Vegetation type is an important predictor of the arctic summer land surface energy budget

The Arctic is warming at a rate faster than the global average, leading to significant changes in various biogeophysical variables. These changes include increased rainfall and permafrost temperatures, permafrost thaw, reduced ice mass and snow cover, and complex alterations in Arctic vegetation composition, structure, and function. These changes impact climate dynamics locally and globally through land-atmosphere feedbacks mediated by the land surface energy budget (SEB). The SEB is defined by its components: R_net = SW_net + LW_net = H + LE + G + M, where R_net is net radiative energy, SW_net and LW_net are net shortwave and longwave irradiances, H is sensible heat flux, LE is latent heat flux, G is ground heat flux, and M is latent heat of fusion. Uncertainties persist in high-latitude climate projections, particularly concerning sensible and latent heat fluxes, which directly feedback to Arctic biophysical variables. The magnitude and seasonality of SEB components depend on a complex interplay of drivers, including vegetation type, snow cover, soil and permafrost characteristics, topography, and meteorological conditions. However, a quantitative understanding of the relative importance of vegetation type compared to other drivers is lacking. Current Earth system models often represent Arctic vegetation using only a few plant functional types (PFTs), despite significant diversity. While previous observational studies demonstrate that Arctic vegetation types influence SEB components, they often lack quantitative assessments or are geographically limited. This study addresses this gap by providing a quantitative, circumpolar assessment of the observed SEB over treeless land >60°N from 1994–2021, comparing the predictive skill of vegetation type with other important SEB drivers.

The existing literature highlights the importance of high-latitude surface energy budgets (SEBs) in land-climate interactions within the rapidly changing Arctic environment. However, a considerable degree of uncertainty remains in their prediction. Previous studies have individually investigated the influence of various factors such as vegetation type, snow cover, soil and permafrost characteristics, topography, and meteorological conditions on SEB components. These studies, however, often focus on qualitative descriptions, cover limited geographical extents, or lack a comprehensive quantitative comparison of the relative importance of different SEB drivers. The current Earth system models frequently simplify the representation of Arctic vegetation, using only a few plant functional types (PFTs), despite the region's notable vegetation diversity. Consequently, this study aims to quantitatively assess the circumpolar SEB, focusing on the predictive power of vegetation type in comparison to other key drivers, with the goal of improving the accuracy of future climate projections.

This study uses harmonized in situ observations from various circumpolar monitoring networks (FLUXNET, AmeriFlux, AON, ICOS, GEM, GC-Net, and PROMICE) covering 64 tundra and glacier sites from 1994 to 2021. Data were aggregated to daily resolution after quality control and outlier filtering, with units and flux direction conventions harmonized. Net radiation (R_net), net shortwave radiation (SW_net), net longwave radiation (LW_net), and albedo were derived where necessary. Normalized fluxes were calculated as a percentage of daily maximum potential incoming shortwave radiation. Vegetation type at each site was classified based on literature descriptions, aligning with the Circumpolar Arctic Vegetation Map (CAVM) categories: barren complexes, graminoid tundra, prostrate dwarf-shrub tundra, erect-shrub tundra, wetland complexes, glacier, and boreal peat bogs. Additional SEB drivers (climate, topography, snow cover, permafrost characteristics, and cloud cover) were extracted from various spatial data products. A variance partitioning analysis was employed on a vegetation subset (excluding glacier sites) to compare the predictive skill of 15 selected SEB drivers for summer (June-August) mean magnitudes of R_net, H, LE, and G. A linear mixed-model analysis was used to estimate mean surface energy fluxes for terrestrial Arctic (including vegetated and glacier sites) in summer and annually. Post-hoc analyses tested pairwise differences among vegetation types. Seasonality was assessed by quantifying the timing of SEB-flux 'summer regimes' (daily mean values exceeding 0 Wm⁻²) relative to snow-free and -onset dates, using a subset of data from 2000-2021. The timing was compared across vegetation types using Welch's t-tests. All data processing and analyses were performed using R software.

The variance partitioning analysis revealed that vegetation type is a significant predictor of summer surface energy flux magnitudes, especially for sensible (H) and latent (LE) heat fluxes. For H and LE, vegetation type explained 56.3% and 71.7% of variance, respectively. For ground heat flux (G), vegetation type ranked among the top three predictors. For net radiation (R_net), latitude and snow cover duration were the most important predictors, but vegetation type showed intermediate importance. However, analysis with normalized fluxes showed vegetation type among the top three predictors for R_net, along with bioclimatic subzone and landscape-scale dominant vegetation. Linear mixed-model analysis indicated significant relationships between vegetation type and summer mean magnitudes of H, LE, SW_net, and LW_net, and a significant relationship with annual R_net. Pairwise comparisons showed strong differences in H, LE, and LW_net among vegetation types, sometimes exceeding the differences between glacier and vegetation types. For example, summer LE was significantly higher in boreal peat bogs compared to barren tundra. The Bowen ratio (H/LE) varied across vegetation types, being >1 for shrub and wetland types and <1 for graminoid, peat bog, and barren tundra. Analysis of normalized fluxes showed vegetation type effects significant for H and LE across all summer months, while other fluxes varied seasonally. Analyses of seasonality demonstrated that the timing of SEB flux summer regimes varied significantly among vegetation types relative to snow-free and snow-onset dates. For R_net, H, and G, transitions into summer regimes occurred significantly earlier than the spring snow-free date, with greater variability at the start than the end of the season, varying among vegetation types. The timing differences imply vegetation's control on snow cover and SEB seasonality.

The strong predictive power of vegetation type for Arctic summer SEB fluxes likely results from a combination of factors. Vegetation types reflect integrated proxies of environmental conditions influencing SEB (temperature, topography, soil moisture, permafrost). Furthermore, CAVM classes differ in SEB-relevant traits (height, productivity, albedo), suggesting direct causal effects. The low predictive power of cloud cover is explained by consistently high fractional cloud cover and low spatial variability in summer. The relatively low predictive power of precipitation suggests energy-limited systems or that precipitation effects are captured by other variables. The significant differences in surface energy flux magnitudes among vegetation types highlight their importance. Differences between barren and shrub-dominated tundra in sensible heat flux are comparable to differences between glacier and other vegetation types. The observed negative sensible heat flux in barren tundra suggests it acts as a heat sink, potentially creating a positive feedback to climate warming upon its reduction due to climate change. High latent heat fluxes in boreal peat bogs suggest a potential shift towards latent heat flux with peatland expansion, affecting water loss, precipitation, and soil water availability. Analyses of seasonality show vegetation type influences the timing of SEB flux summer regimes relative to snowmelt, suggesting effects on snow distribution and snowmelt timing, with implications for the cumulative energy fluxes. However, the limited data on barren tundra warrants further investigation.

This study demonstrates the significant role of vegetation type in predicting Arctic summer surface energy budgets at a circumpolar scale. Differences among vegetation types can be as large as, or even larger than, differences between vegetated and glaciated surfaces. The findings highlight the need for more comprehensive representation of Arctic vegetation types and their functional traits in Earth system models to improve predictions of Arctic surface energy fluxes, particularly latent and sensible heat fluxes. Future research should focus on identifying key vegetation traits related to SEB, investigating vegetation-snow interactions, and addressing data gaps, particularly concerning year-round SEB data and data on barren tundra.

This study acknowledges several limitations. Long-term, year-round SEB data for Arctic vegetation are scarce, with missing data for many sites in autumn and winter. Year-round turbulent flux measurements are especially limited. SEB observations for barren tundra are lacking, despite its significant differences in surface energy fluxes compared to other tundra types. The analysis focused on vegetation-type level seasonality due to the limited availability of data at the study-site level.