ARTICLE DOI: 10.1038/s41467-017-00669-3 OPEN Statistical modelling predicts almost complete loss of major periglacial processes in Northern Europe by 2100 Juha Aalto1,2, Stephan Harrison3 & Miska Luoto1 The periglacial realm is a major part of the cryosphere, covering a quarter of Earth’s land surface. Cryogenic land surface processes (LSPs) control landscape development, ecosystem functioning and climate through biogeochemical feedbacks, but their response to con- temporary climate change is unclear. Here, by statistically modelling the current and future distributions of four major LSPs unique to periglacial regions at fine scale, we show funda- mental changes in the periglacial climate realm are inevitable with future climate change. Even with the most optimistic CO2 emissions scenario (Representative Concentration Pathway (RCP) 2.6) we predict a 72% reduction in the current periglacial climate realm by 2050 in our climatically sensitive northern Europe study area. These impacts are projected to be especially severe in high-latitude continental interiors. We further predict that by the end of the twenty-first century active periglacial LSPs will exist only at high elevations. These results forecast a future tipping point in the operation of cold-region LSP, and predict fun- damental landscape-level modifications in ground conditions and related atmospheric feedbacks. 1 Department of Geosciences and Geography, University of Helsinki, P.O. Box 64, Gustaf Hällströmin katu 2a, 00014 Helsinki, Finland. 2 Finnish Meteorological Institute, P.O. Box 503, FI-00101 Helsinki, Finland. 3 College of Life and Environmental Sciences, University of Exeter, Penryn TR10 9EZ, UK. Correspondence and requests for materials should be addressed to J.A. (email: juha.aalto@helsinki.fi) NATURE COMMUNICATIONS | 8: 515 | DOI: 10.1038/s41467-017-00669-3 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00669-3 eriglacial environments with frost-induced and permafrost- vegetation community structure and productivity12, 13, hydrol- controlled land surfaces processes (LSPs) are vital compo- ogy14 and biogeochemical cycles7, 15, 16. Currently, the response P 1, 2 fi nents of the cryosphere . With current Arctic ampli ca- of periglacial LSPs to climate warming is highly uncertain. tion of climate warming3, substantial alterations in this sensitive Although a rapid response is expected7, because the broad-scale and important area of the Earth’s system are already observed2, distribution of cryogenic ground processes is coupled with cli- including glacier recession4, shrub expansion to alpine tundra5 matic gradients17, 18 and often, but not necessarily, with the and changes in permafrost thermal–hydrological regimes6. presence of permafrost1, 19, the details of this and timing are Importantly, these changes in ground conditions modify, among lacking. This strong LSP–climate response17, 20 is locally modified others, biogeochemical cycles (e.g., terrestrial CO and CH ) and by lithology and edaphic (reflecting, e.g., glaciation heritage)1, 11 2 – 4 reflectance (i.e., albedo) triggering climate feedbacks7 9. Thus, and topographical characteristics17, 21, 22. For example, gelifluc- better understanding of the response of the periglacial climate tion operates on inclined surfaces with frost-susceptible fine- realm to climate change is critical for assessing climate change grained soils, while cryoturbation features are common in flat mitigation, and extensive modelling studies at various geo- valley bottoms and mountain tops1, 17. The development of palsa graphical scales are urgently required7. mires through permafrost mounding is expected to occur on open The combined spatial extent of active cryogenic LSPs con- low-elevation peat lands where strong winds redistribute snow stitutes the periglacial climate realm1. This prevails across high allowing for a deep frost penetration10, 18. latitudes and elevations, at present covering ca. 25% of the Earth’s Here, we use remotely sensed and field-quantified data of LSPs terrestrial areas. Here, LSPs create surface geomorphological at an unprecedented scale focusing on active surface features features which are unique to periglacial regions including pat- related to cryoturbation, gelifluction, nivation and permafrost terned ground and hummocky terrain associated with cryo- peat mounding to investigate the current and future extent of the turbation, gelifluction terraces and lobes, nivation features periglacial climate realm across a high-latitude Fennoscandia associated with erosion by snow patches and palsa mires which region of ca. 78,000 km2 (Fig. 1). We argue that the absence of develop through permafrost mounding10 (Fig. 1). Periglacial LSPs deep permafrost (compared with, e.g., High Arctic Canada and play a crucial role by controlling denudation processes11, Siberia)23, 24 means that in general the thermal response of LSPs a, b Permafrost extent a b Continuous 70 °N Discontinuous Sporadic 50 °N Isolated Glacier Arctic ocean c Elevation (m a.s.l.) >1,200 800–1,200 68 °N 400–800 200–400 70 °N 50 °N 0–200 Sea level Atlantic ocean Observation 40 °W 20 °W 0 ° 20 °E 40 °E 20 °E 25 °E de1 m c 70 °N 1 m f g 68 °N 1 m 1 m 20 °E 25 °E Fig. 1 The location of the study domain and LSP observation sites in northernmost Europe. a, b The study area in relation to the circum-Arctic extent of permafrost23, 24 indicated as: continuous=90–100% of the area covered by permafrost, discontinuous=50–90%, sporadic=10–50% and isolated=0–10%, respectively. c The observation locations (n = 2,917) and the relief of the study area. Black rectangles in b, c depict the model prediction domain. Photos show examples of typical surface features of cryogenic land surface processes (scales are only directive): cryoturbation (d; small-scaled polygonal patterned soil) gelifluction (e; gelifluction lobes), nivation (f, snow accumulation sites) and permafrost mounding (g; palsas), respectively. Photos by J.A. and M.L. 2 NATURE COMMUNICATIONS | 8: 515 | DOI: 10.1038/s41467-017-00669-3 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00669-3 ARTICLE abAUC = 0.80 ± 0.04; TSS = 0.48 ± 0.06 AUC = 0.92 ± 0.02; TSS = 0.71 ± 0.04 n = 873 n = 686 71 °N 70 °N 69 °N 21 °E 24 °E km 010050 cdAUC = 0.93 ± 0.01; TSS = 0.72 ± 0.03 AUC = 0.97 ± 0.01; TSS = 0.84 ± 0.03 n = 294 n = 420 Observed presence Modelled presence Modelled absence Fig. 2 The modelled baseline occurrences of the four LSPs based on majority vote ensemble. The n in the title denotes the number of observed presences (black dots), while all the observation sites (n = 2,917) are presented in Fig. 1c. The modelling performance is measured as the area under the curve of a receiver operating characteristic plot (AUC) and the true skill statistics (TSS). The evaluation statistics show the mean (±s.d.) over four modelling techniques and 100 cross-validation runs conducted for each LSP (a cryoturbation, b gelifluction, c nivation and d permafrost mounding) using a random sampling procedure to future climate change is likely to be rapid (perennial ground ice century, respectively) and two time periods (2040–2069 and enhances soil-ambient air decoupling25), and thus the region is 2070–2099), averaged over a large group (n = 23) of CMIP5 cli- representative of environments that are especially sensitive to mate simulations29. We show potential for a notable reduction in climate change26. Similar sensitive landscapes are expected to the current periglacial climate realm in our study area, and pre- prevail across broad high-latitude areas of discontinuous and dict that by the end of the twenty-first century active periglacial isolated permafrost, including large parts of Canada and Russia LSPs will exist only at high elevations. between 55 and 70° N latitudes. We relate the current occurrence of LSPs with climatic variables of freezing and thawing degree days (FDD and TDD, respectively), water and snow precipitation, Results local topography (potential radiation, slope angle and topo- Present distributions of cryogenic LSPs. Our forecasts of the graphic wetness) and soil characteristics (peat and rock cover)17. current LSPs show high agreement with the observations thus We use an ensemble modelling approach, where methodology- suggesting robust model transferability to similar environments related uncertainty can be controlled by merging predictions from (Fig. 2). The analysis of the current periglacial realm closely multiple statistical algorithms (regression and machine learning) corresponds to earlier definitions1, 20, 30, marking mean annual to a single agreement map27 (spatial resolution 50 m × 50 m; refer air temperature (MAAT) of +2 °C as a rough upper limit for to Methods for a description of data compilation and statistical cryogenic ground processes (Fig. 3). At present, cryoturbation, analysis). After investigating the baseline distributions (i.e., the gelifluction and nivation are active across a broad range of climate current climate of 1981–2010) of the LSPs, we develop climate conditions, while permafrost mounding is most concentrated projections forced by three Representative Concentration Path- with MAAT of ~ −2 °C and low to moderate annual precipitation 28 – way (RCP) scenarios (2.6, 4.5 and 8.5, roughly equal to CO2 sum (400 600 mm) (Fig. 3). The probability of active LSPs concentrations of 490, 650, 1,370 p.p.m. by the end of this increases towards cold air temperatures (Fig. 4). Permafrost NATURE COMMUNICATIONS | 8: 515 | DOI: 10.1038/s41467-017-00669-3 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00669-3 Density 10 0.25 10 10 0.25 a b c 0.20 5 0.20 5 0.15 0.15 0.20 0 0 0.10 0.10 5 −5 0.05 −5 0.05 0.15 −10 0.00 −10 0.00 400 800 1,200 400 800 1,200 0 0.10 10 10 0.35 d 0.20 e 0.30 5 5 0.25 −5 0.15 0.20 Mean annual air temperature (°C) air temperature Mean annual 0.05 0 0 Mean annual air temperature (°C) air temperature Mean annual 0.10 0.15 0.10 −5 0.05 −5 0.05 −10 0.00 −10 0.00 −10 0.00 400 600 800 1,000 1,200 400 800 1,200 400 800 1,200 Mean annual precipitation (mm) Mean annual precipitation (mm) Baseline 2070−2099 RCP 2.6 2070−2099 RCP 4.5 2070−2099 RCP 8.5 Fig.