Ta, 32,417 Sd, and 659 Tg stations. The total reconstructed data amounts to 54.5%, 59.3%, and 74.3% of the daily time series for Ta, Sd, and Tg, respectively, from 1960 to 2021. Our validation results demonstrate the strong capability of the LSTM model in reconstructing the data. The LSTM accurately reconstructed extreme values as well. Furthermore, we assess the quality of the reconstructed data through trend analysis, which reveals that these data can be reliably used in further in-depth analyses. The cross-validation conducted against stations with over 90% available data shows near-consistent trend evaluation results.

Finally, we are delighted to share our work with the scientific community and domain experts in the high-impact journal, Scientific Data. We sincerely hope that this resource can provide valuable research groundwork and further insights for the community.


Ivanov, V. Y., Ungar, P. S., Ziker, J. P., Abdulmanova, S., Celis, G., Dixon, A., Ehrich, D., Fufachev, I., Gilg, O., Heskel, M., Liu, D., Macias-Fauria, M., Mazepa, V., Mertens, K., Orekhov, P., Peterson, A., Pokrovskaya, O., Sheshukov, A., Sokolov, A., . . . Zhou, W. (2024). A Convergence Science Approach to Understanding the Changing Arctic. Earth's Future, 12(5), e2023EF004157.

Rantanen, M., Karpechko, A. Y., Lipponen, A., Nordling, K., Hyvärinen, O., Ruosteenoja, K., Vihma, T., & Laaksonen, A. (2022). The Arctic has warmed nearly four times faster than the globe since 1979. Communications Earth & Environment, 3(1), 168.


Please sign in or register for FREE

If you are a registered user on Research Communities by Springer Nature, please sign in

Follow the Topic

Climate Change Mitigation
Physical Sciences > Earth and Environmental Sciences > Earth Sciences > Climate Sciences > Climate Change > Climate Change Mitigation

Related Collections

With collections, you can get published faster and increase your visibility.

Ecological data for tracking biological diversity and environmental change

This collection presents data contributions addressing topics in biodiversity and ecology.

Publishing Model: Open Access

Deadline: Jan 31, 2024

Remote sensing data for changes in land use

This Collection comprises a series of articles presenting data on changes to land use in urban areas, farmland, forests, and natural environments, as determined using remote sensing techniques.

Publishing Model: Open Access

Deadline: Jan 31, 2024