📄 earth_science

Reconstruction of missing data and analysis of air temperature variability at the Arctowski Station (Antarctic Peninsula region), 1977–2025

This study reconstructs missing daily mean air temperature data for the Arctowski Station (1977–2025) using ERA5 reanalysis and regression models, revealing significant local variability and alternating warming/cooling periods without a statistically significant long-term trend, while noting increased climate oceanity.

Original authors: Marek Kejna, Joanna Plenzler, Sławomir Sulik

Published 2026-07-10

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Marek Kejna, Joanna Plenzler, Sławomir Sulik

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine Antarctica's Antarctic Peninsula as a giant, icy playground where the weather is the most unpredictable game of all. For nearly 50 years, scientists have been keeping score at a specific spot called the Henryk Arctowski Polish Antarctic Station (HA), located on King George Island. But, like a diary with torn-out pages, their weather log had huge gaps. Some years were missing entirely; other months had holes in them.

This paper is the story of how researchers Marek Kejna and his team acted like detective detectives to fill in those missing pages, not by guessing, but by using a super-accurate digital map of the sky called ERA5.

The Great Data Heist: Filling in the Blanks
The team couldn't just use data from nearby stations because the local mountains and wind patterns make King George Island a bit of a weather island of its own. Instead, they grabbed hourly data from the ERA5 reanalysis grid—a massive, global computer model that tracks the atmosphere.

They found that while the computer model was great, it wasn't perfect. It was like a friend who tells you the temperature is "cold," but misses that it's actually "freezing" or "chilly." Specifically, the model was often a bit too cold compared to the real station, especially in the summer. The difference was about 0.6°C on average, but in February, the real station was a cozy 1.1°C warmer than the model predicted.

To fix this, the team built a set of custom "translation rules" (regression equations) for every single month. Think of it like a secret decoder ring: if the computer said it was -5°C in July, the rule told them the real station was actually -5.01°C. If the computer said it was 1°C in February, the rule adjusted it to a toasty 2.1°C. Once they applied these rules to the missing days, they had a complete, unbroken record from 1977 to 2025.

The Weather Report: What Did They Find?
With the full diary restored, the team took a deep look at the climate of the last 48 years. Here is what the numbers say:

The Average Chill: The station is the warmest spot in the entire region, with an average yearly temperature of -1.6°C. That might sound freezing, but in Antarctica, it's practically a heatwave!
The Seasons: The coldest month is July, averaging -5.8°C. The warmest is February, hitting an average of 2.3°C.
The Extremes: The temperature swings wildly. The coldest it ever got was -32.3°C (on July 24, 1986), and the hottest was a balmy 16.7°C (on January 20, 1979).
The "Ocean" Effect: The weather here is heavily influenced by the ocean. The temperature doesn't swing as wildly as it does on a frozen continent; the difference between the hottest and coldest months is only 7.4°C. This is like living in a house with excellent insulation—the ocean keeps the temperature from going crazy.

The Mystery of the Missing Trend
Here is the big twist: If you look at the entire 48-year period from 1977 to 2025, there is no statistically significant trend showing that the station is getting warmer or colder overall. The paper explicitly rules out a simple, steady warming line for this specific location over this specific time.

Instead, the temperature acts like a rollercoaster with no clear direction, just big loops:

It warmed up until 1984.
Then it cooled down until 1994.
It warmed up again until 1999.
It cooled down until 2015.
And finally, it started rising again after 2015.

The authors suggest these ups and downs are caused by big atmospheric patterns, like the Southern Annular Mode and "atmospheric rivers" (which are like giant, invisible rivers of warm air flowing from the tropics to the poles).

What About the Ice?
The team also looked at "Positive Degree Days" (days warm enough to melt ice) and "Frost Degree Days" (days cold enough to freeze things). While the amount of melting or freezing changes a lot from year to year, the paper notes that no statistically significant trend was observed for either Positive Degree Days or Frost Degree Days during the analyzed period. Instead of a steady shift toward more melting or more freezing, the data shows high variability and fluctuations over several years, with no clear long-term direction in the sums.

The Bottom Line
This paper doesn't claim to have solved the mystery of global warming in Antarctica. Instead, it proves that even in a place known for rapid change, the weather can be a chaotic mix of warming and cooling periods that cancel each other out over decades. The data is now complete and reliable, showing that while the Antarctic Peninsula is a "hotspot" for change, the Henryk Arctowski station has been a place of wild swings rather than a straight line up or down. The future? The paper suggests that if we keep looking, we might see more warming in the coming decades, but for the years 1977 to 2025, the story is one of variability, not a single direction.

Technical Summary: Reconstruction of Missing Data and Analysis of Air Temperature Variability at Arctowski Station (1977–2025)

Problem Statement
The Antarctic Peninsula is recognized as a climate "hotspot" undergoing rapid environmental changes, including significant deglaciation and temperature shifts. However, long-term meteorological records from stations in this region often contain data gaps due to operational interruptions, equipment failures, or logistical challenges. The Henryk Arctowski Polish Antarctic Station (HA), located on King George Island and operational since 1977, possesses a nearly five-decade history marked by intermittent observations. These gaps hinder the analysis of long-term climate variability and trends. Furthermore, while reanalysis datasets like ERA5 provide global coverage, they often exhibit systematic biases when compared to local station measurements in polar regions due to differences in orography, surface radiation balance, and local circulation. The primary challenge addressed by this study is the reconstruction of missing daily mean air temperature data at the HA station to create a homogeneous, continuous time series for the period 1977–2025, while accounting for the discrepancies between local station data and ERA5 reanalysis values.

Methodology
The study employed a multi-stage approach to reconstruct missing data and analyze temperature variability:

Data Identification and Source Selection: The authors identified specific periods of missing daily data within the HA station's 1977–2025 record, covering approximately 29.8% of the total days. Missing values were supplemented using hourly data from the ERA5 reanalysis (Copernicus Climate Change Service) corresponding to the grid cell encompassing the HA station (62.00°–62.25° S, 58.25°–58.50° W).
Regression Modeling: To address systematic differences between the HA station and ERA5 values—particularly the finding that HA is generally warmer than the ERA5 grid, especially in summer—the authors developed monthly linear regression equations. These models were calibrated using a complete, homogeneous dataset from 2013 to 2025, where both station and ERA5 data were available. The Pearson correlation coefficients ( $R$ ) for these monthly equations ranged from 0.92 to 0.95.
Validation and Homogeneity Testing: The reconstructed data were validated against the measured data from the 1978–1990 period, yielding a correlation coefficient of $R = 0.949$ . The homogeneity of the reconstructed 1977–2025 series was tested using the Standard Normal Homogeneity Test (SNHT) by Alexandersson (1986).
Climate Analysis: The study calculated various climatic parameters, including:
- Mean, maximum, and minimum temperatures.
- Continentality (Gorczyński index, $K_G$ ) and Oceanity (Marsz index, $O_M$ ) indices.
- Positive Degree Days (PDD) and Frost Degree Days (FDD).
- Trend analysis using Student's t-test for annual and monthly means.
- Comparative analysis with neighboring stations: Bellingshausen and Faraday/Vernadsky.

Key Results

Data Reconstruction: The reconstructed dataset for 1977–2025 was determined to be homogeneous, with the SNHT test statistic ( $T(k) = 7.86$ ) falling below the critical value (8.20) at the 0.05 significance level.
Temperature Characteristics: The HA station exhibits the highest annual mean air temperature in the region at $-1.6^\circ\text{C}$ . The annual cycle shows mild winters (July mean: $-5.8^\circ\text{C}$ ) and cool summers (February mean: $2.3^\circ\text{C}$ ). Extreme values range from $-32.3^\circ\text{C}$ to $16.7^\circ\text{C}$ .
Seasonal Discrepancies with ERA5: Significant differences exist between HA and ERA5. While winter differences are minimal (HA sometimes cooler), summer differences are substantial, with HA being up to $1.1^\circ\text{C}$ warmer than ERA5 in February. This suggests local factors, such as foehn effects and solar radiation on ice-free terrain, significantly influence the station's microclimate compared to the broader ERA5 grid.
Trends and Variability: No statistically significant long-term warming or cooling trend was observed for the entire 1977–2025 period ( $p = 0.74$ ). Instead, the data reveals distinct multi-decadal fluctuation periods: warming until 1984, cooling until 1994, warming until 1999, cooling until 2015, and a subsequent rise.
Oceanic Influence: The region is characterized by extreme oceanic climate features, indicated by a low annual temperature amplitude ( $7.4^\circ\text{C}$ ), a low Gorczyński continentality index ( $12.0\%$ ), and a high Marsz oceanity index ($11.2$).
Degree Days: The mean annual sum of Positive Degree Days (PDD) was $329.6^\circ\text{C}$ , and Frost Degree Days (FDD) was $-898.4^\circ\text{C}$ . No statistically significant trends were found for PDD or FDD over the study period, though high interannual variability was noted.
Regional Consistency: The temperature variability at HA correlates strongly with the Bellingshausen station ( $R=0.95$ ) and moderately with Faraday/Vernadsky ( $R=0.65$ ), reflecting shared regional atmospheric drivers.

Significance and Claims
The paper positions the reconstructed 1977–2025 dataset as a critical resource for understanding climate dynamics in the Antarctic Peninsula. The authors claim that the study successfully addresses the issue of data gaps in long-term polar records, providing a homogeneous series that allows for robust analysis of variability.

Key claims regarding the study's significance include:

Methodological Contribution: The study demonstrates that while ERA5 is a valuable tool for filling gaps, direct substitution introduces bias. The development of monthly regression equations based on local conditions is a necessary step to accurately reconstruct polar station data.
Climate Context: The results corroborate the complex nature of Antarctic Peninsula climate, characterized by high interannual variability and distinct warming/cooling phases rather than a linear trend. This supports the concept of an "Antarctic Hiatus" (a period of stalled warming) between the late 1990s and 2015, followed by a renewed warming phase.
Local vs. Global Drivers: The findings highlight that local factors (orography, sea ice extent, foehn winds) play a decisive role in temperature records, often overriding broader reanalysis trends. The station's specific thermal advantage in summer is attributed to these local mechanisms.
Environmental Baseline: The reconstructed data serves as a baseline for assessing the impact of climate change on the cryosphere (glaciers and permafrost) and the broader ecosystem in the Admiralty Bay region, a site of intensive multi-disciplinary research.

The authors conclude that the observed variability aligns with changes in atmospheric circulation (Southern Annular Mode, atmospheric rivers) and sea ice extent, emphasizing the need for continued monitoring to understand future climate regimes in this sensitive region.

More like this