Projecting Climate Change-Influenced Shirts in Georgian Viniculture Using a Species Distribution Model

Projecting climate change-influenced shifts in Georgian viniculture using a species distribution model. A PAPER SUBMITTED FOR COMPLETION OF SENIOR RESEARCH FOR THE COLLEGE OF ARTS AND SCIENCES STETSON UNIVERSITY BY J.B. Pitts IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF BACHELOR OF SCIENCE ENVIRONMENTAL SCIENCE ADVISOR Dr. Jason M. Evans, Ph.D. April 2017 i Table of Contents  List of Figures……………………………………………………………………………………………………………........ii  List of Tables…………………………………………………………………………………………………………….........iii  Acknowledgements………………………………………………………………………………………………………….iv  Abstract…………………………………………………………………………………………………………….................v  Introduction……………………………………………………………………………………………………………..........1  Literature Review……………………………………………………………………………………………………………..1  Study Area…………………………………………………………………………………………………………….............3  Methods……………………………………………………………………………………………………………................4  Observations and Results…………………………………………………………………………………………………6  Discussion and Conclusion……………………………………………………………………………………………..12  Works Cited……………………………………………………………………………………………………………........14 ii List of Figures  Figure 1.1……………………………………………………………………………………………………………..........7  Figure 1.2……………………………………………………………………………………………………………..........8  Figure 1.3……………………………………………………………………………………………………………..........9  Figure 1.4……………………………………………………………………………………………………………........10  Figure 1.5……………………………………………………………………………………………………………........11 iii List of Tables  Table 1.1……………………………………………………………………………………………………………..........5  Table 1.2……………………………………………………………………………………………………………........12 iv Acknowledgements: I thank Dr. Jason Evans for his guidance in developing this project, Dr. Michael Denner for the idea and for providing background on the Republic of Georgia, Dr. Tony Abbott for his assistance in map design, Stetson University’s Environmental Science department for providing me with a home for the past two years, and everyone who has supported me along the way. v Abstract: While climate change has thus far had a mostly positive impact on global grape growing, it is predicted that as temperatures continue to rise there will be regional shifts in where specific varieties of wine can be produced. In the Republic of Georgia, a mountainous country with three distinct climatic zones and an 8,000 year history of winemaking, there is a strong connection between wine style and place. Using climate data based off of IPCC projections, this project ran a maximum entropy species distribution model on four Georgian grape species in order to analyze how the winemaking regions of Georgia could change in the future. The models were strongly predictive (each AUC value > 0.917) of the current distribution, and projected that each of the four species analyzed will have suitable habitat in both their current locations and a new region by the year 2070. While the high AUC value allows for confidence in these projections, a range of other factors such as soil type and economic viability will also affect future distributions. I therefore suggest these results be used in a future survey of Georgian winemakers. 1 Introduction Wine style for a specific region is generally dependent upon that region’s baseline climate (Jones 2007). As the Earth enters into a period of climate instability likely to be unprecedented in human history, the practice of regional viniculture, or growing grapes to make wine, will likely be affected. To this point, warming temperatures have had a mostly positive impact on grape growing (Jones & Davis 2000). However, it is predicted that as temperatures continue to rise there will be regional shifts in where certain varieties of wine can be produced (Gaidos 2014). In the Republic of Georgia, where evidence of winemaking has been found dating back to the sixth millennium BCE (Glonti 2010), there are very clear cultural expectations regarding the characteristics of regional wines and the techniques used in viniculture. This project aims to use a maximum entropy model to conduct site suitability analyses (Holland & Smit 2010) of several Georgian microregions and make predictions on how winemaking will be affected by changes to their climates over the next 40 years. These models for projecting future climate are mathematical representations of our Earth/Atmosphere system that include spatial and temporal analyses of the laws of energy, mass, moisture, and momentum transfer in the atmosphere and between the atmosphere and the surface of the Earth (Jones 2007). The results of this study will be used in a future qualitative study asking grape growers and winemakers from the analyzed microregions how viticulture and winemaking will change moving forward. Literature Review While climate change is generally viewed as an environmental issue, many of the factors influencing it are social and therefore climate change must also be considered within a socioeconomic context (Fischer et al. 2002). As Pincus (2003) notes, cultural connections between wine and location are a reflection of the local climate. Most of today’s quality wine production occurs in viticultural regions with relatively narrow geographical and climatic niches, putting them at greater risk for both short-term climate variability and long-term climate change (Jones 2007). With the world’s climate being generally stable for approximately 8,000 years – as long as wine has been made in Georgia (Glonti 2010, McGovern 2013) – this has not been of significant concern in the past. However, it is logical to believe that cultural connections to wine, including the specific winemaking practices used in a certain region that have been developed over thousands of years, will be challenged as the climate shifts (Pincus 2003). 2 Viticulture is particularly sensitive to changes in temperature and carbon dioxide (Bindi et al. 1996). In some ways, it is also more vulnerable than other forms of agriculture, as decisions can take years to play out – for example, if a winegrower decided to replant their vineyard, it would take at least a decade to produce the highest quality grapes (Pincus 2003). To this point, climate change has had a mostly positive impact on winemaking, with earlier phenological events, shorter phonological intervals, and a longer growing season being associated with higher grape yields and improved wine quality (Jones & Davis 2000, Gaidos 2014). However, with temperatures expected to rise another two degrees Celsius in the next 50 years, Jones (2007) posits we will likely begin to see more rapid plant growth and out of balance ripening profiles, resulting in loss of acidity and wines with a higher than normal alcohol content. This would mean that some of the world’s most renowned grape growing regions would no longer be suitable for their trademark grape varieties. However, this does not necessarily mean these grapes will stop being produced, only that there will be regional shifts in where certain wines are made, creating a variety of social pressures; for example, Yellowstone is predicted to be one of the best places to grow grapes in 2050 (Gaidos 2014). It will also challenge the sense of place associated with certain varieties of wine. This makes Georgia a particularly interesting case study in climate adaptation for viniculture. In addition to being the potential birthplace of wine (McGovern 2013) and possessing a vibrant food culture (Goldstein 1999), the highly mountainous country, which is about the size of West Virginia (70K km2), nonetheless has three heterogeneous climactic zones of subtropical, hot continental, and cold. This geographic and ecosystemic diversity has allowed for a wide variety of wines – 525 grape varieties have been described there – to develop across its microregions (Glonti 2010). With any type of prediction modeling, there are inevitably inconsistencies regarding its accuracy, as it is a static, two-dimensional representation of a three-dimensional world with time-dependent processes (Goodchild et al. 1996). Climate models are based on IPCC emissions scenarios (Houghton et al. 2001), which reflect estimates of how humans will emit carbon dioxide in the future. Because of the non-linearity of the system and differences in emission scenarios, the models in use today provide a range of potential changes in the Earth’s temperature and precipitation (Houghton et al. 2001) that must be taken into consideration when analyzing future climate projections. Additionally, grape distribution is affected by a variety of cultural factors 3 (Pincus 2003) and biological factors such as soil composition (Andrés-de Prado et al. 2007) that are difficult to account for with a climate model. Holland and Smit (2010) identified three main themes within current research on the implications of climate change on viniculture: (1) impacts of climate change on wine quality; (2) impacts of climate change and variability on grapevine yield and phenology; and (3) vineyard site suitability analyses. Mechanistic growth models have commonly been used to study climate change’s impact on viticulture (Bindi et al. 1996, Webb et al. 2005). However, my study is focused on the third theme, site suitability analysis, and will therefore require a different approach. While grape distributions have been analyzed in Georgia (Nikolaishvili et al. 2014), prediction modeling for the analysis of distribution shifts, as in the Pincus (2003) study, has not been explored in the country. Site suitability analyses are generally carried out within the context of terroir, which is the unique set of environmental factors and farming practices that affect a crop’s phenotype when it is grown in a specific habitat (Holland and Smit 2010). As Evans et al. (2010) points out, species distribution models such as maximum entropy have commonly been used to study natural environments, but rarely to develop landscape models for cultivated crops. Using maximum entropy to study corn and switchgrass for biofuel production, Evans et al. (2010) found that the model was an effective predictor into potential land-use change patterns, making it a strong candidate for studying changes in regional viniculture. Study Area In addition to having an 8,000 year history of domesticated grapes (Glonti 2010, McGovern 2013), the Republic of Georgia is also defined by three distinct climate zones – subtropical, hot continental, and cold. Subtropical climates are characterized by hot, humid summers and mild to cool winters. Hot continental climates are characterized by large seasonal differences, with warm to hot summers and cold winters. Cold climates are characterized by a lack of warm summers. This geographic diversity has led to the development of 525 distinct, endemic vine varieties across Georgia’s winemaking microregions (Glonti 2010). Four of those varieties will be examined in this study, which represent a range of the different socioeconomic and viticultural factors present in the Georgian wine industry. Over 90% of Georgian wine is produced in Kakheti, the easternmost region of the country. Saperavi and 4 Rkatsiteli are the most widely produced red and white varieties, respectively, in this region, and are therefore the most economically important to the country. Aleksandrouli, another of the species being analyzed, is used in the region of Ambrolauri to produce Khvanchkara, arguably the nation’s most esteemed wine and reportedly the favorite of former Russian dictator Joseph Stalin. For two centuries, Georgia’s wine exports were mainly limited to Russia, however, a beverage ban was implemented on the former Soviet state there in 2006 (Anderson 2013). With almost half of the country’s population being rural, Georgia has been trying to develop new export markets for its wine in the time since. Finally, Tsolikouri is the most popular white variety in western Georgia, found in the regions of Georgia along the Black Sea (Georgian Wine 2014). This proximity provides a challenge to vintners due to a high prevalence of phylloxera, a pest that feeds on the roots and leaves of grapevines (Granett et al. 2001). Methods Due to time constraints and the complexities of climate modeling, my study emphasized the use of only one type of model, the MaxEnt model. Maximum entropy takes a set of samples from a distribution over space X, and using a set of environmental variables on this space, estimates the target distribution of the samples – grape varieties in the case of this study grape varieties – as subject to the given environmental variables (Phillips et al. 2004). Required data for this project were current and future climate conditions, elevation, and grape species as it relates to specified regions. Global climate data was gathered from WorldClim, which generates data through interpolation of average monthly climate data from weather stations on a 30 arc-second resolution grid (often referred to as "1 km2" resolution). This data included 19 derived bioclimatic variables as listed in Figure 1.1 (Hijmans et al. 2005). I selected future data for the year 2070 based on the representative concentration pathway (RCP) of 4.5, which represents an increase in radiative forcing by 4.5 W/m^2 by 2100. This is the second lowest increase of the four IPCC scenarios, associated with a predicted temperature increase ranging from 1.1 to 2.6 degrees Celsius (Weyant et al. 2009). Data for Georgia’s elevation was retrieved from DIVA-GIS, which derives the data from NASA’s Shuttle Radar Topographic Mission (SRTM) at a 30 arc-second resolution (Hijmans et al. 2012). Grape species for analysis were identified from the second edition of Georgian Ampelography, a study of 57 vine varieties within Georgia (Ketskhoveli et al. 2012). Google maps was then used to collect the latitude and 5 longitude of wineries where each of the four grape species are grown, and that data was entered into an Excel spreadsheet for the respective species. Table 1.1. Codes for the 19 bioclimatic variables used in the model. In order to run the MaxEnt model, it was necessary to first pre-process the environmental variables using ArcGIS so they would have matching corner dimensions and cell sizes. Because the climatic data used was on a global scale, I first batch clipped the 19 current climate rasters only for Georgia using the Georgia elevation raster. I then projected these files into the Europe Albers Equal Area Conic coordinate system in order to preserve the distance between points. I then performed a batch resample of all of the rasters to a cell size of 1000. Next, I identified which rasters did not have matching values for columns, rows, and corners and found that the elevation, BIO1, and BIO10 rasters were slightly off. I reclassed those three rasters to a value of 1, then multiplied the elevation and BIO10 files together which I saved as a file labeled “Pre-Mask.” I then multiplied the Pre-Mask file by the BIO1 file to create a “Mask” file, which I then batch multiplied all 20 rasters by so that columns, rows, and corners were matching. At this point the bioclimatic variables were saved using their names as identified in Table 1.1 rather than their codes. Once it was confirmed that the files all matched, I batch converted the raster files into ASCII files, which were saved to a “Current ASCII Files” folder, for use in the MaxEnt program. Once the current climate data had been pre-processed, I then did the same with the 19 future climate rasters, starting over with the original elevation file. I used the same mask from the current data to make sure that the future rasters also matched up with the current rasters. Once all the files matched up, I batch converted the future raster files into ASCII files of the same name as those in the “Current” folder and saved them to a “Future ASCII Files” folder. 6 For species data that could be run with the MaxEnt program, I created CSV (comma delimited) files from excel spreadsheets for each species. After identifying latitude and longitude locations of growing locations using Google Maps, I entered the species name, latitude, and longitude into separate columns. Seventeen points were identified for Saperavi, 14 for Rkatsiteli, and 8 for Aleksandrouli and Tsolikouri, respectively. To line the species data up with the environmental layers, I added the CSV files to the map document, created a shapefile from their XY data, and defined their projection as GCS_WGS1984. I then projected the shapefiles to the Albers Equal Area projection. I then opened the attribute tables and calculated field geometry of latitude and longitude, using the y and x coordinates respectively, to convert the values from degrees to meters. I exported the table as a dBase table, then opened that table in Excel and saved it as a new CSV file labeled “Species_Full_Points.” Once all the data had been gathered and converted into a usable format, two runs of each species were done through the MaxEnt program as described in Phillips et al. 2006. The species CSV file was entered into the “Samples” section, current climate data was entered into the “Environmental Layers” section, future climate data were entered as “Projection Layers,” and a folder labeled for the species was used as the output directory. The models were run with response curves and a jackknife in order to measure variable importance. After one run, I analyzed the results to identify variables with a permutation importance of 0. To eliminate those variables and reduce the number of variables that were going into the analysis, I created two new folders for current and future climate data for that species and copied only the ASCII files for those variables with a permutation importance greater than 1 into them. I then ran the model again using these new folders and used this run for final analysis, identifying the value for the Area Under Curve (AUC) test done by the MaxEnt program as a measure of the model’s predictive value. The AUC value is the probability that a randomly chosen presence site will be ranked above a randomly chosen absence site (Phillips et al. 2008). The highest possible AUC value is 1.0; a random ranking has on average an AUC of 0.5, and models with values above 0.75 are considered potentially useful (Elith 2002). Observations and Results Similar to projections of other winemaking regions by Pincus (2003) and Gaidos (2014), the models for each of the four species project an increase in suitability in growing regions where they are not currently found by the year 2070. Saperavi, as seen in Figure 1.1, projects to shift to 7 the north and west, including an increase in suitability in Russia and Armenia. Rkatsiteli, seen in Figure 1.2, has a similar but less pronounced northwestward shift, and an even stronger suitability than Saperavi in Russia. Figure 1.3 highlights the Kakheti region of Georgia for Saperavi, where both of these species are most commonly produced, which projects to be less suited overall for the two species in 2070. Figure 1.1. Maximum entropy model outputs for Saperavi vine variety currently (top; AUC = .945) and in 2070 (bottom). Colors identify the probability that suitable climate will be in that area, with darker colors representing a probability closer to 1. 8 Figure 1.2. Maximum entropy model outputs for Rkatsiteli vine variety currently (top; AUC = .966) and in 2070 (bottom). Colors identify the probability that suitable climate will be in that area, with darker colors representing a probability closer to 1.9 Figure 1.3. Maximum entropy model outputs for Saperavi vine variety currently (top; AUC = .945) and in 2070 (bottom), with a focus on the Kakheti winemaking region. Colors identify the probability that suitable climate will be in that area, with darker colors representing a probability closer to 1. 10 Aleksandrouli and Tsolikouri, the two analyzed species more commonly grown in the west of Georgia, project to have an increasingly suitable climate south of their current ranges, as seen in Figures 1.4 and 1.5, respectively. Tsolikouri in particular projects to undergo a sizable increase in suitability southward around the Black Sea and into Turkey. Figure 1.4. Maximum entropy model outputs for Aleksandrouli vine variety currently (top; AUC = .970) and in 2070 (bottom). Colors identify the probability that suitable climate will be in that area, with darker colors representing a probability closer to 1. 11 Figure 1.5. Maximum entropy model outputs for Tsolikouri vine variety currently (top; AUC = .917) and in 2070 (bottom). Colors identify the probability that suitable climate will be in that area, with darker colors representing a probability closer to 1. 12 AUC values of the current distributions for the four species ranged from .917 for Tsolikouri to .970 for Aleksandrouli, indicating that the model had a high rate of predictive success. Variables of importance as determined by modeled response curves, shown in Figure 1.6, varied among each species. Average precipitation of the driest month was the only of the twenty variables measured to factor into the range of all four species. Saperavi had the most, with 7 factors above 1.0%, however, maximum yearly temperature was overwhelmingly the most important factor for determining its range. Aleksandrouli had the fewest, with only 5 factors above 1.0%, but was the only of the four species where precipitation variables held the highest importance rather than temperature variables, and the only to be significantly affected by elevation. Tsolikouri had the most varied profile, with four variables having an effect of at least 10% and three above 20% in determining its suitable range. Rkatsiteli was the most affected by precipitation levels, with five of its six determining variables relating to it. Saperavi % Rkatsiteli % Maximum Annual Temperature 67.7 Mean Temperature of Coldest Quarter 41 Annual Mean Temperature 9 Mean Temperature of Driest Quarter 28.2 Precipitation in Driest Month 9 Precipitation in Driest Month 18.9 Temperature Seasonality 6.3 Precipitation in Wettest Month 6.8 Precipitation in Wettest Month 3.3 Precipitation Seasonality 4.1 Diurnal Range 2.8 Mean Temperature of Wettest Quarter 1.1 Precipitation in Driest Quarter 1.4 Aleksandrouli % Tsolikouri % Precipitation in Driest Month 41 Mean Temperature of Driest Quarter 37.2 Precipitation Seasonality 32.1 Isothermality 25.9 Temperature Seasonality 19.1 Temperature Seasonality 23.5 Elevation 4.4 Precipitation in Warmest Quarter 10 Diurnal Range 3.3 Precipitation in Driest Month 1.7 Precipitation in Wettest Month 1.6 Table 1.2. Permutation importance of environmental variables for second MaxEnt run of four Georgian grape species. Variables with a percentage below 1.0 following the first run were eliminated for each species. Discussion and Results Wine style is culturally connected (Pincus 2003) because grape species are highly dependent upon the baseline climate of the region where they are grown (Jones 2007). This is 13 particularly true in a nation such as Georgia with a renowned gastronomic culture, diverse microclimates due to its mountainous terrain, and an 8,000 year history of winemaking. Further analysis of these species’ growing conditions would help to further the understanding of the future of Georgian winemaking. Soil type has a major impact on nutrient and water availability for grape vines, which influences growth rates and wine quality (de Andrés-de Prado 2007). Due to data and travel constraints, soil data was not able to be incorporated into this study, which would change if grapes can be grown in an area that has a suitable climate. Grapes are also particularly sensitive to changes in carbon dioxide (Bindi et al. 1996), another variable that was not tested in the models here but would be increasing in any climate change scenario. Finally, it is also possible that climate change could affect the ripening profile of these species (Jones 2007). While this would not necessarily change where the grape is able to grow, it could change the taste and style of the wine, which is ultimately why a certain species is chosen to be grown. Due to these additional factors, I suggest that the results of this study be utilized in a qualitative study of Georgian vintners to better assess how Georgian winemaking could change in the future. In terms of climate projection, the 4.5 RCP used in this study is a relatively conservative projection, so it would be interesting to see projections done using other pathways to compare differences in future projections. Additionally, incorporation of more data, such as soil types, and a more advanced method of species point data collection, such as the use of GPS technology, could help to provide the most accurate possible results. Georgia was a major supplier of wine to Russia for over 200 years, until Russia imposed a beverage ban on the former Soviet state (Anderson 2013). With almost half of Georgia’s population being rural, and most farmers raising a vineyard, the country has been seeking to develop new export markets for its wine. Currently, over 90% of Georgian wine is produced in the Kakheti region, which may become less suitable for growing its trademark varieties in the future due to climate change. Reduced wine production could hinder a developing economy, however, these models suggest that increased investment in winemaking in additional regions to the north and especially to the west of Kakheti may be a worthwhile cause. Since grapevines can take more than a decade to start producing the highest quality grapes (Pincus 2003), however, it is vital that a wine-producing economy like Georgia is prepared for changes in viniculture well before those changes need to be made. 14 Works Cited Anderson, K. 2013. Is Georgia The Next “New” Wine-Exporting Country? Journal of Wine Economics 8(1):1-28. Bindi, M., L. Fibbi, B. Gozzini, S. Orlandini, and F. Miglietta. 1996. Modelling the impact of future climate scenarios on yield and yield variability of grapevine. Climate Research 7(3):213- 224. de Andrés-de Prado, R., M. Yuste-Rojas, X. Sort, C. Andrés-Lacueva, M. Torres, and R. M. Lamuela- Raventos. 2007. Effect of soil types on wines produced from Vitis vinifera L. Cv. Grenache in commercial vineyards. Journal of Agriculture and Food Chemistry 55(3):779-786. Elith, J. 2000. Quantitative methods for modeling species habitat: comparative performance and an application to Australian plants. Quantitative methods for conservation biology. Springer:39-58. Evans, J. M., R. J. Fletcher, and J. Alavalapati. 2010. Using species distribution models to identify suitable areas for biofuel feedstock production. GCB Bioenergy 2:63-78. Fischer, G., M. Shah, and H. van Velthuisen. 2002. Climate Change and Agricultural Vulnerability. Johannesburg: International Institute for Applied Systems Analysis. Gaidos, S. 2014. GRAPE EXPECTATIONS: Climate Change Is Already Transforming the Wine Industry. Science News 185(3):20–24. Georgian Wine. 2014. Georgian Grape Varieties. <https://www.skidmore.edu/~bturner/375_citation_guidelines.htm> Accessed 24 April 2017. Glonti, T. 2010. Traditional Technologies and History of Georgian Wine. Bulletin de l’OIV 83(953):335. Goldstein, D. 1999. The Georgian feast: the vibrant culture and savory food of the Republic of Georgia. Berkeley, CA: Univ of California Press. Goodchild, M. F., L. T. Steyaert, B. O. Parks, C. A. Johnston, D. R. Maidment, M. P. Crane, and S. Glendinning. 1996. GIS and Environmental Modeling: Progress and Research Issues. New York, US: John Wiley & Sons, Inc. Granett, J., M. A. Walker, L. Kocsis, and A. D. Omer. 2001. Biology and management of grape phylloxera. Annual Review of Entomology 46:387-412. Hijmans, R. J., S. E. Cameron, J. L. Parra, P. G. Jones and A. Jarvis. 2005. Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology 25:1965-1978. Hijmans, R. J., L. Guarino, and P. Mathur. 2012. DIVA-GIS Version 7.5 Manual. University of California, California, USA. Holland, T., and B. Smit. 2010. Climate Change and the Wine Industry: Current Research Themes and New Directions. Journal of Wine Research 21(2–3):125–36. Houghton, J. T., Y. Ding, D. J. Griggs, M. Noguer, P. J. van der Linden, X. Dai, K. Maskell, and C. A. Johnson. 2001. Climate Change 2001: The Scientific Basis. IPCC. Cambridge University Press, Cambridge, England. Jones, G. V. 2007. Climate Change: Observations, Projections, and General Implications for Viticulture and Wine Production. Economics Department Working Paper 7(1):15. Jones, G. V., and R. E. Davis. 2000. Climate influences on grapevine phenology, grape composition, and wine production and quality for Bordeaux, France. American Journal of Enology and Viticulture 51(3):249-261. Ketskhoveli, N., M. Ramishvili, and D. Tabidze. 2012. Georgian Ampelography: English Language Edition. Tbilisi: Exclusive Print Press. 15 McGovern, P. E. 2013. Ancient Wine : The Search for the Origins of Viniculture. Princeton, US: Princeton University Press. Nikolaishvili, D., T. Mamukashvili, M. Sharashenidze, and D. Sartania. 2014. The Landscape Analysis of Viticulture and Enology in Georgia. Journal of Earth Science and Engineering 4:667-674. Phillips, S. J., R. J. Anderson, and R. E. Schapire. 2006. Maximum entropy modeling of species geographic distributions. Ecological Modeling 19 (3-4):231-259. Phillips, S. J., and M. Dudik. 2008. Modeling of species distributions with Maxent: new extensions and a comprehensive evaluation. Ecography 31(2):161-175. Pincus, R. 2003. Wine, Place, and Identity in a Changing Climate. Gastronomica: The Journal of Critical Food Studies 3(2):87–93. Webb, L. B., P. H. Whetton, and E. W. R. Barlow. 2005. Impact on Australian Viticulture from Greenhouse Induced Temperature Change. Modelling and Simulation Society of Australia and New Zealand:170-176. Weyant, J., C. Azar, M. Kainuma, J. Kejun, N. Kakicenovic, P. R. Shukla, E. La Rovere, G. Yohe. 2009. Report of 2.6 versus 2.9 Watts/m^2 RCPP Evaluation Panel. Geneva: IPCC Secretariat.