The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand

Transcription

1 The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand NIWA Client Report: CHC November 2010 NIWA Project: SAN09501

2

3 The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand Jordy Hendrikx Einar Örn Hreinsson Prepared for The Ski Areas Association of New Zealand NIWA Client Report: CHC November 2010 NIWA Project: SAN09501 National Institute of Water & Atmospheric Research Ltd 10 Kyle Street, Riccarton, Christchurch P O Box 8602, Christchurch, New Zealand Phone , Fax All rights reserved. This publication may not be reproduced or copied in any form without the permission of the client. Such permission is to be given only in accordance with the terms of the client's contract with NIWA. This copyright extends to all forms of copying and any storage of material in any kind of information retrieval system.

4

5 Contents Executive Summary i 1. Introduction 1 2. PART A: Nationwide Assessment Background Study Area 3 3. Climate Data Current climate data Climate change scenarios Overview Emissions scenarios Downscaling Scenario scaling Empirical adjustment of daily rainfall data Application to this study Mean temperature and precipitation changes Adjustment of daily rainfall data Assumptions and limitations Methods Snow model Simulating historical SWE Simulating future SWE scenarios Results Discussion Conclusions Part B: Site specific assessments and snowmaking potential Background Methods Climate data from VCSN Snow model calibration for current climate Climate change Precipitation adjustment for extremes Snow model for future climates 33

6 9.6. Snowmaking Limitations and uncertainties Sites Results Conclusions Future research needs Acknowledgements References 45 Reviewed by: Approved for release by: Ross Woods James Renwick

7 Executive Summary The purpose of this report is to provide an assessment of the potential impact of climate change on seasonal snow in New Zealand at a national and local scale. The report is structured into two main sections. Section A presents the summary results of the New Zealand-wide snow modelling for the 2040s ( ) and 2090s ( ) for a selection of emissions scenarios. Section B presents the results of detailed snow modelling and snow making potential for a number of selected ski area locations. These specific site-scale results are each available only to the corresponding participating organisations. The results presented here are the first quantitative assessments of the potential impact of climate change on seasonal snow in New Zealand. The future data are based on the average of 12 global climate models used for the recent Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report, and are determined for three different emission scenarios. Data for the current climate (the period ) have been extracted from NIWA s interpolated Virtual Climate Station Network (VCSN). The mean rainfall and temperature changes for the periods and from the baseline period , for all of New Zealand, have been produced by statistically downscaling global climate model output under the A1B emission scenario to the VCSN grid (approximately 5km spatial resolution). These changes in precipitation and temperature are scaled, based on projections to 2100 of the global mean temperature under different emission scenarios, to yield additional lower (B1) and higher (A1FI) emissions scenarios. See section A for a more detailed description. Using these data and the snow model developed by Clark et al. (2009), which has been specifically designed and calibrated for New Zealand conditions, we have modelled the potential impact of climate change on seasonal snow. The results of this work are consistent with our understanding of snow processes and with research from other mid-latitude locations. On average at a national scale, at nearly all elevations, the 2040s and 2090s scenarios result in a decrease in snow as described by all of our summary statistics; snow duration, fraction of precipitation that is snow, and mean maximum snow accumulation in each year. This decrease in snow is more marked at elevations below 1000m but is evident at all but the highest elevations. In part B of this report we have used the same snow model (Clark et al., 2009) as used in Section A of this report, but have re-calibrated it for each ski area location. Where available, we have used observations of snow depth to calibrate the snow model output for the current climate. Where reliable observations of snow depth were not available, our default parameter set for the snow model (as used for the nationwide modelling) has been used. This calibration was necessary to ensure that the model could simulate the current snow conditions as accurately as possible at each of the selected sites. Having generated modelled snow for the current climate, we then proceed to model snow for the The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand i

8 and time periods, for the three emissions scenarios (B1, A1B and A1FI) for an upper and lower site on each participating ski area. Depending on the elevation and location of the specific site in question, our analysis shows that under a mid-range scenario (A1B) by 2040 there will be between 93% and 79% of the current maximum snow depth (on 31 August) at the upper elevation sites, and by 2090 this will be on average further reduced to approximately 80% to 54% of the current maximum snow depth. Under a warmer scenario (A1FI) by 2040 there will be on average between 92% and 72% of the current snow depth (on 31 August) at our upper elevation sites and by 2090 this will be on average further reduced to approximately 79% to 35% of the current maximum snow depth. At the lower elevation sites on each ski area, the change is more pronounced. Our analysis shows that under a mid range scenario (A1B) by 2040 there will be on average around 91% to 65% of the current maximum snow depth (on 31 August) at the lower sites and by 2090 this will be on average further reduced to approximately 68% to 20% of the current maximum snow depth. Under a warmer scenario (A1FI) by 2040 there will be on average around 83% to 45% of the current snow depth at the lower sites, and by 2090 this will be on average further reduced to approximately 48% to 9% of the current maximum snow depth. While elevation is clearly an important factor in determining the potential impact of climate change on seasonal snow, other factors such as the local climatology of each site also play an important role. Additionally, we also simulate the future climate for the 2040s and 2090s, for three future scenarios, to enable calculations of the potential available time for snowmaking in these two future time periods. We use the simulated temperatures (and humidity) to estimate the future wet bulb temperatures, and calculate the total potential snowmaking hours in the future climates. For the snowmaking analysis a worst case year, rather than an average year, was used to assess the snowmaking potential. This was to ensure consistency with snowmaking design practices. For all emissions scenarios and both future time periods a reduction in potential snowmaking hours is observed at all sites. Using manufacturers information on snowmaking flow rates and temperatures we calculate the potential volume, and thereby snow depth, that could be made per 10,000m 2 area in these future time periods. Our assessment calculates the minimum number of snowguns required to supplement the natural snow depth in these future time periods to achieve the target depths of 0.3m by 15 June and 0.5m by 15 July. At all the upper sites by 2090 under an A1FI scenario in our worst case year, our analysis shows that a minimum level of snowmaking (i.e. one snowgun covering a 10,000m 2 area) will be needed to achieve the desired target snow depths. At most upper sites, this minimum level would be sufficient to meet the target snow depths when all snowmaking opportunities are utilised (from 1 May onwards) and combined with the available natural snow cover in our worst case year. At all the lower elevation sites by 2090 under an A1FI scenario in our worst case year, our analysis shows that at least a minimum level of snowmaking (i.e. one snowgun covering a 10,000m 2 area) will be needed. At nearly all sites one snowgun alone will not be sufficient to meet target snow depths and in some cases two, or up to five additional snow guns (i.e. six per 10,000m 2 ) would be needed when all snowmaking opportunities are utilised (from 1 May onwards) and combined with the available natural snow cover The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand ii

9 in our worst case year. Due to limitations and uncertainties contained within the snowmaking analysis and target depth achievement, we also present the relative percentage for the amount of snow that can be theoretically made by one snow gun for each two-weekly period at each site under this worst case 2090 A1FI scenario (relative to the 1990s worst year). At the upper sites by the 2090s in this worst case year, the average percentage for snowmaking for the period 1 May to 31 July will range from 59% to 32 % and at the car park sites for this period it will range from 49% to 17% (relative to the 1990s). No assessments were made for optimal piste preparation at any of the sites, or the ideal number of snowguns to achieve snow depth targets within minimum timeframes or windows of opportunity. No assessment of the hydrological or economic limitations on snow making was provided for each site. While we have confidence in our analysis, we still urge caution regarding reliance on the very precise percentage values of snow depth and snowmaking potential change presented, as we have employed a relatively simple delta change methodology for the climate change analysis, and there is some uncertainty in these values. Future work should consider the use of a regional climate model coupled with the snow model, use of additional snow data, and improving the understanding and representation of alpine meteorology. The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand iii

10

11 1. Introduction The purpose of this research is twofold; first we provide a nationwide assessment of the potential impact of climate change on seasonal snow in New Zealand; second we assess this impact of this on selected individual ski area locations and determine their ability to make snow using snowmaking in these future climates. This report has been split into two sections; Part A: The nationwide assessment of the potential impacts of climate change on seasonal snow, and Part B: The ski area-specific results and snowmaking potentials. These site-specific results are each available only to the corresponding participating organisations. The NIWA Snowmodel (Clark et al., 2009) is used to simulate the current snow conditions in New Zealand under the current climate, for a period from 1980 to 1999 (midpoint reference year 1990), and up to three projected future climates for the periods 2030 to 2049 (midpoint reference year 2040, the) and 2080 to 2099 (midpoint reference year 2090). The snow model is not altered in any way for the future climates and only the input data is changed. The three future climates are based on different greenhouse gas emissions scenarios; a middle-of-the-road (A1B), an upper emissions scenario (A1FI) and a lower emissions scenario (B1). We have mainly used the A1B emissions scenario for our nationwide snow modelling (Part A), and use all three scenarios for most of our site specific modelling (Part B). These emissions scenarios are described in full in the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report (IPCC, 2007) and in the IPCC Special Report on Emissions Scenarios (Nakicenovic et al. 2000). For each of the selected ski area locations, where available, observations of snow depth have been used to calibrate the snow model output for the current climate. This calibration essentially accounts for the processes that the model can not capture at this fine scale (e.g. local scale re-distribution of snow). The calibration was necessary to ensure that the model could simulate the current snow conditions as accurately as possible at each of these selected sites. This is a critical step before we attempted to simulate a future climate. Having generated modelled snow for the current climate, we generated modelled snow for the and time periods for the three scenarios described above. We then used the estimated climate data for these three future scenarios to calculate the available time for snowmaking in these two time periods. This permitted an assessment to be made about the ability of each specific location to adapt to the potential impacts of climate change on seasonal snow.

12 PART A Nationwide Assessment The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 1

13 2. PART A: Nationwide Assessment 2.1. Background The snowfields and glaciers of the Southern Alps of New Zealand are the most significant in the Southern Hemisphere outside of Antarctica and South America (Fitzharris et al., 1999). While there is an extensive body of literature regarding the glaciers of the Southern Alps (e.g. Chinn, 1991; Chinn, 1996; Fitzharris et al., 1999), their inter-annual variability (e.g. Chinn and Whitehouse, 1980; Fitzharris et al., 1997; Chinn, 1999), and their response to climate variability (e.g. Chinn, 1996; Hooker and Fitzharris, 1999; Lamont et al., 1999; Clare et al., 2002; Chinn et al., 2005), there is comparatively very limited research undertaken on seasonal snow (Fitzharris et al., 1999). This is despite the extensive seasonal snow cover in the Southern Alps and its direct effect on New Zealand's economy. Snow contributes up to 24% of the inflows in the major hydro electricity lakes (McKerchar et al., 1998), and as much as 70% in Spring and early Summer. Many agricultural applications rely on the snow melt component of streamflow for their irrigation needs. Snow also provides numerous opportunities for winter recreation: the Ski Areas Association of New Zealand (SAANZ) estimate that about 300,000 New Zealanders ski or snow board regularly and in 2009 the industry announced a record 1.5 million skier days in New Zealand. However, snow is also a hazard in some areas, causing winter time closures on the Milford Road (Hendrikx et al., 2005) and crop and infrastructure damage from low elevation winter snow storms (e.g. Hendrikx, 2007). Depending on the pace of global climate change over the coming decades, the extent, depth, duration and quality of snow in New Zealand is likely to be subject to significant change. Regional climate models currently suggest that the maximum daily temperatures in the Southern Alps could rise by more than 4 C by the end of this century (Dean et al., 2006). It has also been suggested that there will be increased extreme precipitation events and more precipitation in the west (Ministry for the Environment, 2008). The impact of this warming and change in precipitation on seasonal snow in New Zealand and more specifically individual ski areas was, until now, unknown. The potential impacts of these changes in seasonal snow are both direct, through changes in snow cover to a specific operation, and also indirect as impacts on; for example, melt water impacts on hydro-lake intake or alpine biodiversity. The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 2

14 In New Zealand, there has been no published work on the potential impacts on seasonal snow with climate change. The lack of systematic snow observations in New Zealand (Fitzharris et al., 1999) means that information on interannual variability, trends and projections of seasonal snow must be generated using simulation models. Clark et al., (2009) present results of their temperature index snow model for the South Island of New Zealand. We have used this model to assess the potential impacts of climate change in this report. Furthermore, in 2003 the Australian Ski Areas Association in collaboration with several Australian government agencies commissioned CSIRO to undertake a study to assess the impact of climate change on seasonal snow in mainland Australia (Hennessy et al., 2003). That study has proved very useful for the Australian ski, hydro power, tourism and land management industries, and this has reinforced the need to make a similar assessment for New Zealand. Therefore, the aim of this first part of this research project (Part A) is to provide the first quantitative estimates of the potential impacts of climate change on seasonal snow in New Zealand Study Area The area of interest for this study is the whole of New Zealand. The South Island spans latitudes 40 S to 47 S while the North Island spans latitudes 34 S to 41 S However the most northerly area of substantial snow is the Central North Island at a latitude of 39 S. The location of New Zealand, in the Pacific Ocean and away from continental influences and constant exposure to the westerly winds, gives the climate a distinct maritime character. The South Island is dominated by the Southern Alps, a mountain range rising to over 3000m which extends along the entire length of the western portion of the South Island. Precipitation gradients across the Southern Alps are extreme, ranging from over 12,000 mm/year on the west coast to less than 500 mm/year in the rain shadow areas east of the divide. This orographic precipitation pattern occurs because the Southern Alps are aligned almost perpendicular to the prevailing westerly winds blowing across a large ocean expanse (Sturman and Tapper, 1996; Sturman and Wanner, 2001). This westerly-dominated flow, modified by the topography of the Southern Alps, also influences the temperature regime of the South Island climate with generally cooler conditions in the mountains and warmer and drier conditions to the east of the main divide. The North Island has two main alpine regions, the Central North island volcanic plateau and Mt Taranaki. Other, lesser ranges also experience The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 3

15 snow (such as the Tararua, Ruahine, Kaimanawa and Kaweka Ranges), however the total snowfall amounts are relatively small and there is no current skiarea infrastructure in these ranges. Figure 1 shows the median annual precipitation and temperature distribution for New Zealand. Figure 1: Median annual precipitation and average temperature distribution for New Zealand 3. Climate Data To simulate seasonal snow, climate information is required for the current climate, and for the two future time periods, referred to in this work as the 2040s ( , mid point reference 2040) and 2090s scenarios ( , mid point reference 2090) Current climate data Tait et al. (2006) describe a statistical method to interpolate station observations of precipitation to a 0.05 degree grid (approx. 5km by 5km), and generate a set of highresolution daily climate maps for New Zealand. Each daily precipitation map is The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 4

16 created by using thin-plate splines to interpolate 24-hour precipitation observations at each station. The daily maximum and minimum temperature maps are produced by extrapolating observed daily maximum and minimum temperatures to mean sea level (MSL) using a lapse rate of 5 C/km, interpolating the MSL temperatures onto the 0.05 degree lat/long grid using a bilinear interpolation, and then extrapolating the gridded MSL temperatures to the grid elevations using the same 5 C/km lapse rate (Tait, 2008). The precipitation grids (24-hour rainfall total from 9am) begin in 1960, while the temperature grids (Maximum and Minimum temperature over 24 hours from 9am) are available from These grids continue to be automatically updated daily, as an adjunct to NIWA s Climate Database. This set of data is named the Virtual Climate Station Network (VCSN) data. The method used to interpolate the temperature obviously has a significant effect on the snow simulation. We have elected to use the BA5a method (as discussed above), which lapses observed daily maximum and minimum temperatures to mean sea level (MSL) using a lapse rate of 5 C/km, interpolates the MSL temperatures onto the 0.05 degree lat/long grid using a bilinear interpolation, and then lapses the gridded MSL temperatures to the grid elevations using the same 5 C/km lapse rate. Note, other lapse rates and other interpolation schemes are also being tested but are not reported here. A thorough assessment of the various temperature interpolation methods is outside the scope of this current research project. For this study, we have used these gridded products to model snow for three 20 year periods; 1980 to 1999 for the current climate, and 2030 to 2049 and 2080 and 2099 for two future climates. For the snow modelling a year is defined as the period from April 1 to March 31 the following year, as is standard in Southern Hemisphere snow and glaciology community Climate change scenarios Overview Global climate models used for the IPCC Fourth Assessment Report formed the basis of a recently revised set of climate change projections for New Zealand. These are described in detail in a guidance manual for local government (Ministry for the Environment, 2008). This section will make a short summary of the most pertinent issues in relation to this work, but the Ministry for the Environment (2008) report should be consulted for more details. The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 5

17 Emissions scenarios Climate change projections for New Zealand are derived from the results of global climate models which have been used in the IPCC s Fourth Assessment Report to simulate future climate. These global models have been run with a range of possible future greenhouse gas emissions, and the results of the models are sensitive to the choice of emission scenario. For this study, three different emission scenarios have been selected: A1B, A1FI and B1. Each of these scenarios depicts a storyline about global population growth, industrial development, transport development and many other factors. Figure 2 provides an overview of the relative positions of these scenarios with respect to global surface warming. A1B can be described as a middle of the road emissions scenario relative to other IPCC scenarios, neither particularly high nor particularly low. A1FI is a high range scenario, while B1 is a low range scenario. See Appendix A1 of Ministry for the Environment 2008 for a detailed discussion of emissions scenarios. Figure 2: IPCC multi-model temperature projections for selected scenarios. The grey bars to the right show the range in global warming for a range of scenarios. In this report B1, A1B and A1FI are used. The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 6

18 Downscaling The spatial coarseness of the output of a global climate model (grid points typically km apart) requires a method of downscaling to be used when considering regional applications. The global model output has been statistically downscaled to the same 0.05 degree grid used for the VCSN. The methodology for downscaling temperature and precipitation to the Virtual Climate Station Network grid is described in Ministry for the Environment (2008). From the IPCC Fourth Assessment Report, results are available for over 20 different global models. The average scenario adopted for this work is based on the average result of downscaling 12 models that perform acceptably in simulating the past climate of New Zealand and the South Pacific (Mullan and Dean, 2009). Additionally, for the A1B scenario we have also used the downscaled scenarios from each of the 12 models individually as input to the snow simulation model (rather than just using their average), so that a better appreciation of climate scenario uncertainty can be obtained. The downscaled global models do sometimes differ significantly from one another as shown in Table 1 (see Appendix A3 in Ministry for the Environment (2008) for detailed discussion and maps). Table 1: Annual temperature changes (in C) relative to for 12 General Circulation Models forced by the SRES A1B scenario. Changes are shown for different end periods for the global and downscaled New Zealand average (Modified from Ministry for the Environment, 2008, p121). Model (Country) Global Change to Change to change to Global avg NZ avg Global avg NZ avg cccma_cgcm3 (Canada) cnrm_cm3 (France) csiro_mk30 (Australia) gfdl_cm20 (USA) gfdl_cm21 (USA) miroc32_hires (Japan) miub_echog (Germany/Korea) mpi_echam5 (Germany) mri_cgcm232 (Japan) ncar_ccsm30 (USA) ukmo_hadcm3 (UK) ukmo_hadgem1 (UK) model average The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 7

19 Scenario scaling Downscaled changes for monthly mean temperature and precipitation have been calculated on the VCSN 5km grid for 12 different models for 2040 ( ) and 2090 ( ) under the A1B emission scenario and then rescaled for the other two emissions scenarios considered (B1 and A1FI). Rescaling is done by taking the 12 model A1B range in the global temperature increase to 2100 (i.e C), and calculating the factors required to match this to the IPCC likely ranges for the other two scenarios, while maintaining the same relative spacing (in terms of global temperature change) between the models. This scaling factor is then applied to the local change (in temperature, precipitation, etc.) from the downscaling, where only 12 models are ultimately considered. This assumption of a proportional relationship between the global temperature change and a local change is a very common one in integrated assessment modelling (Kenny et al 2001). The scaling factors vary between about 0.6 for the B1 scenario to about 1.0 for A1B and 1.4 for A1FI (but are modeldependent). Thus, the downscaled changes for the more extreme scenarios (B1, A1FI) have exactly the same spatial pattern as A1B, but with a different amplitude Empirical adjustment of daily rainfall data As a result of climate change, heavier and/or more frequent extreme precipitation is expected over New Zealand, especially where the mean precipitation is predicted to increase. The percentage increase in extreme precipitation is expected to be approximately 8% per degree Celsius of temperature increase (Ministry for the Environment, 2008). The precipitation scenarios previously developed in Ministry for the Environment (2008, their section 2.2.2) apply to the monthly time-scale only. A methodology for empirical adjustment of a daily rainfall time series is described in Ministry for the Environment (2010) that allows for the scenarios of mean precipitation change and also adjusts the distribution to increase the most extreme daily amounts. The distributional adjustment has the effect of decreasing the number of days per year when precipitation occurs, and pushing more precipitation into the upper tail of the distribution. This work is based on analysis of extreme precipitation at a few gridpoints in the Wellington region for one of the Regional Model runs. Further work is required to clarify how appropriate this formula is across all of New Zealand. However in the meantime this method is the current best practice to adjust extremes in precipitation so that it is consistent with our current understanding of climate change. The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 8

20 Application to this study Statistically downscaled change projections were developed on the VCSN grid mentioned above. The directly downscaled changes were for monthly mean values of temperature and precipitation. The change referred to is the difference between the period (2040s), and (2090s), when compared with the base period (1990s or current). From the downscaled precipitation and temperature changes, change projections for the 2040s and 2090s periods were developed for use as input for the snow model. In all cases the detailed future climate was developed by applying the downscaled changes to the current climate VCSN data Mean temperature and precipitation changes Figure 3 shows the projected average annual changes in rainfall and temperature for New Zealand. These projections are derived from the average of 12 downscaled global model projections, assuming the A1B emissions scenario. As we have scaled the B1 and A1FI scenarios from the A1B scenario the patterns of change will be identical, only the magnitude will change. Figure 3: Projected annual mean temperature and precipitation change for the 12-model average A1B scenario for 2040 relative to The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 9

21 3.4. Adjustment of daily rainfall data The empirically adjusted daily rainfall method as described above results in more precipitation in the future climate on days with high precipitation, and an increased number of days without rainfall. The magnitudes of these changes are proportional to the temperature change and so the least change is for the B1 scenario (least warming) and the largest change is for the A1FI scenario (most warming). This captures the concept and current understanding of a future climate with larger extreme precipitation events, especially in areas with an increase in the mean precipitation Assumptions and limitations The daily temperature and rainfall VCSN data for the current climate come with the caveats noted in the associated publications (Tait et al., 2006; Tait, 2008). In particular we note that maximum and minimum temperatures are interpolated independently, so on very rare occasions there may be days where the minimum temperature is greater than the maximum. Note also that the temperature downscaling is based on Tmean only, and the same offset is assumed to apply to both Tmax and Tmin. Particularly relevant to the snow simulations are the well-known challenges in estimating precipitation and temperature in alpine regions from a sparse station network (Clark and Slater, 2006; Slater and Clark, 2006). Limitations of the climate change projections are discussed in Ministry for the Environment (2008); see especially the discussion of uncertainty of downscaling in Appendix A.3.1 of that report and the discussion of differences between global models in Appendix A3.2 of that report. The empirical adjustment of daily rainfall is based on analysis of extreme rainfalls at a few grid-points in the Wellington region for one of the Regional Model runs and further work is required to clarify how appropriate this formula is across all of New Zealand. We also do not change the likelihood of particular precipitation events, or storm tracks. 4. Methods 4.1. Snow model The snow model used in this work is the temperature index snow model described by Clark et al., (2009). A temperature threshold discriminates between rain and snow based on whether the input temperature is above or below the threshold, and a melt factor converts the difference in temperature from the threshold to available melt energy (Equation 1a-c): The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 10

22 dswe dt = a s m s (1a) a s 0, = p, T T T < T accm accm (1b) ( T T ) M f melt, T Tmelt ms = (1c) 0, T < Tmelt or SWE = 0 where SWE is snow water equivalent (mm), a s is the rate of snow accumulation (mm day -1 ), m s is the snow melt rate (mm day -1 ), p is the precipitation rate (mm day -1 ), T is air temperature (K), M f is the melt factor (mm K -1 day -1 ), T accm is the temperature threshold to distinguish between rain and snow (K), and T melt is the temperature threshold for snow melt (K). The model also accounts for the temporal variability in the melt factor by considering seas the mean melt factor M seasonal variability δ, changes in snow albedo immediately after fresh snowfall f δm α f M f, and enhanced melt during rain-on-snow ros events δ M f. The resulting melt factor at any given time is therefore a sum of a mean melt factor and the three time varying melt contributions (Equation 2): M f seas ros ( M + δm + δm α + δm,0) = max (2) f f f f The model uses an hourly time step to calculate SWE for all third order-river basins in the South Island (20,042) and North Island (17,712) of New Zealand. As the elevation within model sub-basins often varies considerably, especially in the mountains, estimating a representative temperature for a model sub-basin can be problematic. Therefore, the model uses 100m elevation bands to account for variability in temperature within model sub-basins. So each sub basin is further subdivided by 100m elevation bands, producing 139,629 and 61,096 independent modelling elements for the South Island and North Island respectively, or a total of 200,725 for New Zealand. The use of elevation bands does increase the number of model elements, but provides more realistic snow simulations. The snow model was developed primarily for hydrological applications, and as such calculates snow in terms of the snow water equivalence (SWE). However, many users of snow as a frozen resource (rather than as melt-water) think of snow in terms of snow depth. Therefore, for the purpose of this report, results are either shown in mm of SWE or in m of snow depth. Where snow depth is presented, mm of SWE have The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 11

23 been converted to m of snow depth using a varying snow density / snow depth relationship. This relationship captures our understanding of snow processes in that the early season snowpack has a low density, which gradually increases over time as the snowpack ripens and becomes isothermal. Our simple snow density / snow depth function starts on 01 April with 200mm of SWE representing 1m of snow depth, increasing at a linear rate to 400mm of SWE representing 1m of snow depth by 01 December of that year. This time period (April December) covers all available snow depth observations that were available for comparative use Simulating historical SWE Using the snow model described above, we simulate SWE for the period The simulations are driven by the daily precipitation grids from the VCSN and the daily temperature grids using the BA5a interpolation method (lapse rate of 5 C/km). The precipitation is disaggregated from daily values to hourly values using multiplicative random cascades while the hourly temperature data is obtained from a Sine curve fitted to the maximum and minimum temperatures (Clark et al., 2009). Clark et al., (2009) ran the snow model for the South Island of New Zealand with a range of parameter sets, each with a unique subset of parameters defining the melt and accumulation threshold, the melt factor and the seasonal amplitude. Each parameter set was evaluated using a range of measures for seasonal snow, including water balance estimates, point observations, and the Technical Committee on Snow (1969) snow classification. A thorough discussion of the methods use to evaluate the selection of a parameter set can be found in Clark et al., (2009). The selected ideal parameter set was defined with a temperature threshold of 274K (i.e. +1 C), a melt factor of 5 mm K -1 d -1 and a seasonal amplitude of 5 mm K -1 d -1. Using this parameter set, twenty years of daily precipitation and temperature data were used to simulate historical SWE for the South Island of New Zealand. The simulations produced by Clark et al. (2009) were qualitatively consistent with the available information on both snow duration and seasonal snow storage, and, acknowledging the afore-mentioned limitations of using an empirical snow model to simulate conditions outside the calibration period (Milly et al., 2008), the model has some utility in simulating the potential impact of projected climate change on seasonal snow Simulating future SWE scenarios To simulate SWE for the 2040s and 2090s future climate change scenarios, we repeated the above-described snow simulation, altering only the temperature and precipitation input data in line with the three emissions scenarios for the two time The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 12

24 periods. The snow model and model parameter set were unchanged for the future snow simulations. Future scenarios of SWE were then compared to the current baseline (i.e. the historical SWE simulations). Three key summary statistics; Mean maximum snow accumulation in each year, Snow duration and fraction of precipitation which is snow, were plotted as spatial plots for New Zealand. Difference plots were also created showing the change (absolute and relative) from current to the 2040s and 2090s. Graphs have also been generated to show these three key metrics with elevation and have the current and both the 2040s and 2090s data over-plotted to permit convenient comparisons. A graph comparing relative changes, compared with current SWE, in each of the three metrics with elevation is also generated. 5. Results This section will provide some of the key results from all of the different scenarios, time periods and metrics used. A complete set of outputs is not presented, merely a selected set that provide an overview and summary of the key results. Table 2 provides a summary for the three metrics at a range of elevations for the three emissions scenarios. Figure 4 presents the mean maximum snow accumulation in each year (mm) for the current (1990s), 2040s and 2090s for all of New Zealand using the 12-model average A1B scenario as input for the two future scenarios. Figure 5 presents the percentage difference in mean maximum snow accumulation in each year (mm) for 2040s and 2090s using the 12-model average A1B scenario as input, when compared to the current mean maximum snow accumulation in each year. Figures 6 and 7 present the same plots as Figures 4 and 5, but for the warmer A1FI scenario. Nationwide plots for the B1 scenario have also been generated, but are not presented here as they show the same general pattern, but with a smaller magnitude than the A1B plots (Table 2). The largest changes, in absolute amounts, in both the 2040s and 2090s A1B and A1FI scenarios, is along the Main Divide of the Southern Alps and in the Southern and Western regions of the South Island where a decrease is noted (Figures 4 and 6). These are locations that have the greatest values of mean maximum snow accumulation in each year in the current simulations. We also observe a marked decrease in the Central North Island and in the lesser ranges in the Southern parts of the North Island. In relative terms, the greatest percentage change is at low elevations areas and in regions to the East and North of the Main Divide in South Island and at all low elevation areas in the North Island (Figures 5 and 7). These plots of changes in The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 13

25 mean maximum snow accumulation in each year show a generally similar response as the other metrics with an overall decrease in nearly all areas for both the 2040s and 2090s scenarios. Current 2040s 2090s Current 2040s 2090s Figure 4: Mean maximum snow accumulation in each year (mm of SWE) for the current, 2040s and 2090s scenarios using the 12-model average A1B scenario as input. The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 14

26 2040s 2090s 2040s 2090s Figure 5: Percentage difference in mean maximum snow accumulation in each year for 2040s and 2090s using the 12-model average A1B scenario as input, when compared to the current. The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 15

27 Current 2040s 2090s Current 2040s 2090s Figure 6: Mean maximum snow accumulation in each year (mm of SWE) for the current, 2040s and 2090s scenarios using the 12-model average A1FI scenario as input. The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 16

28 2040s 2090s 2040s 2090s Figure 7: Percentage difference in mean maximum snow accumulation in each year for 2040s and 2090s using the 12-model average A1FI scenario as input, when compared to the current. The three metrics from the snow model can also be plotted as area-weighted frequency distributions for selected elevation ranges for the current, and the two future scenarios. For conciseness we only show one metric, fraction of precipitation that falls as snow for the South Island (Figure 8). Other metrics show a similar general pattern (Table 2). We have grouped the data into five elevation ranges; 0-500m, m, m, m and above 2000m and plotted the area weighted frequency distribution for the 1990s and the 2040 and 2090s using the 12-model average A1B scenario. The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 17

29 (A) 0-500m (B) m (C) m (D) m (E) 2000m + Figure 8: Frequency distributions of fraction of precipitation that falls as snow for the 1990s (red), 2040s (green) and 2090s (blue) for elevation ranges; (A)0-500m; (B) m; (C) m; (D) m; (E)2000m+, using the 12-model average A1B scenario. Median for each time period is shown as the dashed vertical line. Fraction of precipitation that falls as snow is along the x axis and varies for each figure, from 0 to 0.09 to 1.0 and percentage area weighted frequency is on the y axis from 0% to 30. The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 18

30 At each elevation range (Figure 8A-E), we observe a decrease in the median fraction of precipitation that falls as snow. We also note a change in the distribution with a tendency towards more positively skewed distributions (i.e. more low values) in elevations below 1500m (Figure 8A, B, C). This change in the median and the distribution has a substantial impact, especially by the 2090s. For example at the elevation range of m the 2090s median is approximately equivalent to the 10 th percentile (a low snow year) in the current conditions. At elevations above 1500m there is still a clear decrease in the median, but the change in the distribution is less marked (Figure 8D, E). Figure 9 presents a synopsis of the results for all three metrics as percentage change for the 2040s and the 2090s for an A1B scenario, relative to the current, plotted against elevation for the South Island. In presenting these results we will highlight the changes for both the 2040s and 2090s scenarios at three elevations; 2000m, 1000m and near sea level (0 to 100m elevation band). Figure 9A shows an overall decrease in snow duration for elevations below 2900m in both the 2040s and 2090s scenarios. This decrease is evident at all elevations below 2900m and is especially marked at lower elevations. At an elevation of 2000m the average change in snow duration is approximately -5% by the 2040s and -16% by the 2090s. At 1000m the average decrease in snow duration is approximately -11% by the 2040s and -30% by the 2090s (Figure 9A). At lower elevations there is a greater percentage decrease in snow duration, with an average change of -38% by the 2040s and -78% by the 2090s at near sea level (elevation band from 0 to 100m). The fraction of precipitation that is snow in both the 2040s and 2090s A1B scenario also shows an overall decrease (Figure 9B). This decrease is evident at all elevations, but is especially marked at lower elevations. At an elevation of 2000m the change in fraction of precipitation that is snow is on average approximately -11% by the 2040s and -33% by the 2090s. At 1000m the decrease in the fraction of precipitation that is snow is on average approximately -24% by the 2040s and -56% by the 2090s (Figure 9B). At lower elevations there is a greater percentage decrease in the fraction of precipitation that is snow, decreasing on average by -57% by the 2040s and -87% by the 2090s at near sea level. The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 19

31 A B C Figure 9: (A) Percentage difference in snow duration, (B) fraction of total precipitation that is snow and, (C) mean maximum snow accumulation in each year (average within 100m elevation bands) for the 2040s (green) and 2090s (blue) using the 12-model average A1B scenario, compared to the current. Elevation is along the x axis from 0 to 3700m and relative percentage change is on the y axis from +100% to -100%. The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 20

32 The mean maximum snow accumulation in each year within a 100m elevation band also shows a substantial decrease at nearly all elevations, except for the very highest elevation locations (above approximately 2900m) for the 2040s A1B scenario, where a slight increase is observed. At an elevation of 2000m the change in mean maximum snow accumulation in each year is on average approximately -8% by the 2040s and -30% by the 2090s. At 1000m the decrease in the mean maximum snow accumulation in each year is on average approximately -21% by the 2040s and -58% by the 2090s (Figure 9C). At lower elevations there is a greater percentage decrease in the mean maximum snow accumulation in each year, decreasing on average by -49% by the 2040s and -84% by the 2090s at near sea level. A summary for this scenario and other two (B1 and A1FI) is presented in Table 2. As outlined above, we have used the 12-model average A1B scenario for the snow model results shown (Figures 4, 5, 8, 9), and then also scaled the input data for the other two emissions scenarios (B1 and A1FI) (Figures 6 and 7 for A1FI). Additionally, for the A1B scenario we have also used the downscaled scenarios from each of the 12 models individually as input to the snow simulation model (rather than just using their average), so that a better appreciation of the impact of climate scenario uncertainty can be obtained. Figure 10 presents synopsis graphs of the results for each of the 12 snow model runs for the A1B scenario for the South Island, for all three metrics as percentage change for the 2040s and the 2090s relative to the current, plotted against elevation. Figure 10A shows an overall decrease in snow duration in both the 2040s and 2090s scenarios for nearly all 12 models, for elevations below approximately 2900m. At an elevation of 2000m the change in snow duration ranges from approximately +1% to -11% (mean -6%) by the 2040s, and from -8% to -25% (mean -15%) by the 2090s. At 1000m the change in snow duration ranges from approximately -3% to -18% (mean -13%) by the 2040s, and from -15% to -47% (mean -31%) by the 2090s (Figure 10A). At lower elevations there is a greater percentage decrease in snow duration, showing a decrease from -20% to -63% (mean -45%) by the 2040s and from -53% to -93% (mean -76%) by the 2090s at near sea level. The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 21

33 A B C Figure 10: (A) Percentage difference in snow duration, (B) fraction of total precipitation that is snow and, (C) mean maximum snow accumulation in each year (average within 100m elevation bands) for the 2040s (green) and 2090s (blue) using each of the 12 A1B scenario GCMs individually, compared to the current. Elevation is along the x axis from 0 to 3700m and relative percentage change is on the y axis from +100% to -100%. The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 22

34 Table 2: Summary of average percentage difference for the South Island, in each of the three metrics at three elevations for the 2040s and 2090s using the 12-model average B1, A1B and A1FI scenarios as input, when compared to the current. Emissions Scenario B1 A1B A1FI Elevation 2040s now Duration 2090s Snow Duration 2040s Fraction of precip as snow 2090s Fraction of precip as snow 2040s Max. snow* 2090s Max. snow* 0-100m 1000m 2000m 0-100m 1000m 2000m 0-100m 1000m 2000m * Where Max. snow is the mean maximum snow accumulation in each year. If we consider Figure 10C, we observe that the mean maximum snow accumulation in each year within a 100m elevation band, also shows a substantial decrease at nearly all elevations, except for the very highest elevation locations (above approximately 2900m) for the 2040s scenario, where an increase is observed. At an elevation of 2000m the change in mean maximum snow accumulation in each year ranges from approximately +8% to -22% (mean -9%) by the 2040s and from -6% to -51% (mean -26%) by the 2090s. At 1000m the change in the mean maximum snow accumulation in each year ranges from approximately -3% to -44% (mean -28%) by the 2040s and from -32% to -79% (mean -57%) by the 2090s (Figure 10C). At lower elevations there is a greater percentage decrease in the mean maximum snow accumulation in each year, with a change ranging from -23% to -75% (mean 53%) by the 2040s and from -56% to -95% (mean -82%) by the 2090s at near sea level. While we have confidence in our methodology, we still urge caution regarding reliance on the very precise percentage values of changes in the snow metrics as presented. We have employed a relatively simple delta change methodology for the climate change analysis and there is some uncertainty in these values. So, while all of these results are presented to the nearest 1% the values are indicative rather than precise. The Potential Impact of Climate Change on Seasonal Snow Conditions in New Zealand 23