Impact of Early Season Temperatures in a Climate-Changed Atmosphere for Michigan: A Cool-Climate Viticultural Region

Abstract

This study assesses the impacts of observed (2001–2012) and projected climate change on early season heat accumulation for grape production (viticulture) in a cool-climate region of the world (Michigan, USA). Observational data were generated from a weather station located in the center of one of the most important appellations located in the SW part of the state. Climate change projections retrieved from a high spatiotemporal weather model using atmospheric conditions simulated for the end of the 21st Century. All the temperature variables considered demonstrated significant warming trends especially during the months of March, April and May. Temperature differences, increases in heat accumulation, and changes to potential frost events would necessitate new approaches to vineyard management. In fact, it is likely that vine budburst will occur earlier and early season frost develop as a new challenge. However, our study results are comparable with other viticulture regions of the world, with a warming trend of at least 3 to 5 °C in the months leading up to the growing season by the end of the 21st Century. Therefore, effective climate change adaptations will be important to the grape and wine industry in this region. Management strategies are needed to minimize climate risks while taking advantage of new opportunities related to improved climatic conditions for growing more late-ripening European Vitis vinifera grape varieties, relevant for producing higher quality wines.

1. Introduction

Viticulture is an important economic agricultural sector in Michigan (USA) and in many regions of the world. This sector has been developed mainly in regions with specific climatic conditions, namely (1) adequate heat accumulation during the summer, (2) low risk of frost occurrence in the spring and (3) no extreme heat and water stress events. However, world grape production areas are experiencing significant climatic changes in recent years, with warming trends reported in several viticultural areas of the world. California, Washington, Oregon and New York account for the majority, 94%, of U.S.-based grape production. Recently, new wine regions have been developed in regions with a cold winter, short growing seasons and hot-desert conditions. The state of Michigan is a fairly new wine-grape growing region in USA and Vitis vinifera wine grape production started in the late 60s, reaching 1200 hectares in 2018. In this cool-climate region, a warming trend has been already reported placing Michigan in the “zone of transition” from being able to grow only juice grapes to being suitable for production of vinifera varieties. Historically, the southwest region of the state of Michigan is one of the major grape-growing areas in the state, housing two of the five state appellations. This region has been classified as a “cool” viticulture production area [1]. This classification, where average Growing Season Temperatures (GST, from the months of April to October) are 13–15 °C, is reflected in the varieties that typically grow in the region, namely wine grape varieties that produce wines with higher acidity and lower alcohol concentrations. While some evidence has shown that southwest Michigan should be regarded as an intermediate climate (GST = 15–17 °C, [2]), this region typically produces white wines such as Riesling and Chardonnay and lighter red wines such as Pinot Noir. Regardless of GST-based classification, the region does endure routinely severe winter temperatures on a yearly basis and potentially hazardous spring temperatures.

Compared with any wine region that must contend with cooler temperatures, yearly success in Michigan is strongly linked to spring (Mar–May) temperatures [2,3]. Success for a crop on yearly basis, or failure, typically depends on when vines come out of dormancy and move on to the budburst stage with earlier dates leading to higher potential risk of damage from spring frost. If a vineyard can avoid post-budburst frosts, summer temperatures are warm enough to facilitate grape clusters as they progress towards technological maturity. As such, vineyard managers must develop adaptation and mitigation strategies for spring frost conditions, particularly after vine budburst, as this can be severely limiting to the potential end-season crop output [4,5,6,7,8]. Similar trends in early spring warm temperatures and risk of frost events have been studied in northern France [9], Germany [10], Canada [11], Luxembourg [12], Switzerland [13], and other similar wine regions around the world [14,15]. Climate variability is expected in the field of viticulture. The term terroir, or “land characteristics,” encompasses the idea that variation across both space and time can be expressed within wine grapes and resulting wines on a year-to-year basis [16,17]. Climate is one of the components of terroir, and there is evidence to suggest that it is the most important [18]. Since the first study in 1944, which classified regions based on climate [19], the fields of climatology and viticulture have combined to create a better understanding of how the two sciences work together. Numerous papers in the second half of the 20th Century and very early 21st Century documented how wine grapes responded to changes to climate [20,21,22,23,24,25].

However, in the past 20 years, there has been an abundance of published research that used climate models as proxies to identify potential changes to future wine grapes production regions [26,27,28,29,30]. These papers either look to explain global trends [31,32] or focus on one or two regions [33,34,35] with the goal of explaining what climate change could mean for viticulture in those regions in the future. In all cases, the future will see higher temperatures and changes to precipitation trends, which will both influence wine grape production in any corner of the globe. Nevertheless, many of these papers used models with spatial and temporal resolutions that were quite coarse. For a science such as viticulture, where terroir and spatial inhomogeneity is a central tenet, it would be difficult to draw strong conclusions while using climate model data with 50 km spatial resolution at a one-month timestep over an area with complex topography. While these papers produced excellent results, concepts that are highly important to wine grape growing, such as cold-air drainage, simply cannot be resolved by such models. Additionally, many of these papers focused on the effects of climate change on general growing season conditions with few focused strictly on spring conditions. Spring months may not be as important to warm-climate regions such as southern Spain or Sicily, but for cooler regions, this is a timeframe that needs further study, especially when one considers that many of these regions are likely to rise to prominence in the coming decades.

The primary objective of this research was to use climate model data with a very high spatiotemporal resolution in an effort to describe the differences in spring conditions in southwest Michigan, in a climate-changed world where atmospheric chemistry has been altered. By using high-resolution data, we expect to find trends that would not have been able to be resolved using older climate model data. We show these differences and discuss the implication for viticulture both in Michigan and the rest of the world.

2. Experiments

2.1. Study Area

The primary area of study is located in southwest Michigan (Figure 1). The region is classified as Dfa Köppen Climate Classification [36,37]. The region experiences moderated temperatures throughout the year due to its proximity to Lake Michigan. Located downwind of the large waterbody, the region will typically see spring temperatures warm up later and will see colder fall–winter temperatures set in later than in upwind areas such as Wisconsin and Minnesota [38]. In winter, extreme low temperatures are strongly linked to ice coverage on the lake wherein full ice coverage can lead to temperatures far colder than a lake with very little ice coverage [39,40]. During the summer, a mesoscale lake breeze can set up with proper atmospheric conditions which can irregularly increase local precipitation and decrease high daily temperatures, at varying points in the region [41].

For centuries, growers have taken advantage of Lake Michigan’s moderating effects. This has allowed a “fruit belt” to exist along the western coast of Michigan. Temperatures are warmest along the southern end; thus, the highest diversity of fruits and vegetables can be supported in southwest Michigan. In addition to grapes, the region can accommodate peaches, apples, cherries, numerous berries, tomatoes, celery, asparagus, and several staple crops [42]. This diversity could not be as easily maintained further inland from Lake Michigan as harsher winter temperatures would be a limiting factor.

Grapes have been grown for centuries in the southwest Michigan area. The majority of grapes grown in the region are from of the Concord or Niagara variety of Vitis labrusca, an indigenous North American grapevine [43]. However, in recent decades, there has been a shift towards growing Vitis vinifera grapes, grown for wine production. The initial increase in vinifera plantings in both northwest and southwest Michigan started in the 1990s and may be explained by rising temperatures in the region. There has been a steady shift from juice grape production to vinifera production ever since [44]. This shift has come from a desire to access areas that have recently been able to accommodate wine grape production thanks in large part to climate change over the past several decades [45]. The region has warmed significantly [2], has seen the grape growing season expand by more than four weeks since the early 1970s [6], and likely should not be considered to be a “cool-climate” viticultural region as per the average growing season temperature conditions set forth in Jones (2007) [1]. As the region continues to warm, it is likely that new varieties, which require warmer and reliably longer growing seasons, will be introduced in the coming decades [46]. However, during this expansion of vinifera acreage, the importance of spring months has been noted. Should a vineyard get through the months of March, April, and May (MAM), it is likely that the season will produce a viable harvest [3]. As such, growers must use viticultural methods to prevent (or mitigate) potential damage from these hazardous months.

2.2. Data

Gridded data were obtained from a dataset jointly created by the National Center for Atmospheric Research (NCAR) and University Corporation for Atmospheric Research (UCAR). This dataset is principally a Weather Research and Forecasting (WRF) model that operates at 4 × 4 km cell resolution and an hourly timestep from October 2000 to September 2013 for the entire continental United States, plus parts of Canada, Central America, and the Caribbean [47]. The high resolution of this model, in comparison to other climate models, allows for the direct analysis of mesoscale–bordering on the microscale (on the line between Orlansky’s Meso-γ and Micro-α [48]) events such as cold-air drainage and convection. These processes, along with many others, are essential to understanding the potential diverse effects of spatial inhomogeneity over landscapes and how they may influence climates [49]. Resolving these processes are essential to understanding how climate influences terroir.

The following two simulations were performed: (1) a retrospective simulation of weather conditions from October 2000 to September 2013, which used ground-based, observed data to create the gridded model (the WRF-observed model), and (2) a future climate sensitivity model wherein the same weather events and patterns were observed but with an atmospheric chemistry projected at the end of the 21st Century by the CMIP5 RCP8.5 greenhouse gas (GHG) emissions scenario which includes a carbon dioxide concentration of 1380 ppmv and global temperatures at least 3 °C warmer than the present [50,51]. The “pseudo-global warming” (PGW) model run, with higher (GHG) concentrations, represents a “climate changed” atmosphere [47]. While this may be considered a pessimistic scenario, the RCP8.5 scenario was used only because the dataset had only the real-world observations and the high-end emissions scenario. Had a lesser pessimistic scenario been included, the authors would have included it as well. By comparing the two models, a user can observe the effects of having higher GHG concentration by viewing the same approximate weather events in a different atmospheric chemistry.

A previous paper by the authors [46] found that the WRF-observed model was highly accurate when examining temperature data. Over the course of the study, there was a mean average error between the model and nearby mesonet sites of 0.38 °C for the paper’s study sites. Considering the reliability of the WRF-observed model, it was deemed that the inspection of weather events under the PGW-model would show the effects of a climate-changed atmosphere. This study compared the two models at the same point (SWMREC), which allowed the authors to compare “what happened” to “what could happen” at a very high spatiotemporal resolution. The authors’ previous paper [46] discussed the implications mostly for the summer months. However, this study focused on the highly important spring (MAM) months and compared the two model outputs. Long-term data (1950–2000) were also downloaded from the National Climate Data Center’s online mapper for Benton Harbor Airport (42.127° N, 86.429° W) [52].

2.3. Methods

This study compared data in the climatological spring months (MAM) for southwest Michigan. Hourly data were extracted from the WRF-observed and PGW models at the grid cell containing Michigan State University’s Southwest Michigan Research and Extension Center (SWMREC): 42.088° N, 86.356° W. This was completed for both models.

The calculation of thermal time called Growing Degree Days (GDD), a unitless measurement of thermal accumulation over time which allows for comparison of regions or timeframes. This study used the GDD basic calculation with a base of 10 °C for both models [53]:

$$GDD=\frac{Tmax-Tmin}{2}-10°\mathrm{C}$$

(1)

The calculation for “potential frost days” was calculated by counting the total number of days between March and May where the minimum temperature was below −1 °C [2,3,4,5,6]. The total of “potential frost hours” was calculated by totaling the entire number of hours between the same months below −1 °C.

3. Results

3.1. Effects on Spring Temperature

Spring month temperatures (MAM) were found to be at least 5.35 °C warmer in the PGW model run compared to the WRF-observed run (

Table 1. Difference in monthly temperatures between the two model scenarios (PGW–WRF observed) in °C for each spring month during the study time frame.

Year	March	April	May
2001	4.76	4.51	4.82
2002	4.58	4.20	5.05
2003	5.39	5.07	4.57
2004	3.75	4.31	4.49
2005	7.99	4.64	7.66
2006	6.37	5.65	6.18
2007	6.73	7.05	6.04
2008	7.24	4.76	4.92
2009	6.03	6.57	4.70
2010	7.93	4.72	5.32
2011	5.79	5.71	7.42
2012	5.68	6.95	4.69
Mean	6.02	5.35	5.49

). The PGW-model, a climate-changed atmosphere, runs projected average spring monthly temperatures for southwest Michigan to be approximately 5.7 °C (10.26 °F) warmer than the real-world results. This would result in a radically different landscape for the region and its vineyards.

Even the smallest change, March 2004 (3.75 °C, 6.75°F), would see temperatures warmer than what was observed. These warmer temperatures would almost certainly lead to vines exiting dormancy and entering budburst phenological stage early. This follows a trend found in previous papers that warmer temperatures typically force vines in to phenological stages earlier than in cooler conditions [3,6,24].

There does not appear to be a trend wherein one month had a significantly higher rate of warming than the others (ANOVA, p <0.05), though March showed the highest rate of warming (6 °C) compared to the observed model. It does not appear that the model showed a higher rate of warming in years that were significantly warmer or cooler than the climatological norm. For example, the 2012 spring was exceptionally warm for the state of Michigan, where temperatures for the month of May averaged more than 10 °C higher than normal in some places. Fruit crops responded by entering the budburst phase, only to be hit with damaging frosts which destroyed >80% of the state’s apples, cherries, and peach crops. Grapes were hit similarly hard. This study found that while the difference between the PGW and observed results in 2012 was not significantly higher than any other year (5.8 °C warmer), the MAM average temperature in that year was 20.2 °C, or 9.4 °C above average, in a climate-changed atmosphere (Figure 2). This was also true in a particularly cold year (2002) wherein the temperature change was not noticeably higher or lower than any other year.

3.2. Effects on GDD Accumulation

While there appears to be no yearly trend in GDD accumulation, there appears to be a monthly trend. Using base 10 °C, the accumulations are higher in May than in March. The WRF-observed model rarely ever accumulates any March GDD, thus, most of the GDD difference comes simply from the PGW-model total accumulations. However, by May, the accumulation difference was statistically significantly higher (ANOVA, p < 0.05)—more than three times higher.

Traditionally, budburst occurs near the end of April, though it can happen early with enough warmth in the months of March and early April. The WRF-observed average GDD accumulation at the end of April was approximately 150. However, the average PGW-model GDD accumulation at the same was nearly 320. These conditions would almost certainly alter the early phenological stages of grapevine development. Total spring GDD accumulations averaged a difference of 308 GDD (

Table 2. Difference in monthly growing degree day (GDD) accumulation between the two model scenarios (PGW–WRF-observed, base 10 °C) for each spring month during the study time frame.

Year	March	April	May	Total
2001	3	101	147	251
2002	11	55	119	186
2003	46	84	139	269
2004	29	62	124	215
2005	53	103	219	375
2006	45	136	173	354
2007	81	126	185	392
2008	27	83	143	253
2009	35	113	140	288
2010	108	114	152	373
2011	22	102	219	342
2012	98	158	145	400
Mean	46	103	159	308

and Figure 3).

3.3. Effects on Potential Frost

Perhaps the most striking change concerned the amount of potential spring frost that southwest Michigan may encounter in a climate-changed atmosphere. Since frost is the primary concern for a grower in the spring (MAM) months, any change that lessens the potential for frost could be seen as beneficial.

Table 3. Difference in total number of days where daily minimum temperatures were found to be <−1 °C and the total number of hours <−1 °C between the two model scenarios for the March–May timeframe from 2001 to 2013.

Year	Less Days in PGW	Less Hours < −1 °C in PGW
2001	22	284
2002	18	219
2003	12	212
2004	13	161
2005	14	237
2006	22	208
2007	19	287
2008	23	314
2009	17	173
2010	10	109
2011	18	202
2012	6	57
2013	30	380
Total	224	2843

displays the changes in potential frost between the PGW-model and the WRF-observed model. It is clear that the number of days with potential frost are heavily reduced. On average, the number of days between March and May wherein daily minimum temperatures drop below −1 °C decreases by 17 days. One year saw 23 fewer potential frost days in the spring months. It should be noted that the exceptionally warm 2012 spring experienced a small decline in the number of potential frost days. While this seems counter-intuitive, the small decline reflects the small number of frost days recorded in the 2012 spring in the WRF-observed model (7 days). The PGW-model projects that all but one would be eliminated.

In addition to difference in potential frost days, we also calculated the number of total hours below −1 °C. −1 °C is not a guarantee for frost conditions; however, extended exposure to <−1 °C conditions can be damaging to the vines even if frost is not clearly present. As such, we calculated the difference in the total number of hours spent below −1 °C during the MAM time period for each year. The average difference was found to be 257 h per year less in the PGW-model than in the WRF-observed model (Table 3). Growers would have to worry less if their vines were growing in this kind of environment during the ever-crucial spring months.

4. Discussion

4.1. Implications for Viticulture in Study Area

Our results report that the Spring months (MAM) were radically different between the WRF-observed and the PGW runs. This supports the stark differences found in a related paper that looked at summer and winter temperatures in Michigan and other cool-climate viticultural regions [46]. Each of the climatological spring months of March, April, and May are at least 5.35 °C warmer (Table 1, Figure 2), on average, and, as expected, GDD accumulations are significantly higher in those months (Table 2, Figure 3). This marks a considerable change for months that are “make or break” in the region yet it falls within the expectations set by many climate-model-driven wine studies. In our study, we found that the climate of this area is moving towards more optimal conditions for cultivation of more heat-demanding varieties of grapes.

The observed climate change in the region appears to reduce the risk for wine grape cultivation as the region is now more suitable for cultivation of moderately late-ripening grape varieties. A significant increase of about 5 °C during the MAM months coincides with a potential early budburst and early flowering of grape varieties and consequently a longer season for the ripening period.

Perhaps the most noticeable changes were (1) the reduction in days with the potential for spring frost and (2) the reduction in potential total hours of frost that vines may experience in the Spring (Table 3). A reduction of an average of approximately 220 h of potential frost in the MAM months would mean that the regional concern for frost could be considerably lower.

However, this would not necessarily be a universally positive trend. Budburst has been occurring earlier in all corners of the viticultural world—Michigan, included—as the world has warmed up. Vines require a certain number of chilling hours, which need to be accumulated in the dormancy period. Once that requirement is met, a vine will overcome dormancy and restart annual growth once temperatures allow it to do so.

Temperatures suggested earlier in this paper show a region that almost certainly would have budburst occurring earlier than what has been seen historically on the order of months. Our study site’s 1971–2000 average March temperature was approximately 4.5 °C, but in our PGW scenario it would be nearly 10.5 °C—or what is currently expected to be seen in late April and early May. Considering SW Michigan’s average budburst date was April 27 from 1971 to 2011 [6], this implies that budburst could happen at least one month earlier.

This would shift SW Michigan’s phenological cycle ahead one month and would potentially lead to harvests in August rather than late September or even October and would mean that end-season frost events are less likely. Yet, a month-earlier budburst would still have to navigate frosts that would still likely occur in February and March. While frost rates diminish over time in a warming climate, they do not completely disappear. In essence, this would shift the frost problem we experience today a month earlier on the calendar. The problem, risk, and potential solutions are all likely to be the same as in the current MAM months, but now in late winter rather than mid-spring.

4.2. Implications for Global Viticulture

These changes would not be limited solely to Michigan. Other cool-climate viticultural regions around the world would be virtually certain to experience changes in a similar range. Production regions such as Germany, Austria, Switzerland, northern France, Tasmania, New Zealand’s South Island, British Columbia, the State of Washington, and portions of the Baltic, British Isles, and southern Argentina and Chile, can all expect to see their seasons begin earlier. Areas with April to October growing seasons could expect to see a March to September or a February to August date range from budburst to harvest. While growers have some methods that can be deployed to delay budburst for a while, delaying the action by an entire two months is unfeasible.

We conclude that these cool-climate regions can continue to experiment with “warmer” varieties of grape (i.e., Sangiovese or Barbera) because months such as July and August will reliably provide the required heat accumulation for those varieties. However, the growing season will have shifted forward in the calendar a month or two and these regions will still have concerns about average growing season temperatures. As such, cool-climate viticultural regions are not likely to be able to grow exceedingly warm varieties by the end of 21st Century, as would be suggested with a growing season from April to October that is >4.5 °C warmer [46]. In essence, these regions have a “ceiling” for what can be grown until routine early season frosts disappear.

5. Conclusions

This paper sought to highlight the potential changes to spring climates in a cool-climate viticultural area. It is a continuation of several papers produced by the authors as they continue to explore the impacts of climate change on cool-climate viticulture—particularly in Michigan [2,3,6,46]. While it was not surprising to see that Southwest Michigan would get warmer in the March-April-May months, the size of the increase in temperature was notable. Such changes would leave the region nearly unrecognizable compared to the present. Temperature differences, increases in GDD accumulation, and changes to potential frost events would require new strategies and practices to vineyard management.

Yet, it is not all positive. It is likely the budburst will occur earlier and likely change the typical growing season from April to October to a regular March to September cycle. An earlier budburst would still see early season frost as a challenge, and this is likely to be a limiting factor to Michigan, or any other cool-climate viticultural region.

While this study only used one model run, results were comparable to most other climate-wine studies that showed warming of 3 to 5 °C in a growing season by the end of the 21st Century. Further study would use other models that are downscaled to match the spatiotemporal resolution of the model used in this paper. However, this paper gives a glimpse of the potential radical changes that climate change may have on the global wine industry.