Search

Permafrost in Mongolia is located in the transition zone between high-latitude and high-altitude permafrost regions of the Northern Hemisphere, with large temperature differences and complex subsurface characteristics. In this study, the reliability of the skin temperature data from the ERA5-Land product covering Mongolia is assessed via site observations. The ERA5-Land skin temperature dataset shows a cold bias, which is more pronounced in the cold season. Following calibration based on elevation differences, significant improvements are observed at both the annual scale (92 % improvement in RMSE (root mean square error) and 98 % improvement in MBE (mean bias error)) and the seasonal scale (78 % improvement in RMSE and 82 % improvement in MBE). Additionally, the spatial variations in the surface freezing index (SFI) and surface thawing index (STI) are most pronounced in the central and northeastern Mongolia. The SFI exhibits a significant decreasing trend of 7.16 °C·d/year, while the STI shows a significant increasing trend of 4.49 °C·d/year. Furthermore, the permafrost extent in Mongolia is simulated from 1950 to 2022 using the frost number (Fn) model and the temperature on top of permafrost (TTOP) model. The validated results indicate that the accuracy of the Fn model is relatively high, with an overall accuracy of 0.9 and a Kappa coefficient of 0.47. The permafrost extent in Mongolia has declined from 734.7 × 103 km2 in the 1950 s to 480.1 × 103 km2 in the 2010 s, with a prominent decrease of 3.2 × 103 km2/decade after 1994. According to the variations in permafrost extent during past 72 years, the Hovsgol and Khentii Mountain ranges have experienced significant permafrost degradation.

Permafrost soils are a critical component of the global carbon cycle and are locally important because they regulate the hydrologic flux from uplands to rivers. Furthermore, degradation of permafrost soils causes land surface subsidence, damaging infrastructure that is crucial for local communities. Regional and hemispherical maps of permafrost are too coarse to resolve distributions at a scale relevant to assessments of infrastructure stability or to illuminate geomorphic impacts of permafrost thaw. Here we train machine learning models to generate meter‐scale maps of near‐surface permafrost for three watersheds in the discontinuous permafrost region. The models were trained using ground truth determinations of near‐surface permafrost presence from measurements of soil temperature and electrical resistivity. We trained three classifiers: extremely randomized trees (ERTr), support vector machines (SVM), and an artificial neural network (ANN). Model uncertainty was determined using k‐fold cross validation, and the modeled extents of near‐surface permafrost were compared to the observed extents at each site. At‐a‐site near‐surface permafrost distributions predicted by the ERTr produced the highest accuracy (70%–90%). However, the transferability of the ERTr to the sites outside of the training data set was poor, with accuracies ranging from 50% to 77%. The SVM and ANN models had lower accuracies for at‐a‐site prediction (70%–83%), yet they had greater accuracy when transferred to the non‐training site (62%–78%). These models demonstrate the potential for integrating high‐resolution spatial data and machine learning models to develop maps of near‐surface permafrost extent at resolutions fine enough to assess infrastructure vulnerability and landscape morphology influenced by permafrost thaw. Plain Language Summary: Accurate spatial assessments of the extent of near‐surface permafrost in regions of discontinuous permafrost are essential for evaluating the potential impacts of permafrost thaw. Most current estimates of permafrost extent are provided at the scale of 10s of meters to kilometers, which is often too coarse to understand the impacts thaw might have on infrastructure or geomorphic processes. We trained machine learning models to generate maps of near‐surface permafrost at 3 m spatial resolution in three sites in western Alaska. These maps provide an advancement in evaluating the extent of permafrost in the region as well as highlighting the landscape and ecological factors most important in driving predictions. Key Points: High‐resolution maps of near‐surface permafrost on hillslopes were generated from machine learning modelsThe normalized difference vegetation index was the most important feature in prediction of near‐surface permafrostThe support vector machines generally generated the highest balanced accuracy for predictions at sites where the model was not trained [ABSTRACT FROM AUTHOR]

Abstract Permafrost soils are a critical component of the global carbon cycle and are locally important because they regulate the hydrologic flux from uplands to rivers. Furthermore, degradation of permafrost soils causes land surface subsidence, damaging infrastructure that is crucial for local communities. Regional and hemispherical maps of permafrost are too coarse to resolve distributions at a scale relevant to assessments of infrastructure stability or to illuminate geomorphic impacts of permafrost thaw. Here we train machine learning models to generate meter‐scale maps of near‐surface permafrost for three watersheds in the discontinuous permafrost region. The models were trained using ground truth determinations of near‐surface permafrost presence from measurements of soil temperature and electrical resistivity. We trained three classifiers: extremely randomized trees (ERTr), support vector machines (SVM), and an artificial neural network (ANN). Model uncertainty was determined using k‐fold cross validation, and the modeled extents of near‐surface permafrost were compared to the observed extents at each site. At‐a‐site near‐surface permafrost distributions predicted by the ERTr produced the highest accuracy (70%–90%). However, the transferability of the ERTr to the sites outside of the training data set was poor, with accuracies ranging from 50% to 77%. The SVM and ANN models had lower accuracies for at‐a‐site prediction (70%–83%), yet they had greater accuracy when transferred to the non‐training site (62%–78%). These models demonstrate the potential for integrating high‐resolution spatial data and machine learning models to develop maps of near‐surface permafrost extent at resolutions fine enough to assess infrastructure vulnerability and landscape morphology influenced by permafrost thaw.

[Display omitted] • The calibrated reanalysis data with elevation differences shows significant improvement. • The surface freezing and thawing index changes with −7.16 °C·d/year and 4.49 °C·d/year, respectively. • The current Mongolian permafrost extent is 503.9 × 103 km2 and declined by 34.65 % from 1950 to 2022. Permafrost in Mongolia is located in the transition zone between high-latitude and high-altitude permafrost regions of the Northern Hemisphere, with large temperature differences and complex subsurface characteristics. In this study, the reliability of the skin temperature data from the ERA5-Land product covering Mongolia is assessed via site observations. The ERA5-Land skin temperature dataset shows a cold bias, which is more pronounced in the cold season. Following calibration based on elevation differences, significant improvements are observed at both the annual scale (92 % improvement in RMSE (root mean square error) and 98 % improvement in MBE (mean bias error)) and the seasonal scale (78 % improvement in RMSE and 82 % improvement in MBE). Additionally, the spatial variations in the surface freezing index (SFI) and surface thawing index (STI) are most pronounced in the central and northeastern Mongolia. The SFI exhibits a significant decreasing trend of 7.16 °C·d/year, while the STI shows a significant increasing trend of 4.49 °C·d/year. Furthermore, the permafrost extent in Mongolia is simulated from 1950 to 2022 using the frost number (Fn) model and the temperature on top of permafrost (TTOP) model. The validated results indicate that the accuracy of the Fn model is relatively high, with an overall accuracy of 0.9 and a Kappa coefficient of 0.47. The permafrost extent in Mongolia has declined from 734.7 × 103 km2 in the 1950 s to 480.1 × 103 km2 in the 2010 s, with a prominent decrease of 3.2 × 103 km2/decade after 1994. According to the variations in permafrost extent during past 72 years, the Hovsgol and Khentii Mountain ranges have experienced significant permafrost degradation. [ABSTRACT FROM AUTHOR]

This paper presents new field observations that assist in further defining the extent of Last Permafrost Maximum (LPM) permafrost and frozen ground in north China. In that respect, this region is specific by its arid environment and continental position. Bedrock exposures and sand and gravel pits were examined along a south-north transect from southern Ordos as far as the vicinity of Dongsheng District, Ordos City on the Ordos Plateau (North China). Attention focussed upon the recognition of sand-wedge remains and possible cryoturbations. Two types of wedge structures were recognized: type A wedges extend to depths in excess of 3.0 m and are thought to have formed under conditions of continuous permafrost. Type B wedges are restricted to the upper 1.5 m and are thought to reflect discontinuous permafrost and/or deep seasonal frost. OSL dating of wedge infills indicates that type A wedges formed between 23–19 ka and younger type B wedges formed until the early Holocene. It is now possible to delineate a more exact position of the boundary between continuous permafrost and discontinuous permafrost and/or deep seasonal frost on the Ordos Plateau during the LPM. This boundaries occur between Dongsheng and Jinbian at 37°-39°N. • Precise location of southern limit of continuous permafrost during the Last Permafrost Maximum (c.19-23 ka) in north China. • Narrow zone of discontinuous/sporadic permafrost during the Last Permafrost Maximum (c.19-23 ka) in north China. • Permafrost extent in N China at LPM was specific by its aridity, continentality and proximity to the Siberian High Pressure. [ABSTRACT FROM AUTHOR]

Abstract Due to sparse data and discontinuous time observations in the circum-Arctic region, freezing index and thawing index, as useful indicators, are widely used in permafrost distribution, climate changes and cold-region engineering analysis. However, previous researches on freezing/thawing index over this region were estimated based on mean monthly air temperature. In this paper, we analyzed the spatial and temporal variations of the freezing/thawing index over the circum-Arctic from 1901 to 2015 based on the daily datasets, besides monthly datasets. The results showed that freezing index had a downward changing trend and thawing index had an upward trend during 1901–2015. More important, the change trend in freezing/thawing index after 1988 was more significant than before. Furthermore, different freezing/thawing index based on daily datasets and the monthly datasets were assessed and compared according to daily data from 17 meteorological stations, comprehensive relative errors evaluation implied that freezing/thawing based on daily datasets was more accurate generally, although both of other datasets were available in calculating the freezing/thawing index. As the daily datasets are better in calculating annual freezing/thawing index, therefore, the permafrost extent was estimated by a climate-based predictive model combined with snow depth data from Canadian Meteorological Centre (CMC). Finally, considering that the published permafrost map of the circum-Arctic only shows the past permafrost distribution, but it cannot reflect the permafrost distribution after 2000 under the climate warming. Hence, we simulated the current (mean from 2000 to 2015) permafrost area which is 19.96 × 106 km2, and the results showed some discrepancies between published and simulated permafrost extent mainly located in isolated permafrost regions. Graphical abstract Unlabelled Image Highlights • Freezing index has a downward changing trend during 1901–2015. • Thawing index has an upward changing trend during 1901–2015. • The simulated current (mean from 2000 to 2015) permafrost extent area is 19.96 × 106 km2. [ABSTRACT FROM AUTHOR]

The framework of Quaternary permafrost in China was reconstructed for the first time on the basis of available periglacial, glacial, and other proxies. During the Early Pleistocene (2.68–0.80 Ma BP), permafrost advanced southwards to 47–50° N in northern China and possibly occurred in alpine regions in western China. During the Middle Pleistocene (800–130 ka BP), permafrost occurred extensively on the Qinghai-Tibet Plateau (QTP) and in alpine or mountainous regions of northern, western, central, and northeastern China. The Great Interglacial occurred afterward and before the Last Glaciation, but the evidence of permafrost for this period has been seldom found. Permafrost evolution of the Last Glaciation (72–19 ka BP) in China is divided into: Expansion (72~50 ka BP), degradation (50–26 ka BP), and intensive expansion during the Last Permafrost Maximum (LPMax, 26–19 ka BP) with a permafrost extent of 5.3 × 106~5.4 × 106 km2, and when major features of present permafrost took shape. Permafrost fluctuated during the Younger Dryas (12.9–11.7 ka BP). Since the Holocene, permafrost in China expanded and retreated to lesser extents, forming the current permafrost environment. The Holocene evolution of permafrost was divided into: Unstable climate but stable permafrost during the early Holocene (11.7~8.5–7.0 ka BP); permafrost degradation during the Last Permafrost Minimum (LPMin, or the Holocene Megathermal; 8.5–7.0~4.0–3.0 ka BP) and the Medieval Warm Period (MWP; 1.0~0.5 ka BP); permafrost expansion during the Neoglaciation (4.0–3.0~1.0 ka BP) and the Little Ice Age (LIA; 0.5~0.1 ka BP); and recent permafrost degradation (20th century to the present). However, this review paper only provides the framework of Quaternary permafrost in China and some preliminary discussions. Many key questions await further investigations.

This study quantitatively evaluated how insulation by snow depth (SND) affected the soil thermal regime and permafrost degradation in the pan-Arctic area, and more generally defined the characteristics of soil temperature (T) and SND from 1901 to 2009. This was achieved through experiments performed with the land surface model CHANGE to assess sensitivity to winter precipitation as well as air temperature. Simulated T, active layer thickness (ALT), SND, and snow density were generally comparable with in situ or satellite observations at large scales and over long periods. Northernmost regions had snow that remained relatively stable and in a thicker state during the past four decades, generating greater increases in T. Changes in snow cover have led to changes in the thermal state of the underlying soil, which is strongly dependent on both the magnitude and the timing of changes in snowfall. Simulations of the period 2001-2009 revealed significant differences in the extent of near-surface permafrost, reflecting differences in the model's treatment of meteorology and the soil bottom boundary. Permafrost loss was greater when SND increased in autumn rather than in winter, due to insulation of the soil resulting from early cooling. Simulations revealed that T tended to increase over most of the pan-Arctic from 1901 to 2009, and that this increase was significant in northern regions, especially in northeastern Siberia where SND is responsible for 50 % or more of the changes in T at a depth of 3.6 m. In the same region, ALT also increased at a rate of approximately 2.3 cm per decade. The most sensitive response of ALT to changes in SND appeared in the southern boundary regions of permafrost, in contrast to permafrost temperatures within the 60°N-80°N region, which were more sensitive to changes in snow cover. Finally, our model suggests that snow cover contributes to the warming of permafrost in northern regions and could play a more important role under conditions of future Arctic warming. [ABSTRACT FROM AUTHOR]

There is mounting evidence that permafrost degradation has occurred over the past century. However, the amount of permafrost lost is uncertain because permafrost is not readily observable over long time periods and large scales. This paper uses JULES, the land surface component of the Hadley Centre global climate model, driven by different realisations of twentieth century meteorology to estimate the pan-arctic changes in near-surface permafrost. Model simulations of permafrost are strongly dependent on the amount of snow both in the driving meteorology and the way it is treated once it reaches the ground. The multi-layer snow scheme recently adopted by JULES significantly improves its estimates of soil temperatures and permafrost extent. Therefore JULES, despite still having a small cold bias in soil temperatures, can now simulate a near-surface permafrost extent which is comparable to that observed. Changes in snow cover have been shown to contribute to changes in permafrost and JULES simulates a significant decrease in late twentieth century pan-Arctic spring snow cover extent. In addition, large-scale modelled changes in the active layer are comparable with those observed over northern Russia. Simulations over the period 1967-2000 show a significant loss of near-surface permafrost-between 0.55 and 0.81 million km per decade with this spread caused by differences in the driving meteorology. These runs also show that, for the grid cells where the active layer has increased significantly, the mean increase is ~10 cm per decade. The permafrost degradation discussed here is mainly caused by an increase in the active layer thickness driven by changes in the large scale atmospheric forcing. However, other processes such as thermokarst development and river and coastal erosion may also occur enhancing permafrost loss. [ABSTRACT FROM AUTHOR]