January 2019 Issue of Wines & Vines

The Limitations of the Winkler Index

by Patrick L. Shabam

The Winkler Index or Winkler Scale is a standard for describing regional climates for viticulture in the United States. Developed by A.J. Winkler and M.A. Amerine at the University of Cali-fornia, Davis in the first half of the 20th century, the index was constructed to corre¬late wine quality with climate, focusing on California viticulture. Wine-producing regions of California were broken into five climatic regions using heat summations above 50° F, or growing degree days (GDD). Heat summa¬tions are a way of looking at accumulated temperatures over a given time period. De¬spite the common usage of the Winkler Index, the classifications offer greater uncertainty than the system suggests.

An obvious drawback to the Winkler Index is the focus on temperature alone. Winkler’s groundbreaking resource “General Viticulture” notes the influence of “rainfall, fog, humidity and duration of sunshine,” but Amerine and Winkler’s work and subsequent research found that temperature plays the greatest role in the development of wine grapes. As the focus of Winkler and Amerine’s wine-grape research was on California viticulture, their climatic regions had the luxury of ignoring precipita¬tion, as little rain falls during the growing season. Indeed, Winkler et al. in “General Vi¬ticulture” suggest that the vinifera grape “is not suited to humid summers, owing to its suscep¬tibility to certain fungus diseases and insect pests that flourish under humid conditions.”

Advances in viticulture have expanded the range of successful commercial vinifera pro¬duction, while more recent research notes the importance of wind on photosynthesis rates. Winkler et. al. also point out the importance of ongoing refinement of local variations, stat¬ing, “It is hoped that refinements will be de¬veloped so as to delimit subregions within the present regions, thereby ensuring the greatest potential for quality when the most favorable climatic subregion for a given variety is planted to that variety.”

Therefore, the Winkler Index was designed at a state scale, and not necessarily intended for smaller viticultural subregions or single vineyards.

Methodology of the index
Modern-day calculations of GDD most often utilize daily accumulations of degree days. That is to say that they sum the total degrees of aver¬age daily temperatures above 50° F (10° C) for every day from April 1 to Oct. 31 [Σ Oct 31 Apr 1 daily (T_mean-50,0)]. If the average temperature for a given day is 75° F, then the total degree days added to the total sum for that specific date would be 25° F. Any average below 50° F con¬stitutes zero GDD for the given day.

An initial problem with some calculations is how “average” is determined. In most cases, average is calculated as the mean of the lowest recorded temperature and the highest re-corded temperature, so a high temperature of 85° F and a low temperature of 65° F would produce 75° F [(Tmax-Tmin)/2]. Another method to calculate average temperatures is to take the mean of all temperature readings for a given day. If a weather station reports hourly temperature readings, the mean of those tem¬perature readings would constitute the average of the 24 daily temperature readings, while if a station reports in 15-minute intervals, the average temperature would be the mean of the 96 daily temperature readings. The simple difference in methodology for calculating aver¬age temperatures can have an impact on GDD readings. A recent study of five weather sta¬tions in the Livermore Valley AVA, for example, found variations in 10-year average GDD as high as 206 based on the methodology uti¬lized. As many weather stations have software generating GDD figures for the user, the meth¬odology deployed is not always obvious. Fur¬ther, many software programs will vary on GDD calculations, including but not limited to growing season, threshold (e.g., 50° F or some other number), and how temperatures below that threshold are handled. Some weather station software programs, for example, will not include temperatures below 50° F in the average temperature calculation.

Daily accumulation methodologies, how¬ever, are not consistent with the methodology deployed by Winkler and Amerine. Rather, Winkler and Amerine used monthly means. Specifically, the mean monthly temperature above 50°F was multiplied by the number of
days in the month for each month from April to October, then summed for the entire grow¬ing season [ΣOct Apr monthly((T_mean-50)•30)].

Further, a 1998 assessment of temperatures in the city of Sonoma, Calif., determined that Amerine and Winkler’s original calculations may have been simplified to account for only 30 days in each month.7 Hence, GDD originally calculated in 1944 may have had four fewer days figured into the equation than what mod¬ern assessments typically apply.

The discrepancy between daily accumula¬tion of degree days and monthly accumulation of degree days is most pronounced in early-season and late-season numbers, when mean daily temperatures may be below 50° F. As a simple example, assume that the average tem¬perature for each day from April 1 to April 15 is 48° F and mean temperature for each day April 16 through April 30 is 54° F. Using the daily accumulation method, each day from April 1 to April 15 would have a degree-day total of 0, while each day from April 16 to April 30 would have a degree-day total of 4, for an April total of 60 degree days. With an average monthly temperature of 51° F, using the monthly average would yield a GDD total of 30 degree days for the month.

Even with consistency in methodology, other factors impact the overall GDD total, and hence the climatic region assigned to that total. Equipment and placement of weather stations are considered, although most government and research weather stations, and even most commercial weather stations in¬stalled in vineyards, are assumed to be placed to limit variables that may impact climatic data. The duration of the data (i.e., number of years of data), however, can have a significant impact on num¬bers. Typically, meteorological normals (i.e., what is considered to be the average weather for a particular location) are based on 30-year averages, updated every new decade (e.g., 2010-1981, 2000-1971, etc.). Finding com¬plete data sets covering 30 years in a given area can be difficult, so GDD averages used in viticulture are most commonly given in some significantly shorter duration.

Further complicating data sets is what happens around a given weather station during that 30- year period. Urban development can create warmer conditions that, combined with a changing climate, lead to migration from one climatic classification to an¬other over the course of historical data sets. Further, academic stud¬ies are often done in Celsius, with the Winkler climatic regions con¬verted to Celsius equivalents. For example, Region II is converted to 1,389-1,667° C (from 2,501- 3,000° F), giving the impression that the climate-region break¬downs are based on something more accurate than the conve¬nience of even 500 degree day intervals. (Since the original re¬gion breakdown is in Fahrenheit, and this article is a discussion of those regions, Fahrenheit is used throughout this article unless otherwise noted.)

To demonstrate how varied GDD numbers can be for a given location, data were assessed from a California Irrigation Manage¬ment Information System (CIMIS) station at Camino in El Dorado County, Calif. This station was chosen because it is one of the longest continuously operating CIMIS stations, having been in¬stalled Oct. 19, 1982. Data from this station are also nearly com-plete, especially after 1989. The station is not near a growing urban center. Camino is listed in “General Viticulture” (Region III with 3,400 GDD). This station is not the same weather station used in the Winkler calculations but offers some historical context for the Camino area in general. Thirty total years of data were assessed from 2017 to 1987, with 1988 removed from the analysis, as data were incomplete. The results in Figure 1 show that daily accu¬mulations led to higher GDD to¬tals than using monthly means.

Variations in averages
Both the daily accumulations and monthly accumulations to¬tals at the Camino station are relatively consistent with the total found in Winkler et al., although the numbers are 139 degree days different, based on methodology. If a grower had been given a single year’s data, however, and that year hap¬pened to be either the year with the lowest GDD (1998), or the year with the highest GDD (2017), very different conclu¬sions could have been reached about the climate of Camino.

The averages for the five most recent years show both a warmer climate than the 30-year average would suggest and greater vari¬ance between methodologies. Keeping with standards of clima¬tology and meteorology would discredit the five-year average as insufficient, as weather varies sea¬son to season, and climate is a long-term averaging of weather. Yet getting five years of data is often much more achievable than obtaining 30 years of data. The climate is also warming. Figure 1 demonstrates the long-term tem¬perature trend in GDD for Camino, with every academic indication that this trend will continue. The prospects are very likely that on¬going climatic conditions at Camino will more likely reflect GDD above the 30-year average. Which Winkler Index climate re¬gion is most appropriately applied to Camino? Any interested party could have a choice between Re¬gion III or Region IV, or if they were really looking to cherry-pick, Region II or Region V.

Comparative analysis is where a more appropriate usage of GDD may come into play. How does Camino compare to other nearby areas? A CIMIS station is also located at Diamond Springs, ap¬proximately nine miles southwest of Camino. Diamond Springs is a much newer station, installed Sept. 20, 2010, so a 30-year aver¬age is not possible. The five-year average at Diamond Springs based on daily accumulations is 3,997 degree days compared to 3,821 at Camino. The seven-year daily accumulation average (2011-2017) at Diamond Springs is 3,913 compared to 3,679 at Camino. Every year with the ex¬ception of 2017 shows higher GDD at Diamond Springs than at the Camino weather station (2017 shows Camino 30 degree days higher than Diamond Springs). Even if you could not give a good long-term average GDD for Diamond Springs, it would be safe to say that the cli¬mate at the Diamond Springs station is warmer than at the Camino station. As viticultural growing districts become more defined at subregional levels and as we test new environs for com¬mercial viticulture, local com¬parisons are valuable.

Duration of high temperatures
Another consideration with Win¬kler climatic regions is the dura¬tion of high and low temperatures. GDD calculations based on mean daily temperatures treat high and low temperatures equally, even if the high temperature is reached for only for a few minutes, but the low temperature persists for sev¬eral hours. Areas impacted by afternoon marine inversions may be especially susceptible to briefly maintained high temperatures. Figure 3, created for a study done in the Livermore Valley AVA, shows the cumulative number of hourly readings at or near the daily high tem¬peratures for several CIMIS weather stations in Northern California over a one-week period in July 2016. Cumulative high temperatures were maintained for a shorter period at Santa Rosa and Pleasanton, two areas known for variable summer marine intrusions, than at Point San Pedro, Brentwood and Modesto. Point San Pedro, at the immediate coast, has greater consistency in marine stratus layers than Santa Rosa and Pleasanton, which experi-ence fluctuations in coastal fog. Brentwood and Modesto, both inland from the coast, are known to be relatively free of coastal fog. Many areas transitioning from coastal to in¬land may show GDD numbers similarly influ¬enced by brief durations of high temperatures. Comparatively, similar GDD summations for areas at higher latitudes or greater elevations may show greater total hours near high tem¬peratures or longer periods of solar radiation. Using average air temperature rather than daily mean temperature may create more ac¬curate comparative numbers, assuming that the methodology is consistent across the areas being assessed.

Variations in growing season
A final consideration is the April 1 to Oct. 31 growing season used in GDD calculations. In areas of the West Coast, bud break typically occurs in March, sometimes as early as Febru¬ary, while harvest dates in August and Septem¬ber are more common than in October. Grape variety and microclimatic conditions in indi¬vidual vineyards also play a role.

A review of 50% bud break from 2013 to 2018 and harvest dates from 2013 to 2017 was conducted in seven vineyards in the Livermore Valley AVA, three planted to Chardonnay, three planted to Cabernet Sauvignon and one planted to Sauvignon Blanc. The average 50% bud break ranged from March 28 to April 5, depending on vineyard, for the three Cabernet Sauvignon vineyards, but the Chardonnay vineyards had average dates of March 14 to March 15. The Sauvignon Blanc vineyard ex¬perienced 50% bud break around March 26. Harvests for Cabernet Sauvignon approxi¬mated Oct. 8 to Oct. 21, depending on the vineyard. Chardonnay saw average harvest dates of Sept. 20 to Oct. 7. The vineyard with Sauvignon Blanc saw average harvest dates around Aug. 28. In many cases, bud break began prior to the April 1 growing season start, while harvest was most often completed prior to the Oct. 31 growing season end.

The rationale for calculating heat summa¬tions from April 1 through Oct. 31 are not clear, at least not in Amerine and Winkler’s best-known works on the subject. The 50° F thresh¬old is based on shoot growth, so April 1 may represent an approximate estimation of bud break, but harvest for many varieties occurred, even in the first half of the 20th century, prior to Oct. 31. Amerine and Winkler also calcu¬lated heat summations from bloom to harvest and noted the role of heat summations in the 30 days preceding harvest, therefore not limit¬ing their assessment to just April through Oc¬tober. Heat summations based on bloom and harvest are only possible when grape produc¬tion already exists within an area and would be difficult to use as a tool to match varieties to an area yet to see vines. Differential timing and lengths of growing seasons across North America further complicate the equation.

Is there a better index?
Other indexes also exist for assessing tempera¬ture. The Heliothermal Index, also known as the Huglin Index (HI), was introduced by Pierre Huglin in 1978. HI is the April 1 to Sept. 30 sum of the mean of the daily mean temper¬ate above 10° C and the high temperature above 10° C, multiplied by a coefficient indica¬tive of the latitude. Other indices include Aver¬age Growing Season Temperatures (GST) and Biologically Effective Degree Days (BEDD), along with several other formulas for calculat¬ing heat summations.

A 1999 study by Jorge Tonietto and Alain Carbonneau suggests that a multicriteria system of calculating HI, a dryness index (DI), and a cool night index (CI) are needed to account for potential water balance in soil and nocturnal temperatures in grape development.8 Gregory Jones and others used PRISM climate models to assess spatial climatic distribution across California, Idaho, Oregon and Washington based on the four most commonly used indices in viticulture.3,4 The results suggested that GST and GDD had the greatest correlations, but HI and BEDD demonstrated better differentiation of climate types. A similar study by Rosalyn Francine MacCracken and Paul R. Houser used the PRISM model to predict how well these four indices, along with a modified version of the GST accounting for the common length of a growing season (Modified-GSTavg or Mod- GSTavg), performed in U.S. states east of Cali-fornia, Oregon and Idaho.5 The MacCracken and Houser study found all common indices to be less than perfect at predicting the viability of successful viticulture when comparing the indi-ces’ results with existing viticulture. Mod- GSTavg showed more promise as a predictor of successful viticulture by taking into account the length of the growing season.

The Jones et al. study and the MacCracken and Houser study looked at viticulture on a super-regional scale, albeit using climatic models with 400m to 400m resolution. One could argue that the Winkler Index was never meant to be utilized on a super-regional scale, but rather as a statewide guide specific to California. The original 1944 paper looked only at California locations, with Winkler later introducing other U.S. and international loca¬tions into the scale for comparison. Winker et al. are clear that the regional scale was not meant to be definitive at a subregional scale: “These divisions into climatic regions should be considered as general demarcations.”

Winkler et. al. suggest more extensive work be conducted on more local areas and specifi¬cally note the work of Robert Sisson, who, dur¬ing his 35-year career as viticulture farm advisor to Sonoma County from the 1950s to the 1980s, created a climate classification system unique to Sonoma County. Sisson’s model broke the county into Marine, Coastal Cool and Coastal Warm climate types, based loosely on heat summations of hours spent between 70° F and 90° F. While much work has been done to delineate subre¬gions of viticultural areas, both formally through the creation of new, smaller AVAs and informally through the identification of subregional dis¬tricts by winegrowing associations, work to create climatic classification systems at more local levels unfortunately is scarce.

Use of the Regions I-V designations of the Winkler Index presents several limitations and errors. GDD calculations, on which the Winkler Index is based, can be a valuable tool for comparing growing areas, but GDD is just one aspect of climatic analysis that alone gives an incom¬plete picture (as are other indices designed to summarize heat summation). The methodology, daily climatic patterns, period assessed and length of growing season can all tell a different picture of mesoclimates that might otherwise appear to be similar by Winkler designation. Ideally, each viticultural region would find a climatic classification system that best suits the unique characteristics that define it, while re¬serving simpler indices like the Winkler Index for broader generalizations. Alternatively, GDD makes a good comparative tool when consistency in methodology and comparison are made within regional locations.

1. Amerine, M. and Winkler, A. (1944). Composition and quality of musts and wines of California grapes. Hilgardia ,15, 493–675.

2. Greenspan, Mark, “Row Direction—Which End is Up?” Wine Business Monthly, July 2008.

3. Jones, G., Duff, A. and Hall A., “Updated Analysis of Climate-Viticulture Structure and Suitability in the Western United States,” Proceedings of the 16th International GiESCO Symposium, Davis, Calif., July 2009.

4. Jones, Gregory V., Duff, Andrew A., Hall, Andrews and Myers, Joseph W., “Spatial Analysis of Climate in Winegrape Growing Regions in the Western United States,” American Journal of Enology and Viticulture, September 2010, 313-326.

5. MacCracken, Rosalyn Francine and Houser, Paul R., “Spatial Analysis of Climate-Viticulture Indices for the Eastern United States,” International Journal of Applied Geospatial Research, Volume 7, Issue 4, October-December 2016.

6. Shabram, Patrick L., “Mesoclimate Patterns of the Livermore Valley AVA,” report prepared for the Livermore Valley Winegrowers Associations (unpublished), October 2017.

7. Shabram, Patrick L., Redefining Appellation Boundaries in the Russian River Valley, California, thesis, San Jose State University, 1998.

8. Tonietto, Jorge and Carbonneau, Alain, “A Multicriteria Climatic Classification System for Grape-Growing Regions Worldwide,” Agriculture and Forest Meteorology, 2004.

9. Winkler, A.J., Cook, J.A., Kliewer, W.M., Lider, L.A., “General Viticulture,” University of California Press, 1962, 1974.

Print this page   PRINTER-FRIENDLY VERSION   »
E-mail this article   E-MAIL THIS ARTICLE   »
Currently no comments posted for this article.