SIMCOL – A Decision Support System for integrated control of anthracnose on blue lupin

SIMCOL – Ein Entscheidungshilfesystem zur integrierten Bekämpfung der Anthraknose an Blauer Lupine

During warm and wet periods, conidia from acervuli and micro-sclerotia are splashed by rain from diseased to healthy plants. The conidia germinate to form appressoria on plant surfaces, from which penetration hyphae develop into plant cells. Infection may occur through almost any plant surface. Under suitable conditions, the fungus can grow rapidly inside the plant and cause severe symptoms very quickly. Once the fungus has developed sufficiently inside the plant, dark fruit-bodies are produced, causing typical anthracnose symptoms. A pale pinkish or sometimes orange spore mass develops within the lesions. The spores are normally dispersed by water splash (Agrios, 2005; Davidson et al., 2005).

Warm and humid weather (Römer, 2007), rainfall and prolonged leaf wetness lead to a strong and rapid spread of the infestation (Thomas and Sweetingham, 2004). The disease affects all parts of the lupin plant but the inflorescences and pods are more susceptible than the young plants (Davidson et al., 2005; Klocke and Nirenberg, 2007).

Some fungicide experiments demonstrate the impossibility to stop a Colletotrichum epidemic under way. As a consequence an early infection-oriented use of fungicides may be the correct strategy and suggests the development of a prognosis model (Klocke and Nirenberg, 2007; Nirenberg, 1999).

For this reason SIMCOL (SIMulation model for COL­letotrichum), a modular, weather-based decision support system (DSS) was developed in Germany by the Central Institution for Decision Support Systems in Crop Protection (acronym ZEPP). The DSS should be used by the Governmental Crop protection services as well as in seed production.

The DSS SIMCOL, according to the ZEPP nomenclature (Racca et al., 2010), consists of three prognosis models:

Where: EPI = daily epidemic pressure index DE = disease efficiency L = latency period (1 = start, n = end)

DE is dependent on temperature, leaf wetness and growth stages of lupin. During a leaf wetness period, the DE is calculated as follows:

Where: Tw = average temperature during the leaf wetness period (°C) LWD = leaf wetness duration in hours BBCH_S = BBCH-dependent susceptibility stage (from SIMONTO-Lupin)

The influence of temperature and leaf wetness on the DE was assessed by interpolating the values of the maximum disease incidence recorded in the trials using a combination of Richards function (Richards, 1959) and modified beta-Hau function (Hau, 1988) [3] for each of the 4 different BBCH-susceptible stages (21–23 (2–4 leaves), 39 (12 leaves), 60–69 (flowering) and 70–87 (pods development)).

Where: DE _{BBCH_S} = disease efficiency depending on Tw and LWD for each of the 4 different BBCH-susceptible stages LWD = leaf wetness duration in hours Tw = average temperature during the leaf wetness period (°C) Y_max = maximum disease incidence T_min, T_max, T_opt = estimated minimum, maximum and opti­mum temperature for the disease efficiency (°C)
a, b and n = equation parameters

The regression parameters are summarised in Table 2 and one example of the DE calculated for the BBCH 21–23 (2–4 leaves) is shown in Fig. 1.

Fig. 1. Modeled Disease efficiency of C. lupini, depending on temperature (°C) and leaf wetness duration (hour) for the BBCH stages 21–23 (2–4 leaves)

Where: L _(1/d) = latency (expressed as a daily rate 1/day) L_max = maximum daily latency rate T = daily mean temperature (°C) T_min, T_max, T_opt = estimated minimum, maximum and opti­mum temperature for the latency period (°C)
n = equation parameters

The parameter values are summarized in Table 3.

The EPI is calculated as the sum of the accumulated daily average values of DE during the latency period. An example of the EPI is shown in Fig. 2.

Fig. 2. Epidemic daily pressure index (EPI) of C. lupini and simulated BBCH-growth stages (with SIMONTO-Lupin model) for the met. station Rommersheim (Rhineland Palatinate, Germany), for the growing season 2009.

After the data analysis the best result (77% of correct prognosis, 23% of belated prognosis) was obtained with the following criteria:

The model simulation begins with the crop sowing date. When one of the above conditions is verified SIMCOL1 gives the disease first appearance date.

The results of the validation are summarised in Fig. 3.

Fig. 3. Comparison of predicted and observed first appearance on the stems, SIMCOL trials in 2009 and 2010 (n = 9)

In 89% of the cases the prognosis of SIMCOL1 was classified as correct, in 11% of the cases too late. In all cases, the prognosis was before the first appearance on the pods.

In Fig. 4 the observed periods of the disease occurrence on the leaves, stems and pods and the forecasted period of SIMCOL1 are shown in a box-plot. The average prognosis date of SIMCOL1 lies between the mean date of occurrence on the leaves and the appearance on the stems. The date of the forecasted first appearance is recom­mended to use as the date for the first treatment or as a starting date for the field control.

Fig. 4. Comparison of predicted and observed first appearance on leaf, stem and pods. SIMCOL trials in 2009 and 2010 (n = 9)

It was further analysed whether differences occur in the date of the first disease occurrence in trials with healthy and infected seed and in treated or untreated seed. However no clear connection could be determined. In the experiments, the first appearance in the untreated and in the treated variant sometimes was on the same day, in other cases there was a difference of up to 17 days or there was no infestation in the treated variant.

As a basis for the development of SIMCOL1 the data were obtained with artificially infected seed. In practice, how­ever, the farmers mostly use certified healthy seed. To take into account a possible infection on the seed, a factor that changes the conditions for the beginning of disease for non-infected seed was introduced in the model. This factor changes the minimum criteria for a disease onset as follows:

This factor must still be verified by future validation experiments.

Typically, a fungicide treatment is necessary if the disease (incidence or severity) increases in an exponential phase. C. lupini is a very dangerous disease; some small lesions can produce in a short time many spores and, under favourable conditions, the epidemics grow up rapidly. Due to this high aggressiveness and the fact that there is no disease threshold for anthracnose available, infection must be avoided (0-tolerance).

In a further step, disease incidence (DI) data from 66 trials (historical field data) were analysed and divided into two classes of disease increase (class 0, no disease increase, DI = 0; and class 1, positive disease increase, DI > 0). For this data at first the latency period was computed and then the EPI was accumulated during the latency period (EPIsum). The data are presented as a box plot in Fig. 5.

Fig. 5. Classification of the EPIsum in the two classes of disease increase (0 and 1)

The disease increase (class 1, DI > 0) is in the range of 0.1 to about 5 of the EPIsum. 50% of the data range from 1.8 to approximately 4. Class 0 (no disease increase, DI = 0) has a low EPIsum of 0.1 to 0.8. To establish a limit of the EPIsum, which discriminates the two classes, the 66 cases were analysed. A good agreement was achieved using a cutoff of EPIsum = 2. With this threshold in 85% of the cases a disease increase was correctly forecasted. In 12% of the cases the disease increase was underestimated (the EPIsum predicts no increase in infection, but an increase in infection was recorded) and in 3% of the cases overestimation (disease increase was forecasted but not recorded) occurred.

This limit of EPIsum = 2 was used as an internal model threshold. This means that if the EPIsum (after a treatment) reaches this threshold, a new treatment recommendation is issued.

The fungicide effect is taken into account in the model as a reduction factor of the epidemic, depending on the effectiveness of the fungicide.

The fungicide effect was calculated as follows:

Where: FE = fungicide effect (as a reducing factor for the disease efficiency) FE_T = fungicide efficacy depending on the temperature and the relative dose FE_D = fungicide efficacy duration depending on the temperature sum D_ppm = relative dose sumT = sum of the temperature starting from the day of the last treatment (°C)

For the development of the fungicide module in SIMCOL laboratory tests were performed, in which the temperature-dependent fungicidal activity against C. lupini was investigated for the active ingredient azoxystrobin (product Amistar) and tebuconazole (product Folicur).

With the data collected by the laboratory tests the FE_T [6] and the FE_D [7] functions were obtained:

Where: FE_T = fungicide efficacy depending on the temperature and the relative dose T = temperature (°C) T_min, T_max, T_opt = estimated minimum, maximum and opti­mum temperature for the fungicide efficacy (°C)
a, b, pD and n = function parameters

Where: FE_D = fungicide efficacy duration depending on the temperature sum sumT = temperature sum from the date of the last treatment (°C) a, b = shape parameters

The parameters and the functions obtained with [6] and [7] are summarised in Table 4, Fig. 6 and Table 5, Fig. 7, respectively.

Fig. 6. Simulated fungicide efficacy depending on the temperature (°C) for the active ingredient azoxystrobin, tebuconazole and a fungicide mixture.

Fig. 7. Simulated fungicide efficacy duration depending on the temperature sum from the date of treatment (for the fungicide mixture the mean values of both fungicides were calculated).

The efficacy of fungicides reduces the disease efficiency (DE) from the day of treatment, depending on the daily mean temperature and the sum of the daily mean temperature from the date of the treatment and is calculated as follows:

Where: DE_R = reduced disease efficiency DE = disease efficiency FE_T = fungicide efficacy depending on the temperature and the relative dose FE_D = fungicide efficacy duration depending on the temperature sum

With this modified DE_R in a further step, a new EPI can be calculated. If the sum of the EPI reaches the internal model threshold (EPIsum = 2) a new treatment is recommended (Fig. 8).

Fig. 8. SIMCOL3 results for met. station Potsdam in 1999, timing of fungicide treatments and the effect of different treatments with the active ingredient azoxystrobin (A) and tebuconazole (B).

Concerning the influence of the seed treatment, it could be demonstrated with historical trial data (17 trials with 124 surveys) that with fungicide treatment on seed further disease development was reduced significantly. The statistical analysis of these results using ANOVA (not published) showed that the reduction of the disease development can be described by a factor of about 0.5. Since there are no other data to assess the effectiveness of seed treatment a reduction factor was introduced in the model, which reduces the DE by 0.5.

Fig. 9. SIMCOL model results in PASO software.

Since the susceptibility of the lupin plants was related to their growing stage, the simple model SIMONTO-lupin that simulates the appearance of the BBCH growth stages, calculated with a temperature growth rate, was developed and integrated in SIMCOL.

Field data were correlated with an epidemic pressure index in order to obtain the prognosis of the disease occurrence (with model SIMCOL1) and a threshold for the treatment scheduling (SIMCOL3).

The basis for both models SIMCOL1 and 3 is the calculated disease efficiency depending on temperature, leaf wetness and BBCH-growth stages. The estimated regression parameters for the calculation of the DE are described in Table 2: the estimated minimum temperature for the development of the disease is 5°C for all tested BBCH-stages. Exceptions are the stages 60–69, in which even at 0°C a development of the disease is possible. This estimated parameter shows no statistical significance (p > 0.05). This means that the experimental data at low temperature used for the estimation of the DE are not satisfying and need further improvement. The estimated maximum temperature for development is approximately 30°C for all stages. The optimum temperatures are around 23, 21, 26 and 22°C for BBCH 21–23, 39, 60–69 and 70–87, respectively. The statistical significance of these parameters (p < 0.05) shows that the maximum and optimum temperature were estimated correctly.

To get a minimum DE for stages 21–23, 39 and 60–69, a minimum of 4 hours of leaf wetness duration is required. For stages 70–87 only two hours are sufficient. After 7–8 hours of leaf wetness already high values of the DE for the BBCH stages 21–23, 39 and 60–69 are reached. At BBCH 70–87 4 hours of leaf wetness are sufficient to obtain the maximum DE. This confirms the pathogen´s aggressiveness and the higher susceptibility of the plants during the pods development stage (70–87).

The aggressiveness of the pathogen can also be confirmed by the very short latency period which is only about 4 days at the optimum temperature of 25°C.

The excellent quality of the calculated DE, the latency period and the EPI (combination of the two former para­meters) was confirmed by showing correct prognosis of the disease appearance (89% of correct cases) in SIMCOL1 and the disease increase (85% of correct cases) in SIMCOL3.

Laboratory experiments were carried out to obtain data for the modelling of the fungicide efficacy. Fungicide effi­cacy functions depending on temperature where calculated for two active ingredients and introduced in the model SIMCOL3 to obtain the possibility to forecast the dates of the subsequent treatments. The results of the laboratory experiments demonstrate an increased efficacy of the a.i. tebuconazole compared to the a.i. azoxystrobin (Fig. 6). The optimum temperature for the higher efficacy was about 19.8 for azoxystrobin and 17.3°C for tebuconazole (Table 4).

The model also includes a factor which takes into account the influence of the seed treatment (DE reduced mean a 0.5 factor) and the seed infestation on the disease development (changes of the minimum criteria for a disease onset in SIMCOL1). Both factors needed further evaluation.

The validation of SIMONTO-Lupin (Racca and Tschöpe, 2011; Tschöpe and Racca, 2010) and SIMCOL1 with independent data obtained from the Governmental Crop protection services of Brandenburg, Mecklenburg-Western Pomerania and Saxony-Anhalt, the Saatzucht Steinach, the Julius-Kühn-Institut and the ZEPP has delivered good results. A further and appropriate validation of the model SIMCOL3 must still be done. Some parameters like the efficacy of the seed treatment and the influence of the infestation level on the seeds need some further improvement.

The DSS SIMCOL was programmed and included in the software package PASO and can be used to optimise the control strategy against the anthracnose in blue lupin both for the agricultural practice through the advisory and for the seed production. Moreover SIMCOL helps in planning field assessments and fungicide treatments thus saving costs for crop protection products.