Changes in the incidence and onset of potato late-blight epidemics in Finland were investigated and compared with possible changes in climate, presence of soil-borne inoculum, and aggressiveness of Phytophthora infestans populations. Datasets were constructed from leaf blight assessments in cultivar trials or fungicide tests carried out at eight experimental sites during the periods 1933–1962 and 1983–2002. Additional data were obtained from late-blight monitoring projects carried out from 1991 to 2002. From 1998 to 2002, the risk of blight outbreak was 17-fold higher compared with the periods 1933–62 and 1983–1997. Simultaneously, the outbreaks of the epidemics began 2–4 weeks earlier. The changes observed were associated with a climate more conducive to blight in the late 1990s. Lack of rotation also advanced blight epidemics by an average of 9 days in 1998–2002, but it did not have this effect in 1992–1997, suggesting that soil borne inoculum may not have been a significant threat to potato until the late 1990s. The aggressiveness of the P. infestans isolates seemed to have only minor effect on the onset of the epidemics after 1991, as the apparent infection rate remained unchanged despite weather conditions more favourable to late blight in the late 1990s. As a consequence of the more frequent and earlier epidemics, the sales of fungicides used against late blight in Finland increased 4-fold from the 1980s to 2002.
Potato late-blight caused by the oomycete Phytophthora infestans is one of the most destructive potato diseases worldwide (Fry et al., 1993). An unprotected potato crop can be completely destroyed before tuber formation, resulting in total crop loss. Yearly variations in yield losses can be huge and highly dependent on meteorological factors influencing both incidence and severity of leaf and tuber blight (Crosier, 1934; Hirst & Steadman, 1960a; Croxall & Smith, 1976; Zwankhuizen & Zadoks, 2002).
Potato late-blight was introduced into Europe during the 1840s (Fry et al., 1993). In Finland, late-blight epidemics have been reported in newspapers since 1847 (Mäkelä, 1966). Prior to the 1970s, the disease caused considerable yield losses in southern Finland – on average in one year in eight. However, the disease hardly ever appeared in northern parts of Finland (beyond 65°N) before frost had destroyed the haulm (Seppänen, 1971). Chemical control of potato late-blight was rare until the 1970s, from when it was controlled by one to three applications per season of dithiocarbamates or, after 1985, metalaxyl (Seppänen, 1987). After the appearance of resistant isolates to metalaxyl (Hermansen et al., 2000), several new compounds for late-blight control have been introduced since the 1990s.
Besides structural changes in populations of P. infestans the climate has changed. According to Baker et al. (2004), climatic data indicate that the risk of late blight in the upper Great Lakes region in the USA has increased from 1948 to 1999. Climatic observations in Finland in the 20th century and especially since 1980 show increasingly warm springs (April and May), slightly warmer summers (June to August), and a reduced diurnal range of temperatures (Tuomenvirta & Heino, 1996; Tuomenvirta et al., 2000; Tuomenvirta, 2004). In Scandinavia and Finland, the reduced range is related to increased cloudiness and to the strengthening of westerly air flow (Tuomenvirta et al., 2000). This climatic trend means more favourable weather for potatoes in the early growing season, but also weather that is more conducive to late blight. Studies based on simulation models indicate that an increase in mean temperature in southern Finland of 1°C will extend the period when chemical control for late blight is necessary by 10–20 days if soil-borne inoculum is absent (Kaukoranta, 1996).
Newly emerged populations of P. infestans are more aggressive on potato than the old clonal lineage, which is likely to have contributed to the population displacement (Carlisle et al., 2002; Shattock, 2002). Zwankhuizen & Zadoks (2002) have found that factors enhancing late-blight intensity included favourable weather and sources of inoculum, but they were unable to explain the long-term pattern of blight epidemics that emerged in the Netherlands. In the present study, the primary aim was to quantify incidence and earliness of late-blight epidemics over the years and determine the factors associated with the increase in the disease. Five questions were addressed: i) has the onset of epidemics become systematically earlier and their incidence more frequent; ii) have temperature and precipitation regimes become systematically more favourable to occurrence of late blight; iii) are early season outbreaks of disease associated with short crop rotations and thus probably with oospore-derived infections; iv) has the rate of epidemic development changed in a systematic way; and v) have blight management practices, especially fungicide-use, changed?
Materials and methods
Observation sites and experiments
The datasets for the study were constructed from potato cultivar and fungicide efficacy trials at eight experimental sites of MTT Agrifood Research Finland and the Potato Research Institute in the years 1933–1962 and 1983–2002 (Table 1). In addition, a potato late-blight monitoring project carried out from 1992 to 2002 (originally part of a potato late-blight forecast and warning service) produced data on disease incidence on farms and in home gardens within a radius of 50 km from eight fixed observation sites.
Table 1. The earliest and the latest planting dates, the mean temperatures and the number of rain days (precipitation at least 0·1 mm) from June to August at eight fixed monitoring sites in Finland from 1983 to 2002 and type of information available on potato late-blight epidemics in the periods 1933–1962, 1983–2002, and 1993–2002
Earliest and the latest planting date in 1983–2002
From June to August in 1983–2002
Observations and measurements available in datasets A to Ea
Location of the fixed monitoring sites
Mean temperature (°C)
Mean number of rain days
i = disease incidence; f = first blight observation days after planting; w = weather data; v = microscopic verification of Phytophthora infestans; e = epidemic development (apparent infection rate); p = preceding crop.
25 May–14 June
18 May–10 June
23 May–9 June
18 May–9 June
7 May–11 June
15 May–5 June
Pälkäne or Lammi
8 May–30 May
11 May–13 June
The fixed observation sites represent a wide range of climatic and agricultural environments between the latitudes of 60° and 66°N and the elevations of 26 and 106 m above sea level. Weather stations of the Finnish Meteorological Institute situated at the experimental sites provided accurate meteorological measurements for the study. Kokemäki and Ylistaro are surrounded by very intensive starch and table potato production characterized by very short or absent crop rotation. Pälkäne and Lammi regions produce table and organic potatoes on a relatively large scale. There is no commercial potato production near Jokioinen. Ruukki is located between Finland's high-grade seed potato production area and an area where potatoes for table consumption and the processing industry are produced very intensively under short crop rotation. Mikkeli and Maaninka represent small-scale potato production. Potato fields are tiny, 0·1 to 1·0 ha, scattered, and separated by forests. Rovaniemi is a very extreme site characterized by a growth season of 2–3 months with 24 h of daylight. Potatoes there are grown for local consumption.
Dataset A was compiled by Seppänen (1971). It includes the dates of the first observations of potato late-blight symptoms in the years 1933–1962 at eight experimental stations of MTT Agrifood Research Finland (Table 1). The observations originate from unprotected plots of the old cv. Rosafolia, which is as susceptible to late blight as the cv. Bintje in the later datasets.
Dataset B consists of the assessments of leaf blight outbreaks in unprotected plots of cv. Bintje in cultivar and fungicide efficacy trials carried out 1983–1991 at the same fixed experimental sites as the trials in dataset A. Two additional sites, Mikkeli and Kokemäki, were included, and Pälkäne was replaced by trials carried out at the Potato Research Institute in Lammi, 15 km southeast of Pälkäne (Table 1). The data were collected from the original datasheets in the archives of each experimental site. If more than one trial contained untreated plots of cv. Bintje, the earliest late-blight observation of the year was included in the dataset.
Dataset C was produced during late-blight monitoring and warning projects carried out 1992–2002. At each of the eight experimental sites, there were untreated trap plots of cv. Bintje that were monitored two to three times a week for the appearance of potato late-blight lesions. The leaflets containing lesions were sent to MTT, Plant Protection, where the presence of P. infestans was verified with microscopic observations.
For dataset D, blight onset was monitored from 1992 to 2002 on farms and in home gardens within a 50-km radius from the fixed observation sites. The observations were verified similarly to those in dataset C. Annually, five to 20 fields were monitored for outbreak of disease.
Dataset E is composed of leaf-blight progress assessments at unprotected cv. Bintje plots examined from 1991 to 2002. Each year 3–6 field trials were monitored at Jokioinen and Lammi. Disease severity was assessed as the percentage of the total affected leaf area in a plot. Only experiments in which at least five consecutive disease ratings were carried out were included in the data. A total of 48 experiments were included in the dataset.
Incidence and first outbreaks of late-blight epidemics
Logistic regression (Allison, 1999) was used to compare late-blight incidence between observation periods and sites in data combined from sets A, B and C. Utilising the same data, mixed models (Littell et al., 1996) were used to compare the onset of late-blight epidemics days after planting (DAP) between the observation periods and sites. DAP was used because it relates the onset of the epidemics in the stage of development of the potato crop better than the calendar date in Finnish growth conditions. Disease onset data from dataset C were plotted against dataset D to see if the trend observed in the onset of late-blight epidemics was similar in the experimental plots and in the fields in surrounding regions.
Effect of climate change on the onset of epidemics
Onset data from 1983 to 2002 at eight observation sites (datasets B and C) were divided into two equal time spans, the period 1983–1992 and the period 1993–2002, both comprising 80 observations. Daily weather data (temperature and precipitation) from the first period, obtained from the Finnish Meteorological Institute, were analysed by several regression methods for predicting the onset of late-blight epidemics. The estimated regression parameters were applied to see if climate change provides explanation for the changes observed in the onset of epidemics in the period 1993–2002.
Crop rotation and late-blight outbreaks
Two categories of crop rotation were constructed from data sets C, D and E; potatoes grown versus potatoes not grown in the preceding season. Difference in onset time of the epidemics between these two categories was analysed separately for periods 1992–1997 and 1998–2002 using Mixed Models (Littell et al., 1996). The type of crop rotation and the two monitoring periods were set as fixed effects and the individual years and sites were considered random effects.
Changes in epidemic development
The course of the epidemic in dataset E was classified into three phases: i) the symptomless period from planting to the appearance of the first symptoms (DAP), ii) the period from the appearance of symptoms to 5% leaf area blighted, and iii) approximately linear disease progress with 5–95% leaf area blighted. The variables used in the analysis were the duration of each of the three phases of the epidemic, final blighted leaf area and apparent infection rate (AIR) (Fry, 1977). AIR was calculated for the interval between 1% and 99% disease level. When the 99% level was not reached the AIR was not calculated. The differences of means of all variables between the observation years and sites were respectively analysed using mixed models (Littell et al.,1996). Observation year and site were regarded as fixed parameters and individual experiment and replicate as random parameters. The experimental years 1991–1995 and 1996–2002 were respectively grouped together for the final analysis.
The effect of the change in the onset of the epidemics on fungicide use
The total potato area capable of being treated once with a maximum recommended dose of fungicides was calculated from the annual sales of fungicides (data set F). The annual number of fungicide applications was calculated by dividing the potentially fungicide-treated potato area by the total potato production area of the year in question. Copper products were excluded, because they are also used for other crops and their sales volume has been much lower than that of other fungicides.
Statistical analyses were carried out by Proc Logistic, Proc Reg, Proc Lifereg, and Proc Mixed procedures available in SAS/STAT 9·1 (SAS Institute Inc., Cary, NC, USA).
Incidence and first outbreaks of late-blight epidemics
In the historical reference period 1933–1962, late blight was present on average every second year. In the 1980s and up to the year 1997, the average for late-blight attacks on potato crops was two years out of three. During the last pentad of the survey, 1998–2002, late blight occurred in four to five years out of five, depending on the observation site (Table 2). In terms of odds ratios, the risk of disease in the period 1998–2002 was 17-fold greater than for the period 1933–1962 (P= 0·0004). For all periods studied, the risk of the late blight was 5-fold greater in southern Finland compared with the northern parts of the country, though the differences decreased towards the end of the observation period.
Table 2. Incidence of potato late-blight in the periods 1933–62, 1983–92, 1993–97, and 1998–2002 at monitoring sites in Finland as the percentage of seasons in which disease was present
% of seasons when blight was present
Odds ratio for incidence of blight
95% Wald confidence limits
Not included in the means or statistical analysis.
nd = no data available.
ref = reference against which odds ratio is estimated.
The first disease outbreaks were recorded on an average 83 DAP during the monitoring periods 1933–1962 and 1983–1993. The outbreaks from 1993 to 1997 took place on an average 78 DAP, while those from 1998 to 2002 only averaged 57 DAP (Fig. 1). During the last 5 years, late blight appeared on crops on an average 24 days earlier than during the other pentads. The shift towards an earlier attack was greatest (43 days) in Mikkeli, whereas the shift was smallest (15 days) at the northernmost site, Rovaniemi. The trend towards earlier epidemics was also obvious in untreated potato fields within 50 km of the eight fixed observation sites. The observations from the potato fields are in good agreement with those from the fixed monitoring sites, with a very rapid shift towards earlier epidemics between the years 1996 and 1998 (Fig. 2). Before 1996, the onset of epidemics fluctuated between 60 and 80 DAP, and it dropped to values of 18–60 DAP after 1998.
Climate change and the onset of epidemics
Daily weather data for the 40 DAP in the years 1983–1992 were used to estimate regression parameters for predicting blight onset DAP, and the regression model was applied to examine the possible role of climate change in blight onset during the period 1993–2002. The reason for including weather data for only 40 days of after planting was that the earliest blight outbreaks of disease during the period 1993–2002 were observed 18, 35, 36, 38, and 40 DAP. Linear regression using either third degree polynomial terms (SAS Proc Reg) or life regression (SAS Proc Lifereg) were first used to develop a regression model for the period 1983–1992. Both regression methods predicted qualitatively well (early versus late onset) but under-predicted very late onsets, whose weight in turn forced early onsets to be predicted slightly too late. Obviously, this was because weather later than 40 DAP could not be taken into account in the years when the weather was not conducive to late blight. As only qualitative prediction was achievable, the logistic regression (SAS Proc Logistic) was used to predict early (77 DAP) or late onset of disease (>77 DAP). The cut-off point was chosen at 77 days because 40 days of weather data did not allow classification of onset dates using a much later cut-off point.
Regression analysis and the inspection of data showed that the onset dates observed at Kokemäki were systematically about 10 days earlier than elsewhere in southern Finland, without any obvious weather-related cause. It was concluded that the systematic difference stems either from observation methods or from the high percentage of cultivated area that has been under potato production around the site. In the life regression and the logistic regression, Kokemäki versus other observation sites was treated as a class variable. Two cases of extremely early onset dates between 1983 and 1992, which did not seem to be related to weather, were taken to be outliers and removed from the data. Different periods of weather data, variables (mean minimum temperatures during 5, 7, or 10 DAP; mean temperatures; temperature ranges; precipitation sums; number of rain days) and combinations of the variables were tried. The model that best assigned cases to the correct category (onset at day77 or >77 DAP) and was biologically meaningful was Logit(P) = intercept +B′[T3 T3P3 P3 T4P4 P4], where T3 and T4 are the mean temperatures during 21 to 30 and 31 to 40 DAP respectively, P3 and P4 are the number of rain days (at least 0·1 mm) 21 to 30 and 31 to 40 DAP respectively, and B is the vector of slope parameters b0 to b4. (Table 3). This model correctly assigned 89% of 78 cases between 1983 and 1992.
Table 3. Modelling probability (P) that potato late-blight onset is later than 77 days after planting. Logit (P) = intercept + B′X, where B is the vector of slope parameters b0 to b4. Site class is a classification variable with two values, Kokemäki or other observation site
R2= 0·3889; max-rescaled R2= 0·5800.
Both the observed and the predicted early onset of disease have become more common in the 1990s. Since the late 1990s, the early onset has been observed more often than predicted (Fig. 3), suggesting that an additional factor besides weather favoured early onset of epidemics in the late 1990s. The predicted onset corresponds to the almost continuous increase in the frequency of rain (Fig. 4b), at the same time as temperatures in the early season have fluctuated upwards (Fig. 4a) after a slump in the early 1990s.
Crop rotation and late-blight outbreaks
Data on cropping history available from 235 fields and from experiments carried out during the years 1992–2002 made it possible to distinguish categories of crop rotation: namely potatoes grown or not grown in the preceding season. The data analysis revealed that cropping history did not have any effect on the onset of the epidemics from 1992 to 1997 (Table 4). From 1998 to 2002, the epidemics started on average 9 days earlier if the preceding crop was potato instead of other crops, which suggests the appearance of a soil-borne primary inoculum source. There was a clear shift towards more potato-intensive production; from 1992 to 1997, potatoes were grown after a preceding potato crop at 42% of the monitored fields, and this figure was 63% in the years 1998–2002.
Table 4. Effect of potato as preceding crop on onset time of potato late-blight epidemics in fields in Finland, from 1992 to 2002
First observation of late blight, days after planting
A major change observed in epidemics was that these began 15 days earlier on average from 1996 to 2002 than from 1991 to 1995 (Table 5). In all years from 1991 to 1995 except 1993, late blight was found statistically significantly later than in 2002 (P 0·0001) (Fig. 5a). In this respect, however, the years 1996–2001 did not differ from 2002 (P= 0·036–0·808). In addition, the final disease rating at the end of the experiments had increased from a mean defoliation of 70% in the years 1991–1995 to 99% in the years 1996–2002. The lower disease ratings from 1991 to 1995 are due to the fact that epidemics in those years started in the second half of August and crop was usually either harvested or defoliated by frost before an epidemic reached a defoliation level of 100%. Once an epidemic had started, there was no significant difference between the monitoring periods or individual years in the rate of disease development from the first symptoms to 5% defoliation or from 5% to 95% defoliation. The values or AIR remained unchanged during the period 1991–2002 (Fig. 5b).
Table 5. Epidemiological measurements associated with the rate of development of potato late-blight epidemics from 1991 to 2002 in field trials in Finland
Estimate for difference b-a
P for b-a
Number of trials
Days from planting to first symptoms
Days from first symptoms to 5% leaf area blighted
Days from 5% to 95% leaf area blighted
Apparent infection rate
Final leaf area blighted (%)
The effect of the change in the onset of epidemics on fungicide use
The statistics on the sales of fungicides against potato late-blight showed increasing fungicide use over the period when the onset of disease was advanced. In the 1980s, the amount of fungicides sold annually would have been enough to spray 80% of the total area of potato production once only. On the basis of the sales at the end of the 1990s, the area of potato acreage in Finland (approximately 30 000 ha) could have been sprayed four times (Fig. 6a). Since organic farming and production for personal consumption are also included in the total production area, the actual number of sprays on professional potato farms is somewhat higher than the estimated figure. The highest increase in fungicide sales took place simultaneously with the observed shift towards earlier epidemics in the years 1996–1998.
The most widely used active ingredients for potato late-blight control have been mancozeb and fluazinam. Fluazinam rapidly replaced maneb (Fig. 6b). In consequence, the number of sprays probably decreased because the spraying interval of fluazinam (10–14 days) is considerably longer than that of dithiocarbamates (7–10 days). Due to the former's excellent rainfastness, there was also no need to respray the crop after rain, as opposed to the situation with dithiocarbamates (Kudsk & Mathiassen, 1991). Therefore, the actual increase in fungicide use against late blight due to earlier epidemics in the late 1990s is underestimated in calculations based on official statistics.
Analysis of data from fixed sites and actual potato fields as well as from field experiments shows that late-blight epidemics were likely to start 2 to 4 weeks earlier in the years 1996–2002 than during the historical period 1933–1962 reviewed by Seppänen (1971). During the 1990s, the shift towards earlier epidemics was rapid and took place earlier in southern than in northern Finland. Late-blight also invaded areas north of latitude 65°N, where the disease had been practically unknown until the late 1990s.
Before the 1990s, disease onset and epidemic development in Finland were very similar to those in other European countries (Croxall & Smith, 1976; Kolbe, 1982; Kluge & Gutsche, 1990; Zwankhuizen & Zadoks, 2002). Schepers (2004) has collected data on the first outbreaks in most European countries in the years 1999–2003. During this period, there was great temporal variation in late-blight onset without any obvious trend. According to Zwankhuizen & Zadoks (2002), late blight was almost absent from the Netherlands in the 1970s and early 1980s, and although blight in the Netherlands became a serious problem again during the 1980s, the authors did not report any systematic trend towards earlier epidemics.
Prevailing weather conditions during the growing season determine the incidence and severity of late-blight epidemics (e.g. Crosier, 1934; Hirst & Stedman 1960b; Kluge & Gutsche, 1990). Therefore, it was appropriate to assume that the increasing likelihood of early epidemics in Finland during the 1990s could be explained by the warming of the climate (Tuomenvirta et al., 2000; Tuomenvirta, 2004), as indicated by Kaukoranta (1996). The analysis of the data collected at eight observation sites between 1983 and 2002 supports the hypothesis that the increased frequency of rain and higher early-season temperature has affected the onset time of the epidemics. Yet, in the regression models developed in this study, weather parameters could only partly explain the earlier outbreaks of late blight.
In addition to more favourable weather parameters, the shift towards earlier outbreaks of late blight coincides with increasing evidence of oospores as a new primary source of inoculum (Anderson et al., 1998; Lehtinen & Hannukkala, 2004). It was evident that the risk of early attacks of late blight increased significantly in fields where potatoes were also grown in the preceding year in comparison to fields with a break of one year or longer in potato production. During the years 1998–2002, there was an increase in the proportion of fields where potatoes were grown after a preceding potato crop, which emphasizes the role of soil-borne inoculum. Bødker et al. (1998) have reported earlier blight outbreaks in Danish potato fields where potatoes had been grown during one of two previous years in comparison to fields with a break of at least 3 years in potato production. The preceding potato crop did not pose a threat to a subsequent potato crop before the shift towards earlier outbreaks. This suggests that the increased threat posed by the preceding potato crop was due to oospores, or an increased impact of volunteer plants on the onset of epidemics. There is no evidence, however, that conditions have become more favourable to the survival of tubers (giving rise to volunteer plants) in soil after the shift towards earlier outbreaks. However, some of the present P. infestans strains might be more capable of causing early epidemics as variation in aggressiveness to tubers has increased (Lambert & Currier, 1997; Flier et al., 1998).
Despite the higher aggressiveness of the new P. infestans population (Shattock, 2002; Carlisle et al., 2002), Zwankhuizen & Zadoks (2002) did not find evidence of increased severity of the late-blight epidemics in the Netherlands. However, they found that the intensity of epidemics was less predictable in the years 1980–1996 than before. The present analysis of disease progression failed to find changes in the apparent infection rate during the period 1991–2002, which saw an increase in the disease outbreaks and a shift towards their earlier onset. This indicates that variables like infection efficiency, length of latent period, and sporulation efficiency did not change sufficiently to have a significant effect on the epidemic rate.
Due to the earlier epidemics in the 1990s, the use of late-blight fungicides has dramatically increased. In practice, an epidemic onset that is 2–4 weeks earlier means two to four additional sprays to achieve sufficient control of late blight. Accordingly, 1·5 more sprays were applied at the end of the 1990s than at the beginning of the decade. The difference would have been greater without the introduction of fluazinam with its longer spray interval and better rainfastness than those of dithiocarbamate products. The increased use of fungicides might also be partly due to a more professional attitude towards late-blight control resulting from the steady increase in the potato farming area per grower.
It is a challenge for future potato late-blight research to find a balance between the public demands for a reduced use of pesticides (Schepers, 2005) and the pressure to increase pesticide use due to changes in climate and challenging pathogen populations. Early attacks require the alertness of growers, but there is currently no warning system capable of predicting oospore-derived late-blight infections. Growers can reduce the risk of an early attack by improving crop rotation, but there is an urgent need to study the factors determining the onset of oospore-derived epidemics and to implement the knowledge in decision support systems for late-blight control.
The authors wish to thank the personnel of the South Ostrobothnia, North Ostrobothnia, and Lapland Research Stations, especially Arjo Kangas, Elina Virtanen, and Antti Hannukkala, for providing blight monitoring and experimental data. We are grateful to Ms Tuija Vihervirta, Ms Mari Helminen, and Ms Tuula Viljanen for technical assistance and Mr. Hannu Ojanen for expertise in geographical information. The Finnish Ministry of Forestry and Agriculture has provided funding for the blight monitoring projects at several stages during the course of the study.