Vol.

76: 203–223, 2018 CLIMATE RESEARCH

Published online October 25
https://doi.org/10.3354/cr01532 Clim Res

OPEN
ACCESS

Analysis of several bioclimatic indices for

viticultural zoning in the Pacific Northwest
Golnaz Badr1, 4,*, Gerrit Hoogenboom1, 5, Mohammad Abouali2, Michelle Moyer3,
Markus Keller3
1
AgWeatherNet Program, Washington State University, Prosser, Washington 99350, USA
2
Data Scientist at Ford Motor Company, Dearborn, Michigan 48120, USA
3
Viticulture and Enology, Department of Horticulture, Washington State University, Prosser, Washington 99350, USA
4
Present address: Cornell Lake Erie Research and Extension Lab, Portland, NY 14769, USA
5
Present address: Institute for Sustainable Food Systems, University of Florida, Gainesville, Florida 32611-0570, USA

ABSTRACT: The growth and development of grapevines Vitis vinifera L. are highly dependent on
the weather dynamics of a region. The goal of this study was to use long-term historical weather
data to obtain and develop several bioclimatic indices that are of viticultural importance. In this
study, several bioclimatic indices were computed using 30 yr (1983−2012) of daily weather data
that were obtained from the gridded surface meteorological dataset at the University of Idaho (UI
GSM). The bioclimatic indices were determined for each of the American viticultural areas (AVA)
located in the State of Washington and parts of northeast Oregon. Several new indices including
dynamic minimum temperature (DyMin.Temp.), cold damage index (CDI), and wind speed index
(WSI) were developed based on previous indices and models. Considering all AVAs, the mean
frost free days (FFD) ranged from 146 to 230; the mean growing degree days (GDD) ranged from
948 to 1662; the mean biologically effective degree days (BEDD) ranged from 318 to 1590; the
mean Huglin Index (HI) ranged from 1452 to 2425; the mean length of growing season (LGS)
ranged from 164 to 189 days; and the mean growing season suitability (GSS) ranged from 82% to
92%. Once these data were obtained, the bioclimatic indices were then used to categorize the
AVAs. These relative categories provide knowledge of climate potential in the study area for opti-
mal grape production, potentially enabling us to match the best cultivars to a site based on its spe-
cific climate dynamics.

KEY WORDS: American viticultural areas · AVAs · Cold damage index · Wind speed index ·
Dynamic minimum temperature index · GIS

1. INTRODUCTION perature and total precipitation during the growing

season for a specific region (Makra et al. 2009, Santos
Weather plays a primary role in grapevine (Vitis et al. 2013). Jones et al. (2004) described climate as
spp.) growth and development (Van Leeuwen et al. the most profound factor in determining the ability of
2004, Fraga et al. 2013), and specific environmental a region to produce quality grapes. The macro-
conditions determine grapevine physiology and de- climate requirements for vines stipulate that the low-
velopment (Jackson 2008, Magalhães 2008, Santos et est temperature during the winter does not drop be-
al. 2012, Tóth & Végvári 2016). A strong correlation low −15 to −20°C and that there is a minimum of 1000
has been reported between grape yield and climate growing degree days with a base temperature of
traits for a specific region, such as monthly mean tem- 10°C (Van Leeuwen & Seguin 2006). The meso-
© The authors 2018. Open Access under Creative Commons by
*Corresponding author: [email protected] Attribution Licence. Use, distribution and reproduction are un-
restricted. Authors and original publication must be credited.
Publisher: Inter-Research · www.int-res.com
204 Clim Res 76: 203–223, 2018

climate is mainly determined by the topography of a The HI and BEDD use a coefficient (k) to represent
region (Keller 2010), while the micro-climate is pri- the changes in day length imposed by latitude
marily decided by local soil and canopy management (Table S1). The increase in day length during the
and is focused on the fruit zone (White 2015). growing season increases potentially relative to an
increase in latitude. A clear definition of k does not
exist; however, Huglin (1978) categorized k based on
1.1. Bioclimatic indices the latitudes between 40° and 50° into 5 classes,
where k had a range of 1.02 to 1.06. The k for lati-
Several indices have been developed based on the tudes equal to or lower than 40° was reported to be 1
heat load (daily accumulated temperatures above a (Huglin 1978, Tonietto & Carbonneau 2004). The lat-
threshold of 10°C for a fixed period) and temperature itude temperature index (LTI) (Jackson & Cherry
requirements of grapevines. The bioclimatic indices 1988) is the result of multiplying the mean tempera-
were placed into 5 major categories: (1) bioclimatic in- ture of the warmest month by the subtraction of lati-
dices focusing on heat unit accumulation; (2) biocli- tude (ϕ) of the location from 60 (60 − ϕ).
matic indices focusing on length of growing season;
(3) bioclimatic indices focusing on minimum tempera-
ture; (4) bioclimatic indices focusing on precipitation; 1.1.2. Bioclimatic indices focusing on length of
and (5) bioclimatic indices focusing on wind speed. growing season

Growing season suitability (GSS) (Malheiro et al.

1.1.1. Bioclimatic indices focusing on heat unit 2010, Santos et al. 2012) is the fraction of days from
accumulation April to September with a daily mean air tempera-
ture above 10°C. The time frames are based on the
The Winkler index (WI) (Winkler et al. 1974, Jones standard formulation of the indices to be able to facil-
et al. 2010), more commonly recognized as growing itate the comparison among various regions. The
degree days (GDD), refers to the degree day units ac- length of growing season (LGS) (Jackson 2008) is the
cumulated during the growing season with a base number of days with mean temperatures above 10°C
temperature of 10°C. This required heat load (GDD) is for a growing season ranging from April 1 to October
a function of grapevine variety; therefore, the re- 31. The frost free days index (FFD) (Magarey et al.
ported minimum value and optimal range of the GDD 1998) is used to determine growing season length; it
can vary. In our study, GDD was computed based on a is the period between the last frost (temperatures
fixed growing season from April to October in order to below 0°C) in spring and the first frost in fall.
facilitate comparisons with other grape-producing re-
gions. The Huglin index (HI) (Huglin 1978) combines
the air temperature during the active period of vege- 1.1.3. Bioclimatic indices focusing on minimum
tative growth with a coefficient of day length that temperature
varies according to the latitude (see Table S1 in the
Supplement at www.int-res.com/articles/suppl/c076 The cool night index (CI) (Tonietto & Carbonneau
p203_supp.pdf). It provides valuable information on 2004) takes into account the minimum temperature
the local heat summations by considering the average during grape maturation, which is normally the aver-
and maximum temperatures and weighting the accu- age minimum air temperature in September. Mean
mulated temperatures to the daytime period. Huglin thermal amplitude (MTA) (Mullins et al. 1992, Ramos
& Schneider (1998) classified grape varieties based on et al. 2008) is the difference between the minimum
HI. Growing season temperature (GST) (Jones 2005) and maximum temperature in September and is as-
is the average temperature of the growing season sociated with grape quality and composition (Montes
from April to October. Biologically effective degree et al. 2012). Adjustments to MTA should be made
days (BEDD or E°) (Gladstones 1992) account for heat based upon phenology and not calendar. Tempera-
accumulations that are defined by upper and lower tures below −17°C are commonly considered the
temperature thresholds (between 10 and 19°C), and lethal lower temperature limit for grapes; defined as
the BEDD formulation also modifies the heat accumu- the minimum temperature index (Min.Temp.) (Hid-
lation for diurnal ranges (Table S1). Gladstones (1992) algo 2002), a vital constraint of growing grapevines.
applied BEDD to define corresponding maturity However, cold acclimation is a physiological process
groupings of various grape cultivars (Table 1). that happens over a period of time, so using a single
 Badr et al.: Bioclimatic indices for viticulture 205

Table 1. Wine grape maturity groupings and corresponding biologically effective degree days (BEDD) to ripeness (Gladstones
1992)

Red White or Rosé

Group 1
1050 day° – Madeline, Madeline-Sylvaner
Group 2
1100 day° Blue Portuguese Chasselas, Muller-Thurgau, Siegerrebe, Bacchus,
Pinot Gris, Muscat Ottonel, Red Veltliner, Pinot Noir,
Meunier
Group 3
1150 day° Pinot Noir, Meunier, Gamay, Dolcetto, Bastardo, Traminer, Sylvaner, Scheurebe, Elbling, Morio-Muskat,
Tinta Carvalha, Tinta Amarella Kerner, Green Veltliner, Chardonnay, Aligote, Melon,
Sauvignon, Blanc, Frontignac, Pedro Ximénez,
Verdelho, Sultana
Group 4
1200 day° Malbec, Durif, Zinfandel, Schiava (=Trollinger), Semillon, Muscadelle, Riesling, Welschriesling,
Tempranillo, Tinta Maderia, Pinotage Furmint, Leanyka, Harslevelu, Sercial, Malvasia
Bianica, Carbernet Franc
Group 5
1250 day° Merlot, Carbernet Franc, Shiraz, Cinsaut, Chenin Blanc, Folle Blanche, Crouchen, Rousanne,
Barbera, Sangiovese, Touriga Masanne, Viognier, Taminga, Carbernet Sauvignon
Group 6
1300 day° Cabernet Sauvignon, Ruby Cabernet, Colombard, Palomino, Dona Branca, Rabigato,
Mondeuse, Tannat, Kadarka, Corvina, Nebbi- Grenache
olo, Ramisco, Alvarelhao, Mourisco Tinto,
Valdiguie
Group 7
1350 day° Aramon, Petit Verdot, Mataro, Carignan, Muscat Gordo Blanco, Trebiano, Montils
Grenache, Freisa, Negrara, Grignolino, Souzao,
Graciano, Monastrell
Group 8
1400 day° Tarrango, Terret Noir Clairette, Grenache Blanc, Doradillo, Biancone

temperature threshold may not be effective in cap- fore the start of frost, crop load and plant nutrient sta-
turing grapevine behavior in response to low temper- tus have no significant effect on the cold hardiness of
ature. In reality, lethal low temperatures are dynamic grapes. (Davenport 2008). The temperature at which
and vary based on the cold hardiness status of each 50% of the organ or sample tissue is injured due to a
grapevine cultivator. Grapevine cold hardiness is a freeze is called LT50 (Ferguson et al. 2011), and the
dynamic phenomenon that responds to seasonal prediction of cold hardiness is primarily based on the
changes in air temperature (Ferguson et al. 2011). measured LT50 of the dormant buds of multiple grape
Cold hardiness starts with the acclimation at the end cultivars (Ferguson et al. 2011).
of the growing season (fall), which coincides with a
decrease in the air temperature. During the winter,
when temperatures are low and stable, cold hardi- 1.1.4. Bioclimatic indices focusing on precipitation
ness remains steady. De-acclimation begins when
the temperature rises at the end of winter/early The growing season precipitation index (GSP)
spring (Ferguson et al. 2011). The injuries caused by (Blanco-Ward et al. 2007) provides the general suit-
frost and freezes on grapes are well documented ability used in climate zoning for viticulture that
(Wample et al. 2001, Fennell 2004, Davenport et al. accumulates precipitation during the growing season
2008); however, there are differences among culti- (Table S1). However, the GSP is only relevant for
vars in terms of their cold hardiness (Clark et al. regions where grapes are normally not irrigated. The
1996, Wolfe 2001, Fennell 2004, Mills et al. 2006, hydrothermic index (HyI) (Branas 1974) combines
Davenport 2008, Ferguson et al. 2011, 2014). In re- the effect of air humidity and temperature using pre-
gions where plants complete their development be- cipitation as a surrogate during the growing season
206 Clim Res 76: 203–223, 2018

to assess the risk of grapevine exposure to certain Carbonneau (2004) undertook a climatic classifica-
diseases, such as downy mildew, as well as water tion that combined HI, CI, and dryness index (DI),
stress. and found that HI and CI were able to describe most
of the variability within the climatic zones of grape-
producing regions around the world. However, they
1.1.5. Bioclimatic index focusing on wind speed used World Meteorological Organization (WMO) sta-
tions that were located near each grape-producing
The impact of high wind speeds on grapevine region. This weather dataset was based on the point
growth and yield is well documented (Takahashi et measurements provided by the weather stations and,
al. 1976, Freeman et al. 1982, Hamilton 1988, Jack- therefore, did not spatially cover the study area.
son 2008, Keller 2010), but no comprehensive index Jones et al. (2009) also determined the climatic
has been developed using wind speed as a supple- indices for grape-growing regions worldwide using
mental tool for viticultural zoning. Previous studies several bioclimatic indices, including HI, CI, DI,
have indicated that strong winds can cause physical BEDD, and GST. They utilized WorldClim (World-
damage to grapevines (Hamilton 1988) as wind Clim 2009) 1 km resolution for the period from 1950
speed impacts evapotranspiration due to its impact to 2000 with a temporal resolution of monthly data.
on stomatal resistance (Dry et al. 1989, Campbell- Jones et al. (2009) recommended use of climate grids
Clause 1998, Tarara et al. 2005, Keller 2010) and with a finer spatial resolution and recent weather
boundary layer thickness (Keller 2010). Jackson data to update bioclimatic indices. With an improved
(2008) discussed the impact of wind velocity on the climate structure, the variability and change of suit-
heating of grape berries, canopy water deficiency, ability can be monitored more accurately and effi-
irrigation systems, soil erosion, disease dispersion, ciently (Jones et al. 2009). Santos et al. (2012) pro-
physical damage to vines, shoot length, leaf size, vided a macro-climate and classification analysis for
stomatal density, number of clusters per vine, ripen- grape-growing regions in Europe. They used the
ing, and the solid solubility. The impact of wind is ENSEMBLES observational gridded data set for Eu-
increased by the number of wind perturbations rope (E-OBS) from the EU-FP6 project ENSEMBLES
(Williams et al. 1994, Tarara et al. 2005, Keller 2010). (http://ensembles-eu.metoffice.com) for a period of
Gladstones (1992) reported 2 types of damage result- 59 yr (1950 to 2009) for the European continent and
ing from strong winds: (1) injury caused by strong calculated the GSS, GSP, Min.Temp., CI, WI, HI, DI,
winds in spring and early summer that mainly affects HyI, and composite index (CompI) in order to deter-
young and tender growth; and (2) injury caused by mine any potential trends in the individual indices.
hot, dry winds in summer that damages the vines, They identified significant trends in WI and HI, and
and leads to imperfect ripening and the collapse of using a canonical correlation analysis, demonstrated
berries. that the observed inter-annual variability of the HI
was strongly controlled by large-scale atmospheric
circulations during the growing season. Santos et al.
1.2. Bioclimatic zoning of grape-growing regions (2012) also analyzed the inter-annual variability and
long-term trends in the bioclimatic indices and up-
1.2.1. Climatic zones outside the USA dated the indices using high-resolution datasets for
Europe.
Several studies have used bioclimatic indices to Hall & Jones (2010) calculated GST, GDD, HI, and
classify potential grapevine-growing regions across BEDD for Australian wine grape growing regions
the globe. Jackson & Cherry (1988) calculated and from 1971 to 2000 and found that knowledge of cli-
compared 14 bioclimatic indices for 78 locations in mate dynamics helps to better understand the culti-
Europe, Australia, and New Zealand to determine var suitability within each region. Anderson et al.
the most useful index among those proposed for the (2012) used daily historical data from weather sta-
classification of grape-growing regions. They re- tions (1971 to 2000) to calculate bioclimatic indices
ported LTI and HI to be the best indices for differen- such as GST, WI, HI, and BEDD, and then interpo-
tiation of grape-growing regions, based on the in- lated the indices to a spatial grid with a 500 m resolu-
formation provided on the heat requirements of tion. They found that elevation plays an important
different cultivars. The climate dataset was obtained role in determining the climate suitability of a partic-
via weather stations, and, therefore, the temporal ular region for grape production. They argued that
and spatial resolutions were not optimal. Tonietto & GST and WI functionally capture the same informa-
 Badr et al.: Bioclimatic indices for viticulture 207

tion, and that HI is most capable of representing the using the PRISM monthly dataset for the Pacific
actual structure of suitable regions for grape produc- Northwest (PNW) region of the USA. Yau et al. (2013)
tion in New Zealand. However, they suggested the used principal component analysis (PCA) to deter-
need for within-region assessment of potentially suit- mine the dominant factors influencing the AVAs in
able areas in future studies. Montes et al. (2012) used the PNW region, and found that the combination of
CI, DI, HI, and MTA for a multi-criteria climate clas- elevation, GDD, FFD, and precipitation was the most
sification of 7 Chilean viticultural valleys represented important. They concluded that the climate compo-
by 54 different weather stations. They reported that nent of AVAs is the most difficult to obtain and ana-
CI did not accurately represent climate variation in lyze, implying that any research capable of providing
Chile and lacked discriminating capacity. Montes et an improved climate component would enable the
al. (2012) found a similar spatial trend for MTA and advancement of characterization of AVAs in the
HI, and concluded that MTA was a suitable index for PNW region. Badr et al. (2018) used several biocli-
characterizing the thermal regime in Chile. Con- matic indices coupled with soil and topography infor-
ceição & Tonietto (2005) evaluated the climate poten- mation to delineate the suitable areas for growing
tial of 3 regions in Brazil by calculating HI, CI, and wine grapes in the State of Washington.
DI using average historical meteorological data ob- The objectives of our study were to conduct statis-
tained from weather stations in those areas. tical analysis of several bioclimatic indices for the
AVAs located in the State of Washington and parts of
northeastern Oregon, and to develop new indices
1.2.2. Climatic zones in the USA based on the cold hardiness dynamics of grapes, min-
imum temperature, and wind speed thresholds.
The most suitable climatic zones for viticulture in
California were first classified by Amerine & Winkler
(1944) and Winkler et al. (1974). They were later up- 2. MATERIALS AND METHODS
dated for the western USA by Jones et al. (2010)
using PRISM (parameter-elevation relationships on 2.1. American Viticultural Areas (AVAs)
independent slopes model) (Daly et al. 2008) for the
period from 1971 to 2000 with a spatial resolution of Officially recognized appellations that allow vint-
15 arc-seconds (400 m) and a monthly temporal reso- ners and consumers to attribute wine characteristics
lution. Jones et al. (2010) calculated 4 bioclimatic to the specific geographic origin of its grapes are
indices, including GDD, HI, BEDD, and GST, and called AVAs (TTB 2015). AVAs are acknowledged by
emphasized the importance of updating the long- the TTB, and the US Department of the Treasury
term weather data for the calculation of bioclimatic allows vintners to describe the origin of their wine to
indices. Yau (2011) computed GDD, LTI, and FFD consumers (TTB 2015). In the PNW region of the

Table 2. Area, elevation and year of recognition and other properties of American viticultural areas (AVAs) in the study area

AVA State(s) Area Elevation (m) AVA

(ha) Min. Max. Mean recognition
year

Columbia Valley Washington-Oregon 4 597 090 20 1442 451 1984

Puget Sound Washington 1 179 606 0 1046 109 1995
Yakima Valley Washington 289 775 128 1099 339 1983
Horse Heaven Hills Washington 233 160 81 671 321 2005
Walla Walla Washington-Oregon 129 059 122 696 315 1984
Ancient Lakes of the Columbia Valley Washington 68 607 173 583 377 2012
Columbia Gorge Washington-Oregon 48 431 23 840 377 2004
Wahluke Slope Washington 32 631 122 504 251 2005
Rattlesnake Hills Washington 29 934 258 922 434 2006
Lake Chelan Washington 13 291 278 1152 445 2009
Naches Heights Washington 5315 359 647 543 2012
Red Mountain Washington 1837 167 429 228 2001
Snipes Mountain Washington 1585 224 399 280 2009
The Rocks District of Milton-Freewater Oregon 1483 241 307 278 2015
208 Clim Res 76: 203–223, 2018

Fig. 1. Geographic locations of the Washington-Oregon American viticultural areas (AVAs)

USA, a total of 14 AVAs have been established in the the USDA published statistics for 2012, and it is ex-
State of Washington and eastern and north-central panding annually (USDA-NASS 2015).
Oregon (Table 2, Fig. 1). The Columbia Valley is the
largest, covering an area of 4 597 090 ha, while the
Yakima Valley was the first AVA to be federally rec- 2.2. Weather data
ognized in Washington in 1983 (Washington State
Wines [WSW] 2015). The most recent AVA to be fed- We used weather data obtained from the University
erally recognized was Oregon’s Rocks District of Mil- of Idaho’s gridded surface meteorological data (UI
ton-Freewater in 2015, which is enclosed by the GSM 2015). The UI GSM (Abatzoglou 2011) employs
Walla Walla AVA. the spatial attributes of PRISM (Daly et al. 2008), with
Several of the AVAs in the PNW are shared among temporal attributes of regional-scale reanalysis and
Washington and eastern and north-central Oregon daily gauge-based precipitation from the North
(Fig. 1). The study area was, therefore, expanded to American land data assimilation system phase 2
include parts of Oregon in order to calculate the bio- (NLDAS-2; Mitchell et al. 2004). The dataset has a
climatic indices for all the shared AVAs. The latitude spatial resolution of 4 km for the contiguous United
ranged from 45.25° to 49° N and the longitude ranged States and a daily temporal resolution. The UI GSM
from 116.8° to 124.78° W. The total area under grape was evaluated by Abatzoglou (2013) through com-
cultivation in Washington was 27 186 ha, based on paring the dataset with the observed data recorded
 Badr et al.: Bioclimatic indices for viticulture 209

by weather station networks such as RAWS (RAWS 2.3. CDI

2011), AgriMet (AgriMet 2011), AgWeatherNet (Ag-
WeatherNet 2011), and USHCN-2 (USHCN 2011). Acclimation of grapevines to cold weather in fall
The daily minimum and maximum air temperatures for regions located in the northern hemisphere is
of the dataset have a reported median error of 1.7 to called cold hardiness and it varies greatly based on
2.3°C, and the median correlations of the daily maxi- variety and region. The cold hardiness (Hc,i) for day i
mum and minimum air temperatures are reported to is computed based on the cold hardiness of the previ-
be 0.94 to 0.95 and 0.87 to 0.90, respectively (Abat- ous day (Hc,i–1) plus the changes in the cold hardiness
zoglou 2013). For the wind speed, median correlations (ΔHc) over the course of a single day (Eq. 1). The cold
of 0.54 and 0.68 were reported during the cold season hardiness was computed for a hypothetical cultivar to
(October to April) and correlations of 0.52 and 0.62 for ensure that the most sensitive cultivars were taken in
the warm season (May to September), revealing an to account.
overestimation ranging from 5% to 30% (Abatzoglou
Hc,i = Hc,i–1 + ΔHc (1)
2013). Still, the dataset has great potential for land-
scape-scale modeling in areas where there is a limited The changes in cold hardiness (ΔHc) were computed
amount of comprehensive, long-term, daily historical as follows:
weather data (Abatzoglou 2013).
ΔHc = (DDc + ka × c log,a) + (DDh + kd × c log,d) (2)
In our study, the bioclimatic indices were calcu-
lated using several variables obtained from the UI where ka is the constant for acclimation rate, kd is the
GSM, which covered a period of 30 yr (1983 to 2012). constant for de-acclimation rate, c log,a is the dimen-
The variables included precipitation, temperature, sionless logistic component for acclimation, c log,d is the
and wind velocity at 10 m above ground. The initial de-acclimation logistic component, DDc represents
resolution was 4 km, and was downscaled to a spatial the chilling degree days, and DDh the heating degree
resolution of 482 m using a bi-linear interpolation days (Ferguson et al. 2011, 2014). The initial cold har-
algorithm (MATLAB V2014a 2014); the indices were diness (Hc,initial) was modified to −3°C, which is the re-
developed for the period between 1983 and 2012. For ported hardiness of green tissue (Fennell 2004).
our study, the day length/latitude coefficient (k) The CDI was developed to evaluate the minimum
required for obtaining HI and BEDD was calculated daily temperature in a specific location with respect
by applying a linear interpolation to the previously to the predicted LT50 for that location. The index uti-
reported k for each region. This interpolation modi- lizes a predicted LT50 to count the number of events
fied k based on linear changes in latitude. The k was that occur when the daily minimum temperature falls
linearly increased from 1.03 to 1.06 by increasing the below the LT50 threshold. Hence, the CDI compares
latitude from 44° to 50°. the minimum daily temperature with the predicted
The bioclimatic indices for each AVA were ex- LT50 for a given day. If the minimum temperature
tracted using ArcGIS v.10 (ESRI 2015) based on the falls below the threshold, it is counted as 1 event.
AVA boundaries. All the shapefiles, except for the Consequently, the number of events can be counted
Rocks District of Milton-Freewater, were obtained for each month, and the total number of incidents per
courtesy of the Davenport Laboratory, Washington year can then be summed up to calculate the annual
State University Irrigated Agriculture and Research CDI. Depending on the availability of long-term
and Extension Center, Prosser, WA, USA. The com- weather data, CDI can be computed for several
puted bioclimatic indices were statistically compared years. It is also possible to sum up the CDI for periods
among AVAs to detect significant differences at the of 5, 10, 20, and 30 yr, and then discuss the probabil-
95% confidence level. The statistical analysis was ity of the occurrence of critical low temperatures in a
conducted to compare the mean values using the region based on the results. In fact, CDI could be
ANOVA test with SAS v.9.4. at a 95% level (SAS computed for 30 yr and then used to calculate the
2015). It should be noted that the AVAs differed in CDI for 100 yr.
area and number of pixels, and were also not inde-
pendent, as Columbia Valley encompasses most of
the AVAs located in eastern Washington (Fig. 1). 2.4. DyMin.Temp.
However, use of statistical techniques can assist with
further categorizing these AVAs to obtain more pro- The DyMin.Temp. was created as a modification of
found insight into how each AVA compares with the the Min.Temp. (Table S1). By comparing the daily
others. minimum temperatures with the dynamic low thresh-
210 Clim Res 76: 203–223, 2018

old for each month, grapevines’ various levels of sen- velocity (⎯Uy,m,d). WSI essentially uses the same proce-
sitivity to low temperatures during both the growing dure as described for DyMin.Temp., but uses ew,y,m
season and winter can be monitored for each loca- instead of eT,y,m (Eq. 7).
tion. The DyMin.Temp. compares daily low air tem-
ew ,y ,m = ∑ d =1 condw (U y ,m,d ,U th )
eom(y ,m )
peratures with the assigned threshold for a particular (7)
month (Tt); consequently, it calculates the number of
events (d) when the temperature fell below a certain where ⎯Uy,m,d is the average wind velocity for the
threshold during a particular month. The total num- specified day, Uth is the wind velocity threshold, and
ber of incidents occurring in a single month ranges the condw (⎯Uy,m,d,Uth) is obtained as follows (Eq. 8):
from 0 (‘no events’) to the number of days in a partic- ⎧⎪ 1 U
y ,m,d > U th
ular month, ultimately leading to the determination condw (U y ,m,d ,U th,m ) = ⎨ (8)
of the total number of incidents occurring within a ⎩⎪ 0 otherwise
specific month. Depending on the availability of a The ⎯Uth used in our study was 4 m s−1 (Freeman et
given region’s long-term weather data, the total al. 1982, Hamilton 1988, Hunter & Bonnardot 2011).
number of events recorded for each month must then Since the daily wind velocities obtained from the UI
be averaged for each year of the study. The GSM were based on 10 m wind speeds, the logarith-
DyMin.Temp. was ultimately obtained by summing mic wind profile conversion equation was used to
up the average number of events that occurred convert the threshold to a 2 m wind velocity (Allen et
within each individual month for a specific 12 mo al. 1998).
period (Eq. 3).

TDynamic min = ∑ m = Jan e m

Dec
(3) 2.6. Precipitation

where TDynamic min is the dynamic minimum tempera- Spatio-temporal analysis of the precipitation used
ture and⎯em is the average number of events in month in our study was divided into 2 groups: (1) precipi-
m (Eq. 4). tation during the growing season (GSP) (Blanco-
Ward et al. 2007); and (2) out-of-growing-season
1 2012
em = ∑ eT ,y ,m
n 1983
(4) precipitation (OutGSP). The motivation for dividing
precipitation into 2 distinct groups was to obtain a
where n is the number of years and eT,y,m is the num- better understanding of local water availability
ber of events occurring within a specified year and prior to the start of a new growing season. This
month, as denoted by y and m, respectively (Eq. 5). knowledge can help decision-makers, extension
specialists, and growers gain more insight into the
eT ,y ,m = ∑ d =1 condT (Tmin,y ,m,d ,Tth,m )
eom(y ,m )
(5) need for irrigation of grapes in specific regions. For
each AVA, the total precipitation was calculated
where eom(y,m) (end of month) returns the number and the percentage of precipitation was also ob-
of days in a given month and year, represented by m tained. OutGSP is a particularly useful index as it
and y, Tmin,y,m,d is the minimum temperature for a can potentially indicate the amount of water avail-
specific day, Tth,m is the temperature threshold for a able for filling the soil profile prior to the start of a
chosen month (the thresholds were as follows: Janu- new growing season.
ary and December = −10°C; October, November, and
February = −5°C, March = −3°C, and April to Sep-
tember = 0°C), and condT (Tmin,y,m,d,Tth,m) is defined as 3. RESULTS AND DISCUSSION
follows (Eq. 6):
⎧⎪ 1 Tmin,y ,m,d < Tth,m 3.1. Bioclimatic indices focusing on thermal heat
condT (Tmin,y ,m,d ,Tth,m ) = ⎨ (6) unit accumulation
⎩⎪ 0 otherwise
3.1.1. BEDD
2.5. Wind speed index (WSI)
BEDD were computed for each AVA. Puget Sound
The WSI was developed to help address the impact had the lowest BEDD (318°C BEDD), while the
of wind speed on viticultural zoning. The WSI ini- highest was obtained for Snipes Mountain (1590°C
tially takes into account the 10 m daily average wind BEDD) (Table 3, Table S2 in the Supplement). Jones
 Badr et al.: Bioclimatic indices for viticulture 211

Table 3. The average biologically effective degree days (BEDD) calculated for each their GDD, the AVAs of this
American viticultural areas (AVAs) in the study area. The classification only pertains study were primarily catego-
to this index (based on Jones et al. 2010). (–) Not applicable
rized (Jackson, 2008) into re-
gion I (≤1390) and region II
AVA BEDD Classification Maturity group (1391 to 1670). Jones et al.
(°C) (Gladstones 1992)
(2010) also reported that most
Puget Sound 318 Too cold – of the regions in Washington
Lake Chelan 838 Too cold – are categorized as region I,
Columbia Gorge 941 Too cold – while the Columbia Valley
The Rocks District of Milton-Freewater 969 Too cold – (Oregon) is categorized as re-
Ancient Lakes of the Columbia Valley 976 Too cold – gion II (Fig. 2b). The higher the
Columbia Valley 1086 1 Group 1
Walla Walla 1108 1 Group 2
heat unit accumulation in a re-
Wahluke Slope 1129 1 Group 2 gion, the greater the ability to
Horse Heaven Hills 1160 1 Group 3 ripen grapes, especially culti-
Red Mountain 1208 2 Group 4 vars that mature late in the
Naches Heights 1221 2 Group 4 growing season (Wolfe 1999).
Rattlesnake Hills 1286 2 Group 5
Our results also confirm previ-
Yakima Valley 1339 2 Group 6
Snipes Mountain 1590 3 Group 8 ous reports on the western
USA. (Jones et al. 2010).

et al. (2010) recommended a classification of grape- 3.1.3. HI

growing regions based on BEDD and suggested that
if the BEDD is below 1000°C then the region is too The HI calculated for our study indicated that,
cold for grapes; however, 5 AVAs within our study among the chosen AVAs, Puget Sound had the low-
area had BEDDs lower than 1000°C, including Puget est (1452), while Snipes Mountain had the highest
Sound, Lake Chelan, Columbia Gorge, the Rocks (2425) (Table 5). The different values for HI indicate
District of Milton-Freewater, and Ancient Lakes of that there was a significant difference at the 95%
the Columbia Valley. The rest of the classes start at confidence interval for some AVAs. However, no
1000°C BEDD, with a range of 200°C BEDD for each such differences were found for the following pairs:
class (Table 3). Gladstones (1992) reported the wine Ancient Lakes of the Columbia Valley and Yakima
grape maturity groupings and their corresponding Valley, Horse Heaven Hills and Walla Walla, Colum-
BEDD to ripeness for making dry or semi-sweet table bia Valley and Rattlesnake Hills (Table S2). We next
wines (Table 1). Among our chosen AVAs, only categorized the AVAs based on the HI (Jones et al.
Snipes Mountain matched Group 8, for which culti- 2010) (Table 5), with the accepted classification stat-
vars such as ‘Tarrango’, ‘Terret Noir’, ‘Clairette’, ing that areas with an HI below 1200 are too cold for
‘Grenache Blanc’, ‘Doradillo’, ‘Biancone’ are recom- grapes. The classes start at 1200, each comprising
mended (Fig. 2a, Table 3). 300 units. The first HI class (1200 to 1500) is ‘very
cool’, the second is ‘cool’ (1500 to 1800), the third is
‘temperate’ (1800 to 2100), the fourth is ‘warm tem-
3.1.2. GDD perate’ (2100 to 2400), and the fifth is ‘warm’ (2400 to
2700). Our results indicated that the majority of AVAs
The calculated GDD indicated that Puget Sound in the study area are located in a ‘warm temperate’
had the lowest accumulated thermal units (948°C region (class 4) based on their HI (Fig. 2c).
GDD), while Wahluke Slope had the highest (1662°C
GDD). The multiple comparisons of GDDs among
AVAs indicated that there was no significant differ- 3.1.4. LTI
ence in the GDD with a 95% confidence level be-
tween Wahluke Slope and the Rocks District of The LTI was computed for all the AVAs; Puget
Milton-Freewater. There were also no substantial dif- Sound had the lowest (223.2) while the Rocks Dis-
ferences between Red Mountain and Snipes Moun- trict of Milton-Freewater had the highest (336.9)
tain, or Horse Heaven Hills and Walla Walla, with a (Table 6). Because this index also considers the lati-
95% confidence level (Table 4, Table S2). Based on tude of each location, the LTI was significantly differ-
212 Clim Res 76: 203–223, 2018

Fig. 2. Key bioclimatic indices primarily dealing with thermal heat unit accumulation for American viticulture areas (AVAs) lo-
cated in the State of Washington and parts of Oregon. Class limits in legends are not directly comparable. (a) Biologically ef-
fective growing degree days (BEDD); (b) growing degree days (GDD); (c) Huglin index (HI); (d) latitude temperature index
(LTI); (e) growing season temperature (GST). See Fig. 1 for coordinates
 Badr et al.: Bioclimatic indices for viticulture 213

Table 4. Mean growing degree days (GDD) calculated for Table 6. Calculated latitude temperature index (LTI) and
each American viticultural area (AVA) based on daily corresponding LTI groups for American viticultural areas
weather data from 1983 to 2012. The classification only (AVAs) in the study area using daily weather data from 1983
pertains to this index to 2012. Grouping based on AVA LTIs was also conducted,
based on the grouping recommendations from Jackson &
Cherry (1988). LTI grouping does not correspond to group-
AVA Mean Region
ings in other indices. Examples of cultivars that match the
GDD
groupings are provided
Puget Sound 948 I
Columbia Gorge 1089 I AVA LTI LTI Examples of
Naches Heights 1189 I group cultivars
Lake Chelan 1329 I
Rattlesnake Hills 1386 I Puget Sound 223.2 B Riesling, Pinot Noir
Columbia Valley 1414 II and Chardonnay
Yakima Valley 1488 II
Lake Chelan 269.3 B
Ancient Lakes 1525 II
Walla Walla 1566 II Naches Heights 281.3 C Cabernet Sauvi-
Horse Heaven Hills 1566 II gnon, Cabernet
Red Mountain 1622 II Franc, Merlot,
Snipes Mountain 1624 II Malbec, Sauvignon
The Rocks District of Milton-Freewater 1649 II Blanc, and Semillion
Wahluke Slope 1662 II Columbia Gorge 285.3 C
Ancient Lakes 299.3 C
Table 5. The calculated Huglin index (HI) for American viti- Rattlesnake Hills 301.8 C
cultural areas (AVAs) of the study area. The classification Columbia Valley 304.4 C
only pertains to this index Yakima Valley 312.0 C
Wahluke Slope 319.4 C
AVA HI Classification Snipes Mountain 322.3 C
Red Mountain 324.8 C
Puget Sound 1452 Very cool
Columbia Gorge 1740 Cool Horse Heaven Hills 328.3 C
Naches Heights 1931 Temperate Walla Walla 328.6 C
Lake Chelan 2019 Temperate The Rocks District 336.9 C
Columbia Valley 2124 Warm temperate of Milton-Freewater
Rattlesnake Hills 2138 Warm temperate
Yakima Valley 2248 Warm temperate
Ancient Lakes of the 2248 Warm temperate
Columbia Valley
ranging from 200 to 275; (3) group C with an LTI
Walla Walla 2274 Warm temperate
Horse Heaven Hills 2289 Warm temperate ranging from 275 to 370; and (4) group D with an LTI
The Rocks District of 2332 Warm temperate > 370. The results of our study indicated that the
Milton-Freewater majority of AVAs in Washington and Oregon are
Red Mountain 2362 Warm temperate within group C, except for Puget Sound and Lake
Wahluke Slope 2397 Warm temperate
Snipes Mountain 2425 Warm Chelan (Table 6). Jackson & Cherry (1988) reported
the favored cultivars for group B to be ‘Riesling’,
‘Pinot Noir’, and ‘Chardonnay’, and the favored culti-
ent (Fig. 2d, Table S2) for all of the AVAs except for vars for group C are ‘Cabernet Sauvignon’, ‘Caber-
Walla Walla and Horse Heaven Hills because that net Franc’, ‘Merlot’, ‘Malbec’, ‘Sauvignon Blanc’, and
they are located within the same latitude range (45.6° ‘Semillion’. This classification can thus be used as a
to 46.2° N). tool for assisting in the selection of cultivars based on
Yau (2011) also calculated the LTI for the AVAs in their LTI groupings.
this study area, and a comparison between the 2 sets
of results showed that the AVA rankings were almost
identical for both studies, although the LTIs calcu- 3.1.5. GST
lated for our study were generally slightly higher
than those reported by Yau (2011). Jackson & Cherry The GST for all the AVAs was greater than 13.7°C.
(1988) reported an LTI grouping based on the culti- Puget Sound had the lowest GST (13.7°C) while the
vars grown in each group. The 4 main groups were: Rocks District of Milton-Freewater had the highest
(1) group A with an LTI < 196; (2) group B with an LTI (16.9°C) (Table 7; Table S2). Comparison of the mean
214 Clim Res 76: 203–223, 2018

Table 7. Mean growing season temperature (GST) calculated for each American viticulture area (AVA) in the study area from
1983 to 2012, and some examples of possible grape cultivars that match the climate maturity groupings

AVA Mean Climate Examples recommended of grape cultivars

GST maturity
(°C) grouping

Puget Sound 13.7 Cool (1) Riesling, Muller-Thurgau, Pinot Gris, Gewürztraminer,
Pinot Noir, Chardonnay, Sauvignon Blanc
Columbia Gorge 14.0 Cool (1)
Naches Heights 14.4 Cool (1)
Lake Chelan 15.1 Intermediate (2) Riesling, Pinot Gris, Gewürztraminer, Pinot Noir,
Chardonnay, Sauvignon Blanc, Semillon, Cabernet
Franc, Tempranillo, Merlot, Malbec, Syrah, Viognier,
Dolcetto,
Cabernet Sauvignon
Rattlesnake Hills 15.5 Intermediate (2)
Columbia Valley 15.6 Intermediate (2)
Yakima Valley 16.0 Intermediate (2)
Ancient Lakes of the Columbia Valley 16.2 Intermediate (2)
Horse Heaven Hills 16.4 Intermediate (2)
Walla Walla 16.5 Intermediate (2)
Red Mountain 16.7 Intermediate (2)
Snipes Mountain 16.7 Intermediate (2)
Wahluke Slope 16.9 Intermediate (2)
The Rocks District of Milton-Freewater 16.9 Intermediate (2)

GST for each AVA revealed that there was a substan- 3.2. Bioclimatic indices focusing on length of
tial difference in GST among the AVAs, except for growing season
Red Mountain and Snipes Mountain, as well as the
Rocks District of Milton-Freewater and Wahluke 3.2.1. FFD
Slope, which only had small differences at the 95%
confidence level. The calculation of FFD for each AVA showed that
Additional classifications have been made based Naches Heights had the lowest number of FFD
on the average GST as a means of determining grape (146 d) and Puget Sound had the highest (230 d). Red
maturity groupings and cultivars (Jones 2007, Jones Mountain and Wahluke Slope did not show a signifi-
et al. 2010, Yau 2011). These groupings start at aver- cant difference in FFD at the 95% confidence level
age GSTs of 13°C and end at 24°C. The 4 major matu- (Table S2), implying that the average growing season
rity groups within this range are: (1) the cool climate length for these regions is similar. This information
maturity group ranging from 13 to 15°C; (2) the inter- can help decision-makers select cultivars for each
mediate climate maturity group ranging from 15 to specific AVA that will complete their growing cycles
17°C; (3) the warm climate maturity group ranging prior to the first freeze in fall, and start their activity
from 17 to 19°C; and (4) the hot climate maturity after the last frost in the spring.
group ranging from 19 to 24°C. Based on the calcu- Previous studies have indicated that a region re-
lated GST, each AVA in our study can be categorized quires a minimum FFD of 180 d to be considered opti-
as belonging in either climate maturity group 1 or 2 mal for grape production (Prescott 1965, Rosenberg et
(Fig. 2e, Table 7). There are also several cultivars al. 1983, Becker 1985, Jackson & Cherry 1988, Yau et
associated with these climate maturity groupings al. 2013). Among our selected AVAs, Wahluke Slope,
(Jones 2007, Yau 2011). The application of GST can Walla Walla, the Rocks District of Milton-Freewater,
help to distinguish the climate maturity groupings for and Puget Sound all had FFD that were equal to or
AVAs, thereby enabling better cultivar recommenda- greater than 180 d. Knowledge of the ranking of AVAs
tions for each specific climate maturity grouping based on their FFD is important for the assignment of
(Fig. 2e, Table 7). early or late cultivars in these regions. The remaining
 Badr et al.: Bioclimatic indices for viticulture 215

AVAs with an FFD of less than 180 d have already period from April to September the air temperature is
been established, and they are still able to produce higher than 10°C. Santos et al. (2012) reported that a
marketable grapes. However, this may be due in part GSS higher than 90% is required for a region to be
to interventions such as wind machines, and other considered suitable for grape production, in order to
frost protection measures. Still, the reported optimal satisfy the heat requirement of grapevines. Although,
growing season length for regions with a proven the GSS calculated for half of the AVAs in our study
record of sustainable grape production should be up- was lower than 90%, they are still suitable for viticul-
dated based on the chances that the growers are will- ture because they have a GSS ranging from 80% to
ing to take in a region that has a higher risk of damage 90%. Regions with a lower GSS (Table 8) in Washing-
due to low temperature (Fig. 3a). ton and northeastern Oregon are similar to viticultural
regions in Western Europe such as Burgundy, Cham-
pagne, and the Mosel and Rhine Valleys of Germany
3.2.2. GSS (Malheiro et al. 2012, Santos et al. 2012).

GSS was also calculated for the AVAs in our study.

The Columbia Gorge had the lowest GSS (0.82), while 3.2.3. LGS
the Rocks District of Milton-Freewater had the
highest (0.92) (Fig. 3b). All AVAs had a GSS higher The calculated LGS ranged from 164 to 189 d.
than 0.8 (Table 8), indicating that for 80% of the Naches Heights had the lowest LGS (164 d) while the

Fig. 3. Key bioclimatic indices focusing on the length of grow-

ing season for American viticulture area (AVAs) located in the
State Washington and parts of Oregon. Class limits are not di-
rectly comparable. (a) Frost free days (FFD); (b) growing sea-
son suitability (GSS); (c) length of growing season (LGS).
See Fig. 1 for coordinates
216 Clim Res 76: 203–223, 2018

Table 8. The range of growing season suitability (GSS) cal- Table 9. Calculated length of growing season (LGS) for all
culated and averaged over a period of 30 yr based on daily the American viticulture areas (AVAs) in the study area and
weather data from 1983 to 2012. The classification their corresponding frost free days (FFD). Adjusted LGS
only pertains to this index based on the lowest value of either LGS or FFD

AVA GSS Preference AVA LGS FFD Adjusted

based on GSS LGS

Columbia Gorge 0.82 Suitable Naches Heights 164 146 146

Naches Heights 0.83 Suitable Columbia Gorge 165 164 164
Lake Chelan 0.86 Suitable Lake Chelan 169 166 166
Puget Sound 0.86 Suitable Rattlesnake Hills 174 157 157
Rattlesnake Hills 0.87 Suitable Columbia Valley 175 168 168
Columbia Valley 0.87 Suitable Puget Sound 176 230 176
Yakima Valley 0.89 Suitable Ancient Lakes of 179 171 171
Ancient Lakes of 0.90 Most suitable the Columbia Valley
the Columbia Valley Yakima Valley 179 164 164
Horse Heaven Hills 0.90 Most suitable Horse Heaven Hills 183 178 178
Walla Walla 0.91 Most suitable Walla Walla 184 189 184
Wahluke Slope 0.91 Most suitable Wahluke Slope 185 180 180
Snipes Mountain 0.92 Most suitable Snipes Mountain 186 166 166
Red Mountain 0.92 Most suitable Red Mountain 186 179 179
The Rocks District of 0.92 Most suitable The Rocks District of 189 203 189
Milton-Freewater Milton-Freewater

Rocks District of Milton-Freewater had the highest with regions that have shorter growing seasons,
(189 d). Warmer regions located at lower elevations thereby providing useful information to assist deci-
tended to have a higher LGS compared with colder sion-makers in cultivar selection.
regions at higher elevations (Fig. 3c). However, since
this index essentially uses concepts similar to FFD,
we evaluated AVAs by checking both indices simul- 3.3. Bioclimatic indices focusing on minimum
taneously. In comparisons of all AVAs, some had sim- temperature
ilar LGS and FFD (Table 9) while others showed sub-
stantial differences. This resulted in 2 conditions. 3.3.1. CI
When FFD is larger than LGS, the risk of frost is low
and there is a long period between the last frost in Our calculation of the CI indicated that Naches
spring and the first frost in fall. However, if the Heights had the lowest average (6.8°C) while the
selected cultivars in regions with these types of con- Rocks District of Milton-Freewater had the highest
ditions are not able to complete their growth and (10.2°C). Comparison of the CI among all the AVAs
development cycle within the LGS, they may not showed a significant difference among most of the
mature properly and the grapes might not be of opti- AVAs at a 95% confidence level, except for AVA
mal quality. When FFD is less than LGS, which is the pairs Ancient Lakes and Puget Sound, Horse Heaven
dominant case for the majority of AVAs in our study Hills and Red Mountain, Columbia Valley and Snipes
(Table 9), the risk of frost during the growing season Mountain, and Yakima Valley and Lake Chelan
is higher, and because of this, it is better to adjust the (Table S2, Fig. 4a.). Knowledge of regions with simi-
LGS to be equal to the FFD. lar CIs can aid growers and decision-makers in a
All of the AVAs in our study match the second more accurate allocation of cultivars to specific
condition, except for Puget Sound, Walla Walla, and regions, thus improving the potential quality of the
the Rocks District of Milton-Freewater (Fig. 3c). In grapes that are being produced. Based on the CI
regions with shorter growing seasons, grapes may be classifications discussed by Tonietto & Carbonneau
unable to complete their full growth cycles, while the (2004), all AVAs in our study are categorized as CI+ 2
chilling period could be insufficient for the Puget regions with a CI ≤ 12°C. This category of viticultural
Sound AVA. This index can be integrated with other climate has a great potential for producing quality
indices to more accurately match specific cultivars grapes when the thermal heat units in the region are
 Badr et al.: Bioclimatic indices for viticulture 217

Fig. 4. Key bioclimatic indices used in calculation of minimum

temperature for American viticulture area (AVAs) located in
the State of Washington and parts of Oregon. Class limits not
directly comparable. (a) Cool night index (CI); (b) mean ther-
mal amplitude (MTA); (c) dynamic minimum temperature
(DyMin.Temp.). See Fig. 1 for coordinates

sufficient (Tonietto & Carbonneau 2004). A lower CI also had no significant difference at the 95% confi-
causes a lower rate of metabolism for aromatic mate- dence level. This index has been shown to influence
rials and pigments during the night. the quality of grapes, including grape composition,
flavor, and aroma (Mullins et al. 1992, Ramos et al.
2008, Montes et al. 2012). Therefore, if the same man-
3.3.2. MTA agement and irrigation strategies are used, AVAs with
similar MTA can potentially produce grapes with
The calculations indicated that Puget Sound had many similar quality traits. However, further assess-
the lowest MTA (11.4°C) among all the AVAs while ments are needed to evaluate fruit quality traits in
Snipes Mountain had the highest (18.1°C). Multiple AVAs with comparable bioclimatic ranges.
comparisons among the AVAs indicated that there
was no significant difference in the MTAs of Colum-
bia Gorge, Colombia Valley, Red Mountain, and 3.2.3. DyMin.Temp.
Wahluke Slope at a 95% confidence level (Table S2,
Fig. 4b). Ancient Lakes of the Columbia Valley, The number of events when the minimum air
Horse Heaven Hills, and Walla Walla had no signifi- temperature dropped below a certain temperature
cant difference at the 95% confidence level. Lake threshold that changes across multiple seasons were
Chelan and the Rocks District of Milton-Freewater calculated for all AVAs for a period of 30 yr. Puget
218 Clim Res 76: 203–223, 2018

Table 10. Number of incidents for each American viticulture frequent freezes, and evaluate each region based
area (AVA) in the study area when the minimum air temper- on its resilience, adaptability, and technological
ature drops below a cold hardiness threshold or cold dam-
advancement.
age index (CDI) calculated for 5, 10, 20, and 30 yr out of the
historical weather data. The number of events for a century
was calculated based on the CDI for 30 yr
3.2.4. CDI
AVA CDI
5 yr 10 yr 20 yr 30 yr century The total number of events when the minimum
temperature dropped below a dynamic threshold
Puget Sound 0 1 1 1 5
(Bud LT50) was determined for 5, 10, 20, and 30 yr
Red Mountain 4 5 13 15 49
periods and projected for 100 yr (Table 10). Puget
The Rocks District 5 5 5 22 74
Sound proved to have the lowest number of events (0
of Milton-Freewater
within 5 yr and 1 for up to 30 yr), while the highest
Wahluke Slope 6 7 13 18 59
Walla Walla 7 9 47 16 52
number of events was found for Naches Heights (30,
Snipes Mountain 7 12 25 33 112
47, 86, and 119 events for 5, 10, 20, and 30 yr, respec-
Horse Heaven Hills 9 12 20 23 78
tively) (Fig. 5a−d). Our comparison of the CDI analy-
Yakima Valley 11 17 33 46 155
sis indicated that there was a significant difference
Ancient Lakes of 11 13 23 30 99
between the Puget Sound results and the other AVAs
the Columbia Valley (Table S2).
Lake Chelan 13 16 30 38 126 The impact of CDI events for a given location
Columbia Gorge 13 23 43 56 187 depends on various factors, such as the duration of a
Columbia Valley 14 20 36 48 161 cold spell, economic profitability of the vineyard,
Rattlesnake Hills 15 23 13 72 239 availability of affordable technologies, and the miti-
Naches Heights 30 47 86 119 397 gation of the grower to the risks associated with
potential crop loss. Discussing the limiting ranges of
CDI events is beyond the scope of this paper and
Sound was determined to have the lowest number of needs further investigation. Therefore, the results
events (3) (Table S2), while Naches Heights had the presented here can be used as a foundation for future
highest number of events (45) (Fig. 4c). The multiple studies using this index as an effective tool for the
comparison of the DyMin.Temp. showed significant determination of regions that have a high number of
differences between AVAs (Table S2) except for CDI events.
Ancient Lakes of the Columbia Valley and Rattle-
snake Hills besides Red Mountain and Columbia
Gorge. The amount of damage caused by exposure 3.4. Bioclimatic indices focusing on precipitation
to a low temperature depends on where individual
plants are in their phenological stages, showing that Our results indicated that, for the majority of AVAs,
both temporal and spatial analysis of this index are most precipitation occurs during winter and early
critical for both the Naches Heights and Puget Sound spring (~60%), except for the Columbia Gorge
AVAs. The temporal distribution of events was fo- (~40%). This information can be beneficial for deci-
cused on April, with no incidents occurring during sion-makers and growers in determining their options
the summer months (June, July, and August). The for supplemental irrigation. Furthermore, the amount
concentration of most of the events in April might be of precipitation that finally reaches a grapevine’s root
partially due to the assumptions made for the calcu- zone is generally lower than the recorded precipita-
lation of the index, which set the April low tempera- tion, due to the partial interception of raindrops by the
ture threshold to a value of 0. canopy and vineyard cover crops, as well as evapora-
Our study is the first to consider a dynamic mini- tion at the surface. Therefore, the actual available
mum temperature rather than a fixed minimum tem- water to the plants is even lower than the calculated
perature. Although this index is not sufficient to fully values based only on precipitation. Hence, when as-
categorize grape-growing regions, it can be used in sessing the precipitation of various AVAs, 2 questions
conjunction with other indices to better explore cli- must first be answered: (1) does the region require
mate variability and weather extremes in viticultural supplementary irrigation and (2) if the region requires
regions. Future studies should evaluate the risks supplementary irrigation, does a proposed site for a
associated with viticulture in regions that suffer from vineyard have access to water and rights to use it?
 Badr et al.: Bioclimatic indices for viticulture 219

Fig. 5. Cold damage index (CDI) for American viticulture area (AVAs) located in the State of Washington and parts of Oregon for
(a) 5 yr; (b) 10 yr; (c) 20 yr; (d) 30 yr. See Fig. 1 for coordinates

3.4.1. GSP regarded as excessively humid. All AVAs that were

assessed in our study had a GSP lower than 200
GSP was computed for all the AVAs. Snipes mm, except for the Columbia Gorge and Puget
Mountain had the lowest GSP (52 mm) while Puget Sound (Fig. 6a).
Sound had the highest (332 mm) (Table 11). The
GSP for Puget Sound was notably different when
compared to the other AVAs (Table S2). The GSPs 3.4.2. OutGSP
for Columbia Gorge and Walla Walla also showed
significant difference compared with the other AVAs OutGSP provides further insight into the available
at the 95% confidence level (Fig. 6a, Table S2). soil water at the start of the growing season. Our
This index provides an accurate means of determin- results indicated that Wahluke Slope had the lowest
ing the general suitability of a region for viticulture OutGSP (73 mm) and Puget Sound had the highest
(Santos et al. 2012), especially in locations where (415 mm) (Fig. 6b, Table 11). Preference is generally
irrigation is not an option. If a region has a GSP that given to regions that receive the majority of their pre-
is less than 200 mm, it is regarded as extremely dry; cipitation during the winter, thus providing the
while if the GSP is greater than 600 mm, it is plants in these regions with sufficient soil water prior
220 Clim Res 76: 203–223, 2018

Table 11. Growing season precipitation (GSP), out-of-growing-season precipi- 2014). However, downy mildew is not
tation (OutGSP), and average total annual precipitation for the American viti- a cause of concern since the pathogen
cultural areas (AVAs) in the study area and their respective percentage of the
does not yet occur, mainly due to un-
total precipitation
favorable weather conditions for this
pathogen in Washington.
AVA GSP OutGSP Total GSP OutGSP
(mm) (mm) precipitation (%) (%)
(mm)
3.5. Bioclimatic index focusing on
Snipes Mountain 52 78 130 40 60 wind speed
Rattlesnake Hills 57 90 147 39 61
Wahluke Slope 58 73 132 44 56 The WSI was calculated for the
Ancient Lakes of 59 78 137 43 57 AVAs located in our study area, with
the Columbia Valley
Red Mountain proving to be the least
Red Mountain 61 86 147 41 59
windy (16 incidents) and Snipes
Yakima Valley 63 87 150 42 58
Mountain the windiest (56 incidents)
Horse Heaven Hills 65 96 160 40 60
(Table S2). Red Mountain and Snipes
Columbia Valley 81 120 201 40 60
Mountain had significant differences
Lake Chelan 83 108 191 44 56
at the 95% confidence level for WSI
Naches Heights 96 112 208 46 54
compared to other AVAs. Wind speed
The Rocks District of 98 185 283 35 65
Milton-Freewater is influenced by the complexity of
Walla Walla 112 197 309 36 64 topographic elements in a region as
Columbia Gorge 262 195 457 57 43 well as the variability in the thermal
Puget Sound 332 415 747 44 56 properties of different land cover
types, leading to a thermal gradient.
This index cannot be used as a stand-
to the start of growth and development in early alone tool for climate zoning. However, it can serve
spring (Jackson 2008). as a surrogate to help growers and decision-makers
Gladstones (1992) reported an average GSP of 53 gain valuable knowledge about a site. In addition, it
and 70 mm for Fresno and San Jose, California. The can help them make more effective decisions regard-
calculated GSP for the AVAs in the current study ing management strategies such as installing a wind-
showed that Snipes Mountain, Rattlesnake Hills, break. Future studies can use this index to focus on
Wahluke Slope, Ancient Lakes, Red Mountain, Yaki- locations that have higher WSI scores and to conduct
ma Valley, and Horse Heaven Hills are within the more detailed studies on the impact of wind speed on
same range of GSP reported for these locations in vines, such as physical damage, shoot growth, and
California (Gladstones 1992). However, the OutGSP photosynthesis disruption.
for these AVAs is lower than the values reported for
Fresno (178 mm) and San Jose (300 mm). It is impor-
tant to evaluate the potential of a region based on 3.5.1. Meso-climate and terrain features
both the GSP and OutGSP in order to better under-
stand the options for rainfed viticulture versus irri- The similarities and differences between the AVAs
gated viticulture. based on their bioclimatic indices can be described by
the factors that impact the meso-climate, specifically
the topographic features and complexity of the terrain
3.4.3. HyI in a region, including the elevation, slope, and aspect
of the sites. The effect of elevation on minimum tem-
Among all the AVAs, Wahluke Slope had the low- perature was clearly captured for Naches Heights, as
est HyI (833.9) and Puget Sound had the highest this AVA has the highest elevation of all AVAs (543
(3144.6) (Table S2, Fig. 6c). The multiple comparison m), and the lowest CI, lowest FFD, highest CDI, and
indicated that Puget Sound and Columbia Gorge had highest DyMin.Temp. Puget Sound had the lowest el-
a significant difference when compared to all other evation of all AVAs, and the highest CI, highest FFD,
AVAs and between themselves. HyI indicates the lowest CDI, and lowest DyMin.Temp.
potential risk of downy mildew (Santos et al. 2012) in Weather parameters such as relative humidity also
a region as well as the water availability (Fraga et al. indirectly impact the behavior of bioclimatic indices.
 Badr et al.: Bioclimatic indices for viticulture 221

Fig. 6. Key bioclimatic indices focused on the precipitation

of American viticulture area (AVAs) located in the State of
Washington and parts of Oregon. Class limits are not di-
rectly comparable for the hydrothermic Index. (a) Growing
season precipitation (GSP); (b) out-of-growing-season pre-
cipitation (OutGSP); (c) hydrothermic index (HyI). See
Fig. 1 for coordinates

Hence, for an index such as HyI, the value increases important role in capturing the dynamics of weather
in wet regions such as Puget Sound, where the rela- in each AVA. Hence, the results of this study could
tive humidity is much higher compared to other potentially improve our ability to delineate cultivar
regions located to the east of the Cascade Mountains. suitability zones within each AVA in the study
Consequently, future studies should relate the im- area.
pact of relative humidity on the overall dynamics of
bioclimatic indices on a local scale within each AVA.
There is also a need for development of a manage- 4. CONCLUSIONS
ment framework based on the current cultural prac-
tices that are implemented in AVAs with similar cli- This study categorized the AVAs located in the State
mate groupings. of Washington and parts of Oregon based on the
The derived zones based on the bioclimatic dynamics of bioclimatic indices that were evaluated
indices confirmed the previous reports (Jones et al. for a 30 yr period. Several new bioclimatic indices
2010) on the zoning of the AVAs in the study area. were introduced, including CDI, DyMin.Temp., and
The derived ranges for GDD, HI, and BEDD for our WSI. Future studies should focus on the development
study area completely aligned with the ranges and improvement of bioclimatic indices based on
reported by Jones et al. (2010) for the same study recent advancements in sensor technology and the
area. The spatial and temporal resolution of the availability of finer resolution spatial and temporal
data that was used in this current study played an data.
222 Clim Res 76: 203–223, 2018

Acknowledgements. This research was partially supported vines. J Crop Improv 10:201−235
by Washington State University’s AgWeatherNet Program, Ferguson JC, Tarara JM, Mills LJ, Grove GG, Keller M
the Northwest Center for Small Fruits Research, and an IBM (2011) Dynamic thermal time model of cold hardiness for
Fellowship awarded to G.B. The authors thank the Univer- dormant grapevine buds. Ann Bot 107:389−396
sity of Idaho gridded surface meteorological data (UI GSM) Ferguson JC, Moyer MM, Mills LJ, Hoogenboom G, Keller
for providing access to the raw weather data that were used M (2014) Modeling dormant bud cold hardiness and
in this study. The authors are also grateful to Dr. Carol Wilk- budbreak in twenty-three Vitis genotypes reveals varia-
erson for her help reviewing this manuscript. tion by region of origin. Am J Enol Vitic 65:59−71
Fraga H, Malheiro AC, Moutinho-Pereira J, Santos JA
(2013) Future scenarios for viticultural zoning in Europe:
LITERATURE CITED ensemble projections and uncertainties. Int J Biometeo-
rol 57:909−925
Abatzoglou JT (2013) Development of gridded surface mete- Fraga H, Malheiro AC, Moutinho-Pereira J, Santos JA
orological data for ecological applications and modelling. (2014) Climate factors driving wine production in the
Int J Climatol 33:121−131 Portuguese Minho region. Agric For Meteorol 185:26−36
AgriMet (Cooperative Agricultural Weather Data) (2011) Freeman BM, Kliewer WM, Stern P (1982) Influence of
www.usbr.gov/pn/agrimet/wxdata.html (accessed 16 windbreaks and climatic region on diurnal fluctuation of
September, 2017) leaf water potential, stomatal conductance, and leaf tem-
AgWeatherNet (2011) Daily weather data. Washington State perature of grapevines. Am J Enol Vitic 33:233−236
University, Prosser, Washington. http://weather.wsu.edu Gladstones J (1992) Viticulture and Environment. Wine-
(accessed 23 February, 2016) titles, Adelaide
Allen RG, Pereira LS, Raes D, Smith M (1998) Crop evapo- Hall A, Jones GV (2010) Spatial analysis of climate in wine
transpiration — Guidelines for computing crop water grape-growing regions in Australia. Aust J Grape Wine
requirements — FAO Irrigation and drainage paper 56. Res 16:389−404
FAO, Rome Hamilton RP (1988) Wind effects on grape vines. In: Smart
Amerine MA, Winkler AJ (1944) Composition and quality of RE, Thornton RJ, Rodriques SB, Young JE (eds) Proc 2nd
musts and wines of California grapes. Hilgardia 15: International Symposium for Cool Climate Viticulture
493−675 and Oenology, January 1988, Auckland. New Zealand
Anderson JD, Jones GV, Tait A, Hall A, Throught MCT (2012) Society of Viticulture and Oenology, Wine Institute of
Analysis of viticulture region climate structure and suit- New Zealand, Auckland, p 65−68
ability in New Zealand. J Int Sci Vigne Vin 46:149−165 Hidalgo L (2002) Tratado de viticultura general. Ediciones
Badr G, Hoogenboom G, Moyer M, Keller M, Rupp R, Dav- Mundi-Prensa, Madrid
enport J (2018) Spatial suitability assessment for vine- Huglin P (1978) Nouveau mode d’évaluation des possibilités
yard site selection based on fuzzy logic. Precis Agric héliothermiques d’un milieu viticole. CR Acad Agric Fr
2018:1–22 64:1117−1126
Becker N (1985) Site selection for viticulture in cooler cli- Huglin P, Schneider C (1998) Biologie et écologie de la vigne.
mates using local climatic information. In: Proc Interna- Lavoisier, Paris
tional Symposium on Cool Climate Viticulture and Enol- Hunter JJ, Bonnardot V (2011) Suitability of some climatic
ogy. Oregon State University Technical Publication 7628: parameters for grapevine cultivation in South Africa,
20−34 with focus on key physiological processes. S Afr J Enol
Blanco-Ward D, Queijeiro JMG, Jones GV (2007) Spatial cli- Vitic 32:137−154
mate variability and viticulture in the Miño River Valley Jackson RS (2008) Wine science principles and applications,
of Spain. Vitis 46:63−70 3rd edn. Academic Press, Burlington, MA
Branas J (1974) Viticulture. Imprimerie Déhan, Montpellier Jackson DI, Cherry NJ (1988) Prediction of a district’s grape-
Campbell-Clause JM (1998) Stomatal response of grapevines ripening capacity using a Latitude-Temperature Index
to wind. Aust J Exp Agric 38:77−82 (LTI) Am J Enol Vitic 39:19−28
Clark JR, Wolf TK, Warren MK (1996) Thermal analysis of Jones GV (2005) Climate change in the western United
dormant buds of two muscadine grape cultivars and of States grape growing regions. Acta Hortic 41−60
Vitis labrusca L. ‘Mars’. HortScience 31:79−81 Jones GV (2007) Climate change: observations, projections
Conceição MAF, Tonietto J (2005) Climatic potential for and general implications for viticulture and wine produc-
wine grape production in the tropical north region of tion. Climate and Viticultural Congress, 10−14 April
Minas Gerais State, Brazil. Rev Bras Frutic 27:404−407 2007, Zaragoza. OIV, Paris, p 55–66
Daly C, Halbleib M, Smith JL, Gibson WP and others (2008) Jones GV, Snead N, Nelson P (2004) Geology and wine. VIII.
Physiographically sensitive mapping of temperature and Modeling viticultural landscapes: a GIS analysis of the
precipitation across the conterminous United States. Int J terroir potential in the Umpqua valley of Oregon. Geosci
Climatol 28:2031−2064 Can 31:167−178
Davenport JR, Keller M, Mills LJ (2008) How cold can you Jones GV, Moriondo M, Bois BA, Hall A, Duff A (2009)
go? Frost and winter protection for grape. HortScience Analysis of the spatial climate structure in viticulture
43:1966−1969 regions worldwide. Bull OIV 82:507−518
Dry PR, Reed S, Potter G (1989) The effect of wind on the Jones GV, Duff AA, Hall A, Myers JW (2010) Spatial analysis
performance of cabernet franc grapevines. Acta Hortic of climate in winegrape growing regions in the western
143−146 United States. Am J Enol Vitic 61:313−326
ESRI (Environmental Systems Research Institute) (2015) Keller M (2010) The science of grapevines: anatomy and
ArcGIS Desktop10.2. ESRI Academy Redlands, CA physiology. Academic Press, Burlington, MA and San
Fennell A (2004) Freezing tolerance and injury in grape- Diego, CA
 Badr et al.: Bioclimatic indices for viticulture 223

Magalhães N (2008) Tratado de viticultura: a videira, a vinha classification system for grape-growing regions world-
e o terroir. Chaves Ferreira Publicações, Lisboa wide. Agric Meteorol 124:81−97
Magarey R, Seem RC, DeGloria SD (1998) Prediction of vine- Tóth JP, Végvári Z (2016) Future of winegrape growing
yard site suitability. Grape Research News 9:1−2. New regions in Europe. Aust J Grape Wine Res 22:64−72
York State Agricultural Experiment Station, Geneva TTB (Alcohol and Tobacco Tax and Trade Bureau) (2015)
Makra L, Vitanyi B, Gal A, Mika J, Matyasovszky I, Hirsch T American Viticultural Area (AVA). US Department of
(2009) Wine quantity and quality variations in relation to Treasury, Regulations and Rulings Division. www.ttb.
climatic factors in the Tokaj (Hungary) winegrowing gov/wine/ava.shtml (accessed 6 March, 2015)
region. Am J Enol Vitic 60:312−332 UI GSM (2015) The University of Idaho’s Gridded Surface
Malheiro AC, Santos JA, Fraga H, Pinto JG (2010) Climate Meteorological Data. http://metdata.northwestknowledge.
change scenarios applied to viticultural zoning in net/. (accessed 20 February, 2015)
Europe. Clim Res 43:163−177 USDA-NASS (USDA National Agricultural Statistics Serv-
Malheiro AC, Santos JA, Pinto JG, Jones GV (2012) Euro- ice) (2015) http://quickstats.nass.usda.gov
pean viticulture geography in a changing climate. Bull USHCN (2011) United States Historical Climate Network.
OIV 85:971−973 http://cdiac.ornl.gov/ftp/ushcn daily/ (last accessed on
MATLAB V2014a (2014) MATLAB and Simulink 2014a. February 22, 2016)
MathWorks, Natick, MA Van Leeuwen C, Seguin G (2006) The concept of terroir in
Mills LJ, Ferguson JC, Keller M (2006) Cold-hardiness eval- viticulture. J Wine Res 17:1−10
uation of grapevine buds and cane tissues. Am J Enol Van Leeuwen C, Friant P, Choné X, Trégoat O, Koundouras
Vitic 57:194−200 S, Dubourdieu D (2004) The influence of climate, soil and
Mitchell KE, Lohmann D, Houser PR, Wood EF and others cultivar on terroir. Am J Enol Vitic 55:207−217
(2004) The multi-institution North American Land Data Wample RL, Hartley S, Mills L (2001) Dynamics of grape-
Assimilation System (NLDAS): Utilizing multiple GCIP vine cold hardiness. In: Rantz JM (ed) Proc Am Soc Enol
products and partners in a continental distributed hydro- Vitic 50th Anniversary Annual Meeting, June 19−23,
logical modeling system. J Geophys Res 109:D07S90 2000, Seattle, WA. ASEV, Davis, CA, p 81−93
Montes C, Perez-Quezada JF, Peña-Neira A, Tonietto J White RE (2015) Understanding vineyard soils. Oxford Uni-
(2012) Climatic potential for viticulture in Central Chile. versity Press, New York
Aust J Grape Wine Res 18:20−28 Williams LE, Dokoozlian NK, Wample R (1994) Grape. In:
Mullins MG, Bouquet A, Williams LE (1992) Biology of the Schaffer B, Andersen PC (eds) Handbook of environ-
grapevine. Cambridge University Press, Cambridge mental physiology of fruit crops, Vol 1. Temperate crops.
Prescott JA (1965) The climatology of the vine (Vitis vinifera CRC Press, Boca Raton, FL, p 85−133
L.). The cool limits of cultivation. Trans R Soc S Aust 89: Winkler AJ, Cook JA, Kliewer WM, Lider LA (1974) General
5−23 viticulture, 4th edn. University of California Press,
Ramos MC, Jones GV, Martínez-Casasnovas JA (2008) Struc- Berkeley
ture and trends in climate parameters affecting winegrape Wolfe W (1999) Site selection in eastern Washington: opti-
production in northeast Spain. Clim Res 38:1−15 mizing sSite and variety choices. In: Watson J (ed) Grow-
RAWS (Remote Automatic Weather Stations) (2011) RAWS ing grapes in eastern Washington. Proceedings from the
USA climate archive. https://raws.dri.edu (accessed 16 1998 Washington State University shortcourse for estab-
September, 2017) lishing a vineyard and producing grapes. Good Fruit
Rosenberg NJ, Blad BL, Verma SB (1983) Microclimate: the Grower, Yakima, WA, p 27−30
biological environment. Wiley, New York, NY Wolfe W (2001) Vine and vineyard management following
Santos JA, Malheiro AC, Pinto JG, Jones GV (2012) Macro- low temperature injury. In: Rantz JM (ed) Proc Am Soc
climate and viticultural zoning in Europe: observed Enol Vitic 50th Anniversary Annual Meeting, June 19−
trends and atmospheric forcing. Clim Res 51:89−103 23, 2000, Seattle, WA. ASEV, Davis, CA, p 101−110
Santos JA, Grätsch SD, Karremann MK, Jones GV, Pinto JG WorldClim (2009) www.worldclim.org/ (accessed 16 Sep-
(2013) Ensemble projections for wine production in the tember, 2017)
Douro Valley of Portugal. Clim Change 117:211−225 WSW (Washington State Wines) (2015) www.washington-
SAS (2015) SAS software v9.4. SAS Institute, Cary, NC wine.org/wine-101/regions/ (accessed 16 September,
Takahashi K, Kuranaka M, Miyagawa A, Takeshita O (1976) 2017)
The effect of wind on grapevine growth; windbreaks for Yau IH (2011) Developing a grape site selection GIS for the
vineyards. Bull Shimane Agric Exp Stn 14:39−83 Inland Pacific Northwest. MSc thesis, Washington State
Tarara JM, Ferguson JC, Hoheisel GA, Perez Peña JE (2005) University, Department of Crop and Soil Sciences. Avail-
Asymmetrical canopy architecture due to prevailing wind able at www.dissertations.wsu.edu/Thesis/Fall2011/i_yau
direction and row orientation creates an imbalance in _092211.pdf
irradiance at the fruiting zone of grapevines. Agric For Yau IH, Davenport JR, Rupp RA (2013) Characterizing
Meteorol 135:144−155 inland Pacific Northwest American Viticultural Areas
Tonietto J, Carbonneau A (2004) A multicriteria climatic with geospatial data. PLOS ONE 8:e61994

Editorial responsibility: Toshichika Iizumi, Submitted: January 29, 2018; Accepted: August 14, 2018
Tsukuba, Japan Proofs received from author(s): October 10, 2018