Observed trends in winegrape maturity in Australia

Data from wineries and vineyards in Australia

Records of observations made during winegrape harvests from 44 vineyard blocks in 12 winegrape growing regions in Australia, encompassing the periods 1895–2009, have been assessed (Fig. 1). The regions studied represent the full range of temperatures in which most winegrapes are grown in Australia (Smart et al., 1980). The regions, varieties and blocks represented are listed as well as the latitude/longitude of the vineyard sites (Table 1). Where more than one block of the same variety was assessed in a given region the variety was given a block code e.g. Shiraz (L) and Shiraz (M) from the Eden Valley. For the short-term datasets, the latitude/longitude was estimated to be located centrally in the respective region as confidentiality of these sites was requested by the provider of the data.

Data were classified as long term (a range of 25 years or longer, shaded grey in Table 1) and short term where the range of fewer than 25 years of records were available. The number of years in the range of observations was listed with the number of missing years noted in brackets. Data are listed in order of increasing average growing season temperature (Average GST,  °C), of the region, within the overall ‘dataset length’ groupings (Table 1) [see ‘Climate data’ for a definition of Average GST ( °C)].

In most vineyards, representative samples of grapes are collected from separate blocks at short intervals in the lead-up to harvest and sugar concentration measurements are taken from these samples. These measurements assist the grower in planning when to start harvesting a particular block. These records, or similar, were used in this study. Winegrape ripening profiles were derived from the recordings of changing sugar concentrations found in vintage diaries. From this information, the day of year grapes attained ‘maturity’ was derived for each year in every block. The method whereby maturity day was calculated from this data is described below (see ‘Calculation of maturity day’).

In most cases, the records were kept on site, though for the McLaren Vale site the records have also been included in the archival collection at the Adelaide library. Data were obtained through personal visits to wineries with manuscripts and vintage diaries being made available for perusal. Efforts to gain access to additional records from other regions were attempted but were unsuccessful for a variety of reasons such as insufficient time series length, records not kept or lost, or sugar concentration measurements were not available.

Calculation of maturity day

In many phenological studies, the harvest phase is defined as the date that the crop is picked (e.g. Chuine et al., 2004; Meier et al., 2007). However, with regard to winegrapes, harvest day has been described as a false phenophase because this date can be selected to suit changing style preferences or other constraints that can occur through the harvest period, e.g. impending rain or winery logistics (Jones et al., 2005; Petrie & Sadras, 2008). For these reasons, winegrape harvest day is not a physiologically consistent phenophase. In order to make a comparison of interannual ‘harvest date’, we assess the changing timing of the day of year a designated maturity, or level of ripeness, was attained (hereafter known as DOY_M) (after Petrie & Sadras, 2008).

Sugars, or total soluble solids (TSS) (unit: degrees Baume or °Bé) (Iland et al., 2000), accumulate in grapes from véraison to harvest as they ripen (Conde et al., 2007). TSS (°Bé) are typically measured and recorded periodically in the month leading up to harvest to assist with vintage planning, as growers attempt to pick the grapes at the optimum TSS level for their purpose (Krstic et al., 2003). By analysing the records of the TSS concentrations as these accumulate we can produce an ‘interannually comparable’ physiological measure of a phenological stage as close as possible to harvest date. The day of year where the ripening profile reached a ‘designated TSS (°Bé)’ level was the metric used to represent ‘maturity date’ the harvest proxy in this study, i.e. DOY_M. The designated TSS (°Bé) was selected for each block according to the wine style produced and the likelihood of it being recorded through the course of the harvest planning. For example, the TSS level selected to represent the Muscat à Petit Grains variety grown in the Rutherglen region was 15°Bé as these grapes were generally picked riper to produce a fortified wine, while the Marsanne grapes grown in Central Victoria are used to make a lighter style table wine, hence 10°Bé was the designated TSS in this case (Table 1).

The DOY_M had to be derived in cases where the designated TSS (°Bé) for the block (Table 1) was either above or below those recorded for the particular year. Two different methods were employed by which to make adjustments, the approach being guided by the availability and type of data recorded in the diaries. In the majority of cases winegrape harvest planning records were available (i.e. for all of the short-term datasets, Mornington Peninsula, Margaret River and Central Victorian blocks). For these datasets, DOY_M was extrapolated from annual ripening profiles. An example of the extrapolation procedure is demonstrated for the Marsanne block (Central Victoria), over the observation period 1970–1979 (Fig. 2). In this example, the accumulation of sugars is noted for each year of the decade as the grapes progressed through the ripening period. Where grapes did not actually achieve the designated maturity level of 10°Bé (the value selected as representative of maturity for this variety at this site), i.e. in 1972 and 1977, the line was extrapolated with the calculated average ripening rate for the block (This process was also carried out using a decadal ripening rate for each block in case there was some change in the rate through time. Using this method made no difference to the results so the more simple approach of applying the average ripening rate calculated over the entire period of observation is presented here.) (0.9°Bé week⁻¹) (Table 1). [The average rate of accumulation of sugar (°Bé week⁻¹) has been calculated from the records for each block and this varies, ranging from 0.4°Bé week⁻¹ for Chardonnay in the Mornington Peninsula to 1.2°Bé week⁻¹ for the same variety in the Barossa Valley.] Therefore, the derived maturity date for the years 1972 was ‘adjusted day 79’ and 1977 was ‘adjusted day 83’. This method was used to estimate DOY_M in about one-third of the years overall and for half of these, records were <1°Be different from the designated TSS (°Bé).

In the case of the Eden Valley and Rutherglen, diary records were kept of harvest day, TSS level (°Bé) and tonnes harvested for that day (where the grapes were harvested over successive picks), but harvest planning records were not available. For these datasets, the yield weighted average harvest day and TSS level (°Bé) were calculated. The maturity day was adjusted using the average rate of advance calculated for all of the datasets combined [0.8°Bé week⁻¹ (When considering the regions for which we are adjusting the DOY at maturity in this case, they are in the mid-range of climate relative to the other regions in the study. Selection of a mid-range, or average, rate of ripening used to make adjustments therefore seemed sensible.); the average of the ripening rate (°Bé week⁻¹) for all blocks listed in Table 1], with the ‘adjusted day’ estimated to be when the ‘designated TSS (°Bé) level’ would be attained for the particular site. For the Eden Valley site only about 20% of the years had adjustments that involved a greater than 1°Bé shift. For the Rutherglen region, however, the grapes were harvested across a wider range of Baume levels so greater adjustments were necessary; about half the years with adjustments of >1°Bé.

While the harvest date was recorded for McLaren Vale Shiraz, TSS was only noted through the latter part of the time series (1990–2009). Owing to the extremely long time series of harvest dates available for the McLaren Vale Shiraz block (1895–2009), however, we elected to present the ‘harvest date’ data alongside the DOY_M data of all of other blocks in this study. The winery owner stated that approximately 12°Bé was the targeted level of ripeness at harvest commencement throughout the history of the site. Adjustments were made for the 1990–2009 series where necessary using the average rate of 0.8°Bé week⁻¹ (as above). We emphasize however, ‘the maturity date’ was likely to vary from the ‘harvest date’ if grapes were harvested at a different sugar concentration each year.

Trend in maturity day

Trend analysis of DOY_M of all of the time series has been undertaken for:

Linear regression against time is a long established statistical tool widely used in phenological research to study trends (Sparks & Tryjanowski, 2010) and has been applied in this analysis. Other regression models were also tested to determine the presence of higher order relationships.

Calculation of length of harvest period

If DOY_M is more advanced in later ripening varieties than earlier ripening varieties a compression of the harvest period may occur i.e. the time from the beginning to the completion of harvest is reduced. To determine whether this occurred, the range of days from the first DOY_M and the final DOY_M across varieties was assessed for all the blocks in each region, and overall regions, over comparable parts of the time series. These time periods are 1985–2009 inclusive [10 blocks (Central Victoria (Cr2) block was omitted from this grouping as 2008 and 2009 data are missing.) from six regions] and 1993–2009 inclusive [30 blocks (McLaren Vale (Bl6), Eden Valley (L) and Central Victoria (Cr2), Coonawarra (Co2, Co1, Co3), Eden Valley (E1, E2, E3), Clare Valley (Cl1, Cl2), Loxton (R2), Waikerie (R5, R11) are omitted as the vintage year 1993 was not represented in these blocks.) from nine regions] were included. Blocks with missing data at the beginning and end of the series were omitted to ensure a consistent time frame was used (see Table 1).

Climate data

Daily maximum and minimum temperature data from 1911–2009 were obtained from 0.05°× 0.05° gridded data produced for the Australian continent by Jones et al. (2009). These daily surfaces were calculated using Australian Bureau of Meteorology's network of high quality temperature stations. Data were extracted from the daily temperature surfaces corresponding to the latitude/longitude location of the site (Table 1). Average growing season temperature from October to April (°C) [an average of the mean monthly temperatures (monthly maximum temperature+monthly minimum temperature)/2, from October to April inclusive], hereafter abbreviated to Average GST (°C), was selected as a temperature measure for comparison with trends in DOY_M. This temperature measure has been found to be most closely associated with the timing of phenological phases of véraison and harvest in similar studies (Jones et al., 2005). A temperature time series of Average GST (°C) were calculated for each year corresponding with vineyard observations at each location. Linear trends in the temperature measure were calculated for three time periods: period of vineyard observations; 1985–2009; and 1993–2009, as with DOYM described previously.

Average GST (°C) was also calculated for the climatological baseline period of 1976–2005 for each region to illustrate the range of climates and relative differences between the regions (Table 1).

Trend in maturity day

Time series of maturity date for the total period of observation are shown for six blocks from four selected sites (Fig. 3), with calculated trends also shown for all sites (Table 2). For two blocks from the Central Victorian region, Marsanne and Shiraz (M) (Fig. 3a), a trend to earlier maturity is evident through the time series as indicated by the slope of the fitted linear regression lines (also shown in Table 2). The shift in average DOY_M across this time series of about 24 days is observed for these blocks over this period, about 0.3 days earlier per year. A similar response was observed in the Muscat à Petit Grains block in the Rutherglen region, though this trend is not significant (Fig. 3b). Compared with these, the Mornington Peninsula blocks display a significant trend to earlier ripening at a faster rate (1.5 days yr⁻¹), though this trend was noted through a shorter and later period of observation (Fig. 3c). The Eden Valley sites have lower (nonsignificant) rates of change to maturity (Table 2). It is interesting to note that in the Margaret River, Cabernet Sauvignon site, the best-fit linear regression line indicates grapes were ripening slightly later in the year over the time series, and while this is not statistically significant, this was the only block showing this trend direction (Fig. 3d).

For the McLaren Vale site, TSS level was available only from 1990, but harvest date data extended back to 1895. Trends in both harvest date and estimated DOY_M are shown (Fig. 4). For both series a trend to earlier maturity is evident, although it is also clear that after 1990 grapes were harvested at a higher sugar level than 12°Bé (i.e. later in the year). The trend in DOY_M, where sugar levels were recorded (1990–2009) was about 1.3 days yr⁻¹ earlier for this block (not shown in Table 2).

It is apparent that, with the exception of the Margaret River Cabernet Sauvignon block, a trend to earlier maturity for all of the blocks and through all of the time series has occurred. The trend to earlier maturity is statistically significant for six out of the 11 long-term blocks for the time period for which records were available. For the period 1993–2009, 35 of the 44 vineyard blocks assessed have a statistically significant trend to earlier maturity (Table 2).

A cross-regional trend comparison is possible where consistent time frames, 1985–2009 and 1993–2009, are imposed on the data (Table 2). For the 1985–2009 period the trend in the DOY_M ranges from about 1.5 days yr⁻¹ earlier in the Mornington Peninsula, about 1 day yr⁻¹ earlier in Central Victoria, to 0.1 days earlier in Margaret River. The average shift in DOY_M was 0.8 days yr⁻¹ earlier across the 1985–2009 period. The average shift overall the long-term blocks for the time-period of 1993–2009 for was 1.6 days yr⁻¹ earlier, about twice the rate compared with the period 1985–2009 for the same long-term datasets.

This accelerating rate of change was observed across all the long-term datasets where earlier maturity was detected (Table 2) and is illustrated by the comparison of linear trends, fitted for the three time periods for the Marsanne block (Central Victoria) (Fig. 5). In this case, for the observed series (1939–2009) the trend of about 0.3 days yr⁻¹ earlier is found, whereas for the period 1985–2009 the rate increased to 0.8 days yr⁻¹ and finally over the 1993–2009 period, grapes were ripening about 1.4 days yr⁻¹ earlier for this variety in Central Victoria. Because the pattern of the change in the ripening rate may indicate a nonlinear trend, all the long-term datasets were fitted with a quadratic model. However, no improvement to the fit of the data was detected by using the higher order relationship.

Winegrape harvest compression analysis

For the period, 1993–2009, the average range of DOY_M over the 30 blocks was 63 days (from 46 to 79 days) (Fig. 6), with individual vineyards DOY_M period ranging from 24 to 120 days depending on the region and year (not shown). Analysis of the trend in the range of DOY_M for the 30 blocks over this period indicated a compression of 1.1 days yr⁻¹ overall of the regions through this period (P=0.02).

For the 10 blocks, representing the period 1985–2009 the average maturity day range was 55 days (34–84 days), with individual regional minimum of 38 days, and maximum of 135 days. While it was evident that the range of maturity days, from earliest to latest observed, was decreasing at about 0.4 days yr⁻¹ through the time period 1985–2009, this trend was not significant (P=0.37) (data not shown). From this evidence it appears that, to date, there has been no significant trend in compression of the harvest period from 1985 to 2009.

This trend in compression was also studied on a regional basis where more than one block had been assessed in the region. A significant trend to a compression in DOY_M was found for the six Riverland blocks (1.3 days yr⁻¹; P=0.02) and also the three Central Victorian blocks (0.6 days yr⁻¹; P=0.01). There was a nonsignificant trend in compression of maturity evident for the 10 Barossa Valley blocks (0.4 days yr⁻¹; P=0.47) and the five Langhorne Creek blocks (0.1 days yr⁻¹; P=0.9). In the Mornington Peninsula, no compression of maturity was evident. While not exhaustive, this analysis suggests that compression of the harvest period was likely to be site specific and dependent on the varietal mix.

Average growing season temperature trend in relation to the maturity trend

Average growing season temperature for all but one region (Margaret River) has been warming (Table 3), the rate varying by region and by time period, with the more recent period warming faster than over the longer term. When comparing the same period of time 1985–2009 or 1993–2009, it was generally the inland regions that were warming at a faster rate compared with the more coastal sites (Table 3).

In all regions where a warming trend in the Average GST ( °C) was observed, maturity was trending earlier. It was interesting to note that, where there has been a cooling trend in Average GST ( °C) for the period of observation, this has been associated with a trend to later maturity (Margaret River), though the trends were not significant.