The Cryosphere, 13, 3077–3091, 2019 https://doi.org/10.5194/tc-13-3077-2019 © Author(s) 2019. This work is distributed under the Creative Commons Attribution 4.0 License. Simulated single-layer forest canopies delay Northern Hemisphere snowmelt Markus Todt1,a, Nick Rutter1, Christopher G. Fletcher2, and Leanne M. Wake1 1Department of Geography, Northumbria University, Newcastle upon Tyne, UK 2Department of Geography and Environmental Management, University of Waterloo, Waterloo, Ontario, Canada anow at: National Centre for Atmospheric Science, Department of Meteorology, University of Reading, Reading, UK Correspondence: Markus Todt ([email protected]) Received: 7 December 2018 – Discussion started: 3 January 2019 Revised: 30 August 2019 – Accepted: 30 September 2019 – Published: 25 November 2019 Abstract. Single-layer vegetation schemes in modern land (Webster et al., 2016). In contrast, net longwave radiation surface models have been found to overestimate diurnal cy- fluxes are typically negative for snow under clear-sky condi- cles in longwave radiation beneath forest canopies. This tions in unforested areas, as has been observed for evergreen study introduces an empirical correction, based on forest- Canadian boreal forests (Harding and Pomeroy, 1996; Ellis stand-scale simulations, which reduces diurnal cycles of sub- et al., 2010). Moreover, forest cover has been reported to en- canopy longwave radiation. The correction is subsequently hance snowmelt for subarctic open woodland during overcast implemented in land-only simulations of the Community days and early in the snowmelt season (Woo and Giesbrecht, Land Model version 4.5 (CLM4.5) in order to assess the 2000). However, the impact of forest coverage on snowmelt impact on snow cover. Nighttime underestimations of sub- varies regionally as a function of forest density and meteo- canopy longwave radiation outweigh daytime overestima- rological conditions, with the importance of shortwave and tions, which leads to underestimated averages over the snow longwave radiation changing throughout the snowmelt sea- cover season. As a result, snow temperatures are underes- son (Sicart et al., 2004; Lundquist et al., 2013). timated and snowmelt is delayed in CLM4.5 across ever- Meteorological conditions control longwave enhance- green boreal forests. Comparison with global observations ment, as clear skies increase insolation and thereby vegeta- confirms this delay and its reduction by correction of sub- tion temperature while radiative temperature of the sky is re- canopy longwave radiation. Increasing insolation and day duced (Sicart et al., 2004; Lundquist et al., 2013; Todt et al., length change the impact of overestimated diurnal cycles on 2018). Therefore, values for longwave enhancement, i.e. the daily average sub-canopy longwave radiation throughout the ratio of below-canopy to above-canopy longwave radiation, snowmelt season. Consequently, delay of snowmelt in land- are higher under clear skies but close to 1 or even smaller for only simulations is more substantial where snowmelt occurs overcast conditions due to similar vegetation temperature and early. radiative temperature of the sky. Vegetation density impacts longwave enhancement by scaling the respective contribu- tions of vegetation and atmosphere to sub-canopy longwave radiation as well as by governing the impact of meteorolog- 1 Introduction ical forcing on vegetation temperatures (Todt et al., 2018). While observations have shown trunks heating up due to in- Forest canopy cover modulates longwave radiation received solation and emission of longwave radiation consequently in- by the ground, which consequently differs from atmospheric creasing (Rowlands et al., 2002; Pomeroy et al., 2009), di- longwave radiation. This process is called longwave en- urnal variations in tree temperatures depend on exposure to hancement and has been shown to result in substantial posi- insolation and thus vegetation density (Webster et al., 2016). tive net longwave radiation of the surface when snow cover is prevalent, especially under clear skies and during snowmelt Published by Copernicus Publications on behalf of the European Geosciences Union. 3078 M. Todt et al.: Simulated single-layer forest canopies delay Northern Hemisphere snowmelt About a fifth of seasonally snow-covered land over the iii. to quantify the impact of corrected sub-canopy long- Northern Hemisphere is covered by boreal forests (Rutter wave radiation on snow cover and snowmelt across the et al., 2009), indicating that the process of longwave en- Northern Hemisphere. hancement affects a substantial fraction of global snow cover. Considerable challenges persist in the representation of snow Section2 presents methodological details about the treat- cover and snowmelt in the current generation of climate mod- ment of sub-canopy longwave radiation in CLM4.5, the els, as historical simulations from Climate Model Intercom- physical basis for the empirical correction, configuration of parison Project Phase 5 (CMIP5) underestimate observed global land-only simulations, and calculation of the snow- trends and interannual variability of the spring snow cover off date from SWE observations. Calculation of correction extent (SCE; Derksen and Brown, 2012; Brutel-Vuilmet factors is detailed in Sect.3, and their impacts on the sim- et al., 2013; Rupp et al., 2013; Mudryk et al., 2014; Thack- ulated energy balance and seasonal cycle of snow cover are eray et al., 2016). Snow Model Intercomparison Project’s presented in Sect.4. We conclude with a brief discussion in second phase (SnowMIP2) identified less skill in modelling Sect.5. snow for forested than for open sites, which was attributed to complex processes between atmosphere, snow, and vege- 2 Methods tation (Essery et al., 2009; Rutter et al., 2009). Among models displaying deficiencies in simulating snow 2.1 Sub-canopy longwave radiation in CLM4.5 cover evolution across boreal forests is the Community Land Model (CLM) version 4 and its parent, the Community Cli- Vegetation in CLM4.5 is parameterized as a single layer us- mate System Model version 4 (Thackeray et al., 2014, 2015). ing a “big-leaf” approach (Oleson et al., 2013). Sub-canopy CLM uses a one-layer vegetation scheme, and CLM version longwave radiation is calculated as the sum of atmospheric 4.5 (CLM4.5) has been found to show deficiencies in sim- longwave radiation LWatm and longwave radiation emitted ulation of sub-canopy longwave radiation and longwave en- by vegetation LWveg, weighted by vegetation emissivity "v: hancement with overestimated diurnal cycles under clear-sky D − C 4 conditions (Todt et al., 2018). Similar issues have been mit- LWsub .1 "v/LWatm "vσ Tv ; (1) igated in CLM4.5 by subdividing the roughness layer (Bo- using the Stefan–Boltzmann law with the Stefan–Boltzmann nan et al., 2018) and in SNOWPACK, a one-dimensional constant σ D 5:67 × 10−8 W m−2 K−4 and vegetation tem- snow cover model, by partitioning the canopy into two lay- perature T . Vegetation temperature is calculated based on ers with separate energy balances and consequently separate v an energy balance, net radiation minus turbulent heat fluxes. vegetation temperatures (Gouttevin et al., 2015), which re- Radiative transfer of direct and diffuse shortwave radiation sults in different longwave radiation fluxes emitted upward is calculated via a two-stream approximation (Sellers, 1985) and downward from the vegetation. considering one reflection from ground to canopy. Net long- In order to avoid implementing multiple canopy layers in wave radiation is calculated from atmospheric longwave ra- a global land model and associated computational costs, this diation, vegetation temperature, and (ground) surface tem- study presents an alternative guided by the effect of sepa- perature and determined by vegetation emissivity and emis- rate vegetation layers on sub-canopy longwave radiation. A sivity of the ground. Calculation of turbulent heat fluxes in correction to sub-canopy longwave radiation is implemented CLM4.5 is based on Monin–Obukhov similarity theory and in CLM4.5 to reduce overestimated diurnal cycles, damping described by Oleson et al.(2013). Vegetation emissivity de- variations in longwave radiation emitted downward and, con- pends on the leaf area index (LAI) and stem area index (SAI) sequently, increasing variations in longwave radiation emit- and is calculated as ted upward. While simulation of sub-canopy longwave ra- −.LAICSAI/ diation and longwave enhancement by land surface mod- "v D 1 − e : (2) els has so far been assessed using forest-stand-scale forc- ing and evaluation data, this study uses land-only simulations This parameter is a combination of emissivity in the physical of CLM4.5 and snow-off dates derived from global observa- sense and a weighing parameter based on vegetation density; tions of snow water equivalent (SWE) to assess the impact of however, we will stick to this denomination here for consis- overestimated diurnal cycles in sub-canopy longwave radia- tency with the nomenclature of the technical description of tion on simulated global snow cover and snowmelt. There- CLM4.5 (Oleson et al., 2013). fore, this study has three objectives: CLM4.5 subdivides grid cells based on land units, e.g. vegetated, glacier, or urban, and vegetated land units based i. to develop a correction of sub-canopy longwave radia- on plant functional types (PFTs), with up to 16 possible PFTs tion simulated by single-layer vegetation in CLM4.5, as well as bare soil. Sub-canopy longwave