Wheat leaf photosynthesis loss due to leaf rust, with respect to lesion development and leaf nitrogen status

Authors


Author for correspondence: Corinne Robert Tel: +33 1 30 81 55 46 Fax: +33 1 30 81 55 63 Email: robert@bcgn.grignon.inra.fr

Summary

  • • In wheat (Triticum aestivum cv. Soissons) plants grown under three different fertilisation treatments, we quantified the effect of leaf rust (Puccinia triticina) on flag leaf photosynthesis during the whole sporulation period.
  • • Bastiaans’ model: Y = (1 − x)β was used to characterize the relationship between relative leaf photosynthesis (Y) and disease severity (x). The evolution of the different types of symptoms induced by the pathogen (sporulating, chlorotic and necrosed tissues) was evaluated using image analysis.
  • • The β-values varied from 2 to 11, 1.4–2, and 0.8–1 during the sporulation period, when considering the proportion of sporulating, sporulating + necrotic, and total diseased area, respectively. Leaf nitrogen (N) content did not change the effect of the disease on host photosynthesis.
  • • We concluded that leaf rust has no global effect on the photosynthesis of the symptomless parts of the leaves and that the large range in the quantification of leaf rust effect on the host, which is found in the literature, can be accounted for by considering the different symptom types. We discuss how our results could improve disease assessments and damage prediction in a wheat crop.

Introduction

Wheat leaf rust (Puccinia triticina) is responsible for frequent and important crop losses, and small-grain cereals are typically given two or three foliar fungicide treatments per season in Europe. In order to better identify spraying needs, decision systems based on predicted yield losses rather than on epidemic thresholds have to be developed. Quantification of the damaging effects of the pathogens on diseased plants and inclusion of these damaging functions in crop simulation models is of great importance for a more complete understanding of yield response to disease (Boote et al., 1983; Rouse, 1988; Johnson, 1992; Pinnschmidt et al., 1995; Lopes & Berger, 2001). This approach requires knowledge of the quantitative relationship between disease symptoms and the leaf physiological processes affected by the diseases (Bastiaans, 1991; Shtienberg, 1992; Beasse et al., 2000; Bassanezi et al., 2001).

Foliar pathogens reduce the photosynthetic activity in infected leaves by reducing green leaf area. Some pathogens also affect photosynthesis in asymptomatic areas of diseased leaves (Rabbinge et al., 1985; Van Oijen, 1990; Shtienberg, 1992; Bassanezi et al., 2002). In an attempt to describe this effect, Bastiaans (1991) proposed the concept of a virtual lesion, which corresponds to the leaf area where photosynthesis is zero. He developed a simple model that describes the effect of disease at the leaf scale. The model Px/P0 = (1 − x)β is fit to empirical data to describe the relationship between the reduction in photosynthesis in a diseased leaf (Px) relative to a healthy leaf (P0) and the proportion of visible disease (x). The value of β indicates whether the effect of disease on photosynthesis is higher (β > 1), lower (β < 1)or equal (β = 1) to that accounted for by the observed diseased area. Several studies have already used this approach to estimate the net photosynthesis of diseased leaves successfully (Bassanezi et al., 2001; Bastiaans, 1991, Garry et al., 1998; Lopes & Berger, 2001; Erickson et al., 2003). Moreover, consequences of disease on photosynthesis have already been incorporated into a crop growth model through the parameter β to improve accuracy of the estimates of biomass loss at the canopy scale (Bastiaans 1993; Beasse et al., 2000; Bassanezi et al., 2001; Robert et al., 2004a).

For wheat leaf rust, Bastiaan (1991) calculated a β-value of 1.26 when using data from Spitters et al. (1990). However, in a field experiment (see Fig. 2a in Shtienberg, 1992), Shtienberg found that leaf rust severity values of 1%, 10%, 20% and 40% resulted in a decrease of 20%, 50%, 58% and 60%, respectively, in host photosynthesis. When estimated using these data, β = 2.5 for leaf rust. In another paper (MacGrath & Pennypacker, 1990), severity values of 9% and 41% resulted in reductions in photosynthetic capacity of 60% and 100%, respectively, which would give a β-value of c. 9.0. This large variability in the estimations of β indicates that the relationship between leaf photosynthesis and leaf rust severity needs clarification. This is an important issue for our ability to develop robust crop loss models.

Figure 2.

Proportion of sporulating, chlorotic, necrotic and total visible tissues in wheat flag leaf (cv. Soissons) infected by Puccinia triticina. (a) Pictures of a wheat flag leaf infected by Puccinia triticina from day 11 to day 36 after onset of sporulation for a lesion density of 22 lesions cm−2. (b) Proportion of sporulating, chlorotic, necrotic and diseased (= sporulating + necrotic + chlorotic) tissues from day 11 to day 36 after onset of sporulation plotted against lesion density. Plants are grown under three fertilisation treatments: N0 (circle, low fertilisation), N1 (triangles, standard fertilisation), and N2 (squares, high fertilisation).

The purpose of our paper is to improve our understanding of why such variations in the quantification of rust damage on the host's photosynthetic activity occur. We attempted to identify the main sources of variation and to quantify their respective effects. One of the main issues of our study was then to propose a robust model linking leaf rust severity and leaf photosynthesis. From the literature, four possible sources of variation can be identified:

First, the variability may be a consequence of the fungus's developmental cycle, as shown by Scholes & Farrar (1986) and Scholes & Rolfe (1996). This type of variation may be of importance when lesions with different ages colonise the same leaves (Bassanezi et al., 2001). Variation in the effect of the fungus due to lesion age may be linked to symptom development. Indeed, biotrophic pathogens usually develop from chlorotic tissues, into sporulating lesions often surrounded by chlorotic halos, and finally into necroses; each type of symptom may have a different effect on leaf photosynthesis.

Second, the variability could result from the disease assessment method (Madden & Nutter, 1995). Peterson et al. (1948) considered the sporulating area alone as the diseased area for leaf rust on wheat. In a study on bean rust, Lopes & Berger, 2001) assessed disease as the sporulating area plus the chlorotic halo surrounding the lesions. For bean rust (Uromyces appendiculatus) Bassanezi et al. (2001, 2002) included necrotic tissue as well in the disease assessment.

Third, a source of variability may be variation in the host's physiological state. The pathogen's development is influenced by the plant age (Rapilly, 1991) and the leaf age (Johnson & Teng, 1990). Moreover, the nutritional state of the leaves shows a large range of values in field situations, which may influence lesion development and damage expression (Snoeijers et al., 2000; Robert et al., 2004b).

Last, variability in β could be a consequence of the host genotype's response to disease. Host genotype may play a significant role in the impact of disease on host photosynthesis in some plant–pathogen interactions. Erickson et al. (2003) found a large difference in the β-values (Bastiaan, 1991) between poplar lines indicating different consequences of Marssonina brunnea on photosynthesis for different host genotypes. Parameter β has however, been found to be invariant across three rice cultivars for Magnaporthe grisea (Bastiaans & Roumen, 1993) and six pea cultivars for Mycosphaerella pinodes on pea (Le May, 2002). There was also no difference for two bean cultivars in the value of β for rust, angular leaf spot and anthracnose (Bassanezi et al., 2001).

In the present paper, we attempted to improve our understanding of the variability observed in the literature in the quantitative relationship between leaf photosynthesis and leaf rust severity by testing the first three hypotheses. We thus examined photosynthetic response in relation to the severity of leaf rust, using Bastiaans’ model, across the sporulation period and for leaves grown under three different fertilisation treatments. Leaf rust severity (x, in Bastiaans’ model) was considered alternatively as the proportion of sporulating host surface area, sporulating and necrotic areas or sporulating, necrotic and chlorotic areas (total diseased symptoms). Moreover, since the proportion of the different types of symptoms may vary with lesion age and may have different effects on the host (Scholes & Rolfe, 1996), we tested a second approach (Robert et al., 2004a) in which the specific effect of each type of symptom was taken into account (a β-value was computed separately for chlorotic and sporulating tissues), rather than considering a global effect of the diseased tissues. The possible consequences of the disease assessment method for damage evaluation and modelling purposes are discussed.

Materials and Methods

Plant material

Wheat (Triticum aestivum) cv. Soissons (one of the most commonly grown cultivars in France) was used for all treatments. Seeds were left to germinate for 24 h in humid cotton and then sown in Jiffy peat pots, where they were kept for 8 wk with a 16-h light period (350 µE m−2 s−1) at 8°C and an 8-h dark period at 0°C for vernalisation. They were then transferred to square pots (1.1 l) filled with commercial compost (peat substrate, Gebr. Brill Substrate, Georgsdorf, Germany) and placed in the glasshouse. Light was complemented by sodium lamps from 08 : 00 h to 20 : 00 h. The temperature ranged between 8 and 15°C during the night and 15–30°C during the day. Plants were watered daily. They were treated against powdery mildew (Ethyrimol, Sygenta, 2 ml l-1) 3 wk before inoculation. All secondary tillers were cut the day before inoculation. For each treatment, a proportion of the plants (four for N0 and N2 and five for N1) was not inoculated and these were used as control healthy plants.

Three fertiliser (Osmocote 10N + 11P + 18K) concentrations were applied: N0 was a low fertilisation treatment with 2 g of Osmocote per pot; treatment N1 received a standard dose of 7 g per pot; N2 was an over-fertilised treatment with 18 g per pot. The N content of the control and inoculated flag leaves was measured the day before inoculation and at the end of the experiment.

The day before inoculation, the N content assessment had to be nondestructive and we then used an indirect method. Three chlorophyll-meter measurements (SPAD-502, Minolta Camera Co., Tokyo, Japan) were made along each leaf and the average value (m) to estimate the specific leaf N content (N in mg N cm−2) by: N = 0.0233 × exp(0.0426 × m). This equation was determined from data previously obtained on 20 flag leaves grown as described in this section but fertilised with 0–25 g of Osmocote per pot. The N and carbon (C) content of each of these 20 leaves was measured (Dumas, 1831; Nelson & Sommers, 1980). Their N content was then related to the chlorophyll-meter measurement to obtain this equation (Fig. 1a, R2 = 0.91). Their C content was 2.30 mg C cm−2 and was found to be independent of the fertilisation used (P = 0.15). This latter value was assumed to be the C content before inoculation for all treatments of our experiment.

Figure 1.

Specific flag leaf nitrogen content (Triticum aestivum cv. Soissons) in mg N per cm−2 of leaf. (a) Relationship between measured leaf nitrogen content (N) and leaf chlorophyll-meter measurement (C). Chlorophyll content is the mean of three measurements at the base, middle and apex of the leaf. Line indicates equation: N = 0.0233 × exp(0.0426 × C). (b) Specific N content of the leaves inoculated with Puccinia triticina for treatment N0 (circle, low fertilisation), N1 (triangles, standard fertilisation) and N2 (squares, high fertilisation). Closed symbols are for the day before inoculation (estimation from the chlorophyll-meter measurements according to the model shown in (a)) and open symbols are for the end of the experiment (direct measure).

At the end of the experiment, the leaves were harvested and their N contents were directly measured (Dumas, 1831; Nelson & Sommers, 1980).

The N levels in the leaves before inoculation and at the end of the experiment were then compared for the control and the inoculated leaves and for the different N treatments. Because of an unequal number of leaves assessed for each treatment combination we used a REsidual Maximum Likelihood (REML) analysis (Patterson & Thompson, 1971), which is implemented in the GenStat™ (2003) statistical package.

Inoculation

When plants were at the heading stage, adult flag leaves of main stems were inoculated in a settling tower (Eyal et al., 1968) with P. triticina. A wide range of lesion densities was obtained by inoculating the leaves with a mixture of leaf rust uredospores (Isolate B9384–1C1) and talc, in which the spore content was 9, 5, 3, 2, 1 or 0.5 mg (eight plants of N0 and N2 and 11 plants of N1 inoculated per dose). Infected plants were incubated for 24 h under 100%rh at 17°C and then placed back in the glasshouse until the end of the experiment. Control plants were given the same treatment with no inoculation. When chloroses were visible, we selected 17 plants from treatment N1, 18 plants from treatment N0, and 13 plants from N2 to obtain the widest range of lesion densities.

Disease assessments

The number of lesions was counted 16 d after inoculation. Lesion density was calculated as the number of lesions divided by the area of the leaf surface where gas measurements were made (see the next subsection). Digital pictures were taken at 4, 9, 11, 16, 21, 28, 36 and 43 d after onset of sporulation to accurately estimate the diseased surface area by image analysis (Optimas, Media Cybernetics, Silver Spring, MD, USA). For each picture, four color classes were defined: orange (sporulating tissue), light green to yellow (chlorotic), brown (necrotic) and green (apparently healthy leaf).

For determining the effect of N treatment on the lesion development, the proportion of sporulating, chlorotic, necrotic and total visible diseased areas were modelled separately for each assessment date using a linear relationship between proportion of disease tissues (Y) and rust lesion density (dens). A model with no effect of N treatment, that is:

image(Eqn 1.1)

was compared with a model with separate slopes for each N treatments:

image(Eqn 1.2)

For i = 0, 1, 2 (N0, N1 and N2). In this latter case, the three parameters (bi) were compared on the basis of their 95% confidence intervals.

In the model we assume zero lesions induce zero diseased tissue.

Net photosynthesis measurements

Net photosynthesis was assessed in flag leaves on different dates during the rust sporulation period: at 11, 16, 21, 28 and 36 d after onset of sporulation in diseased and control leaves. Assessments of net photosynthetic rate were made with a portable photosynthesis system (LI-6200; Li-Cor, Lincoln, NB, USA) mounted with a red LED light source (6400–02, Licor) at light saturation (1500 µmol photon m−2 s−1). The assessments were made on leaf sections with an area of 6 cm2 (3 cm × 2 cm). Net photosynthesis was calculated from the influx of CO2 occurring when the leaf is in the gas chamber, in agreement with VonCaemmerer & Farquhar (1981). When leaf rust symptoms appeared on the leaves, the precise location of the measurements on each leaf was chosen (one or two assessments per leaf) so that a large range of disease severity was represented. The leaf sections (6 cm2) on which the measurements were performed were tagged with a permanent marker and all further assessments were performed at the same places. On each assessment date, the proportions of green, chlorotic, sporulating and necrotic tissue on the leaf segment analysed for gas exchanges were measured by image analysis. Diseased surface area was defined either as: first sporulating tissue; second sporulating and necrotic tissues; and third sporulating, necrotic and chlorotic tissues.

Analyses of variance were performed to test whether net photosynthesis in control leaves was influenced by assessment dates and fertilisation treatments. In addition, on four dates, two healthy leaves were ‘gas-cartographied’ by measuring net photosynthesis along the whole length of the leaf. For statistical analysis, the position on the leaf was expressed as the ratio between the distance from the base of the leaf to the total length of the leaf. Regression analyses including the factor date (four assessment dates) were performed to test whether net photosynthesis was influenced by position along the leaf.

Modelling the effect of leaf rust

We attempted to model photosynthetic rates of diseased leaves relative to healthy leaves, as a function of leaf rust severity, lesion age and leaf N status.

Disease severity was related to relative net photosynthetic rate using nonlinear regression analysis, according to Bastiaan's model (1991) where Px is the net photosynthetic rate of a leaf with a disease severity x and P0 is the net photosynthetic rate of a healthy leaf:

image(Eqn 2.0)

Leaf rust severity (x) was considered alternatively as the proportion of sporulating host surface area (xS in Eqn 2.1), sporulating and necrotic areas (xSN in Eqn 2.2) or sporulating, necrotic and chlorotic areas (xSNC in Eqn 2.3).

image(Eqn 2.1)
image(Eqn 2.2)
image(Eqn 2.3)

Moreover, since the different types of symptoms can have different effects on the host, we tested a second approach (Robert et al., 2004a), in which a β-value was computed separately for chlorotic (βC) and sporulating tissues (βS1 or βS2). Here, the necrotic tissue was explicitly removed from the calculation (assuming β = 1 for necrosis):

image(Eqn 3.1)
image(Eqn 3.2)

where xN is the proportion of necrosed tissue area on the leaf, xS is the proportion of sporulating tissue and xC is the proportion of chlorotic tissue. In Eqn 3.1, only sporulating and necrotic tissues are considered; in Eqn 3.2, chlorotic tissues are included in the model.

The models were evaluated for each assessment date and fertilisation treatment as well as for the bulked data set. The procedure used was PROC NLIN (method DUD) of the statistical software SAS (SAS 6.12, SAS Institute, Cary, NC, USA). The β-value estimated in each treatment was compared with 1 by a t-test. Specific comparisons were performed using the nested model method to evaluate the influence of leaf N content and lesion age on the β-value. For each model, we compared the disease effects on net photosynthesis for two alternatives: Y = (1 − X)β and inline image, with i = [1,n] and n = 3 for fertilisation treatments and n = 5 for assessments dates. Comparisons were done using the lack-of-fit F-test, in agreement with Weisberg (1985).

To evaluate the quality of each model, predicted values were compared with observed values and the root mean squared errors (RMSE) were computed. We plotted the model residues against the observed values to check for the absence of bias.

Results

Nitrogen content of leaves

The day before inoculation, the N content (mg N cm−2) of the inoculated leaves ranged from 0.214 to 0.274, 0.169–0.218 and 0.160–0.210 in treatments N2, N1, and N0, respectively (Fig. 1b). The average leaf N content was not significantly different for control and inoculated leaves (P = 0.76, Table 1). There was a significant effect of the fertilisation treatment (P < 0.001). Leaf N content was significantly different (P < 0.05, t-test) in treatments N0, N1 and N2 with N2 > N1 > N0.

Table 1.  Specific leaf nitrogen (N) content of the inoculated (with Puccinia triticina) and the control leaves (Triticum aestivum cv. Soissons) for the three fertilisation treatments before inoculation and at the end of the sporulation period
NaInoculated leavesControl leaves
nbBefore inoculationdEnd of experimentenBefore inoculationEnd of experiment
[N]cSD[N]SD[N]SD[N]SD
  • a

    Fertilisation treatments: N1, standard fertilisation level; N0, low level; N2, high level.

  • b

    Number of leaves.

  • c

    Average leaf N content in mg cm−2 of leaf.

  • SD, standard deviation.

  • d

    Leaf nitrogen content is estimated from the chlorophyll meter measurements.

  • e

    Direct measure of leaf nitrogen content.

N0180.1830.0120.0820.02240.1730.0190.0710.059
N1170.1900.0140.1760.02850.1990.0490.1540.053
N2130.2330.0160.1810.04140.2310.0330.1840.037

At the end of the experiment, the N content of the inoculated leaves ranged from 0.119 to 0.262, 0.107–0.226, and 0.049–0.126 in treatments N2, N1, and N0, respectively. The average leaf N content was still not significantly different for control and inoculated leaves (P = 0.69, Table 1). There was a significant effect of the fertilisation treatment (P < 0.001). Leaf N was significantly lower (P < 0.05, t-test) in treatment N0, relative to N1 and N2 but there was no longer any significant difference between N1 and N2.

Disease severity in leaves inoculated with rust

Lesion density in the leaf segments on which net photosynthesis was measured (6 cm2) ranged from 7.8 to 101.9 lesions cm−2 in treatment N2, from 2.8 to 59.5 in N1 and from 3.8 to 33.4 in N0.

Disease development is described as the development of the proportions of sporulating, chlorotic and necrosed tissue in the infected leaves (Fig. 2).

The sporulating surface on diseased leaves increased with lesion density, and with time after inoculation until 21 d after the onset of sporulation, and reached a maximum value of around 35% of the leaf area for lesion densities greater than 20 lesions cm−2. The proportion of sporulating surface (xS) began to decrease from 21 d after the onset of sporulation for high lesion densities and from 28 d for low lesion densities. Forty-three days after the onset of sporulation, there was no longer any sporulating surface present on the diseased leaves, except at the lowest lesion densities (fewer than three lesions cm−2).

The proportion of chlorotic tissue (xC) increased with lesion density and time after inoculation until 21 d after the onset of sporulation. At this date, chlorotic tissue represented between 2% and 50% of the leaf area. Afterwards, the proportion of chlorotic tissue tended to decrease for the highest lesion densities and to remain stable or even increase for the lowest densities.

Necrosed tissue only appeared after the 16th day of sporulation and for high lesion densities first. Twenty-one days after the beginning of sporulation, the necrosed area (xN) was < 2% for lesion densities lower than 28 lesions per cm2. For higher densities (100 lesions cm−2) the necrosed area reached 29% at the same date. Leaves were almost fully necrosed 36 d after the onset of sporulation for lesion densities greater than 50 lesion cm−2.

Eleven days after the onset of sporulation, the green leaf area (1 – xSNC) was 20% for a lesion density of 100 lesions cm−2 and > 98% for the lowest lesion densities (fewer than two lesions cm−2). The green area was reduced to almost 0 for the highest lesion densities after 28 d of sporulation. For the lowest lesion densities, 50% of the leaf remained green at the end of the sporulation period (43 d).

In order to determine the effect of fertilisation treatment on the lesion development, the regression analyses (Eqns 1.1 and 1.2) were performed for lesion densities ranging between 0 and 35 lesions cm−2, which was the common range of disease severity for the three N treatments (Fig. 2). Sporulating, chlorotic and total diseased tissues were significantly linearly related to lesion density (P < 0.001) at all assessment dates (b ≠ 0 in Eqn 1, Table 2). The effect of lesion density on necrotic tissues appeared after the 16th day of sporulation.

Table 2.  Statistical analyses of the effect of nitrogen treatment on the lesion development of Puccinia triticina on wheat (Triticum aestivum cv. Soissons) flag leaves
  Day 4bDay 9Day 11Day 16Day 21Day 28Day 36Day 43
  • Proportion of sporulating, chlorotic, necrotic and total visible diseased areas were modelled for each assessment date (between day 4 and day 49 of sporulation) using a linear relationship between proportion of disease tissues and rust lesion density including the nitrogen treatment as a factor.

  • a

    Figures followed by the same letter are not significantly different (P = 0.05).

  • b

    Days after onset of sporulation.

  • c

    Lesion density effect (b 

  • ≠ 

    0, Eqn 1.1).

  • d

    The lesion density effect depends on the nitrogen treatment (b0 ≠ b1 ≠ b2, Eqn 1.2). Parameters b0, b1 and b2 (for treatments N0, N1 and N2) were compared (95% confidence intervals).

  • e

    Common slope for all N treatments (b) or separate slopes (b0, b1 and b2) for the fertilisation treatments N0, N1 and N2.

Sporulating tissuesR2  0.86  0.76  0.84  0.79  0.43  0.54
pc< 0.001< 0.001< 0.001< 0.001< 0.001< 0.001< 0.001< 0.001
pd  0.36  0.66  0.08  0.02  0.16  0.002  0.007  0.49
be  0.00439  0.0603  0.00718  0.01112  0.00169
b0  0.01010a  0.01145a  0.00639a
b1  0.00952ab  0.00942ab  0.00518ab
b2  0.00839b  0.00651b  0.00258b
Chlorotic tissuesR2  0.49  0.64  0.50  0.54  0.56
pc< 0.001< 0.001< 0.001< 0.001< 0.001< 0.001< 0.001< 0.001
pd< 0.001< 0.001  0.18  0.009  0.01  0.11  0.19  0.98
be  0.00554  0.00867  0.00826  0.00486
b0  0.00612a  0.00407ab  0.00981ab  0.01130a 
b1  0.00393b  0.00572a  0.01090a  0.00938ab 
b2  0.00374b  0.00235b  0.00756b  0.00816b 
Necrotic tissuesR2  0.0002  0.29  0.36
pc  0.26  0.20  0.18  0.02< 0.001< 0.001< 0.001< 0.001
pd  0.24  0.59  0.56  0.07  0.651  0.01  0.07  0.04
be  0.000105  0.000582  0.02225
b0  0.00726b  0.0396a
b1  0.01061ab  0.0292ab
b2  0.01597a  0.0242b
Total diseased tissuesR2  0.84  0.83  0.81  0.79  0.71  0.45
pc< 0.001< 0.001< 0.001< 0.001< 0.001< 0.001< 0.001< 0.001
pd< 0.001< 0.001  0.03  0.002  0.67  0.13  0.14  0.19
be  0.02143  0.02841  0.03548  0.03695
b0  0.01069a  0.01019ab  0.013670a  0.01928ab
b1  0.00812b  0.01158a  0.011610b  0.02048a
b2  0.00830b  0.00835b  0.012443ab  0.01612b

Lesion development was quite similar in the three fertilisation treatments: over a total of 32 cases (eight dates by four types of tissues), 12 showed a significant N effect (Table 2). Moreover, the effect of N on the development of chlorotic, necrotic and total diseased tissues was unclear as the order of the effect of the fertilisation treatments (expressed by the values bi in Eqn 1.2) changed across the assessment dates and no consistent effect could be established. The proportion of sporulating tissues was significantly higher in N0 treatment relative to N2 treatment at days 16 (P = 0.02), 28 (P = 0.002) and 36 (P = 0.007) of sporulation. This effect was however, small at day 16: the proportion of leaf area covered by sporulating tissue would be 0.13, 0.14 and 0.15 in treatments N2, N1 and N0, respectively, for a lesion density of 15 lesion cm−2, according to Eqn 1.2.

Effect of rust on net photosynthesis

In control leaves, the net photosynthetic rate ranged from 18.7 to 27.5 µmoles CO2 m−2 s−1. When considering all the data (five assessment dates by three N treatments), net photosynthesis in control leaves was independent of fertilisation treatment (P = 0.14) and no interaction between date and fertilisation treatment was found (P = 0.77), but there was a significant effect of the date (P < 0.05).

The net photosynthetic rate in control leaves (Fig. 3) was also significantly related to the position along the leaf (P < 0.001) and was dependent on the date (P < 0.001), but with no interaction between position and date (P = 0.25). The effect of the position along the leaf on leaf net photosynthesis was limited, however, with a slope equal to −3.3 µmoles CO2 m−2 s−1 (per proportion of leaf length) and a R2 value of 0.34. According to this relationship, net photosynthetic rate should decrease by only 1.65 µmoles CO2 m−2 s−1 (7% relative to the average value) when the distance to the apex increases by 50%. Moreover, when restricting the data to the positions between 10 and 75% of the total leaf length from the base, the net photosynthetic rate in control leaves was no more related to the position along the leaf (P = 0.38). We therefore ignored the location of the measurement along the leaf when estimating the net photosynthesis rate on control leaves.

Figure 3.

Control flag leaf photosynthetic rate (µmoles CO2 m−2 s−1) measured at different positions along the plant leaf (expressed in proportion of the length of the leaf from the leaf basis) (Triticum aestivum cv. Soissons). Data from two different leaves (closed and open symbols) are presented. Photosynthesis measurements were taken on four dates: 8 d, 15 d, 25 d, and 36 d after onset of sporulation in the leaves inoculated with Puccinia triticina.

For each assessment date and fertilisation treatment, net photosynthesis in diseased plants (Px) was related to the corresponding value (P0) in control leaves and to disease severity (Fig. 4). Eqns 2.1, 2.2, 2.3, 3.1 and 3.2 were fitted to each data set (Table 3).

Figure 4.

Relative photosynthesis of diseased leaves (Px/P0) during the sporulation period (day 11 to day 36 after the onset of sporulation) related to the proportion of diseased leaf area in (Triticum aestivum cv. Soissons) plants. Disease is measured as: (a) sporulating areas (b) sporulating plus necrotic areas, and (c) sporulating, necrotic and chlorotic areas. Plants are grown under three fertilisation treatments: N0 (circle, low fertilisation), N1 (triangles, standard fertilisation), and N2 (squares, high fertilisation).

Table 3.  Values of β estimated for leaf rust (Puccinia triticina) with wheat (Triticum aestivum cv. Soissons) for different fertilisation treatments
Nitrogen treatmentDateP0dEqn 2.1aEqn 2.2bEqn 2.3cEqn 3.1Eqn 3.2
βSsdR2βSNsdR2βSNCsdR2βS1sdR2βS2sdβCsdR2
  • Fertilisation treatments (N1, standard fertilisation level; N2, high fertilisation level; N0, low fertilisation level),

  • at different assessment dates (days after onset of sporulation).

  • a

    disease is computed as the sporulating area.

  • b

    disease is computed as the sporulating and necrotic area.

  • c

    disease is computed as the sporulating, necrotic and chlorotic area.

  • d

    control leaf photosynthesis.

  • e

    pooled data for all nitrogen treatments.

N01127.5 1.720.130.741.720.130.740.880.070.741.720.130.741.320.330.430.340.76
1625.4 1.500.130.681.480.120.700.680.050.721.480.120.700.950.330.460.270.73
2123.6 2.180.140.712.060.120.760.850.040.822.100.130.741.260.250.570.170.84
2825.8 3.030.270.541.910.220.341.090.070.782.270.280.410.950.251.280.260.78
3623.010.81.521.880.170.441.040.110.313.520.580.221.770.790.700.300.45
N11126.2 1.570.140.711.570.140.710.880.080.691.570.140.711.290.260.380.300.73
1624.6 2.350.180.692.340.170.700.900.050.832.340.170.700.510.411.140.260.83
2124.8 2.110.180.672.040.160.721.130.070.822.060.170.711.050.301.240.360.82
2825.1 4.620.670.572.020.320.550.890.080.872.960.430.721.090.460.870.220.88
3622.8 8.531.481.200.090.770.800.060.791.640.300.761.080.360.360.190.81
N21124.3 1.580.210.721.570.200.730.630.060.881.570.200.730.300.330.840.230.89
1621.7 2.470.370.372.330.310.510.860.090.752.370.340.461.160.330.670.190.76
2122.2 2.330.300.451.950.190.690.960.070.842.020.220.650.910.260.990.260.84
2823.6 6.961.791.130.140.700.750.090.741.590.410.731.140.550.310.300.77
3622.121.55.411.360.070.900.910.060.852.780.420.891.750.690.450.240.86
Ne11 1.620.090.721.620.090.720.790.050.741.620.090.721.180.160.430.140.76
16 2.080.130.532.040.120.590.820.040.772.050.130.570.850.180.790.120.77
21 2.200.120.622.010.090.740.970.040.822.060.100.721.220.150.750.130.83
28 4.210.440.131.650.140.460.940.050.792.430.220.631.220.220.790.140.79
3611.31.351.440.070.680.900.040.722.490.270.661.240.330.620.150.73
Pooled data  2.680.140.011.760.050.720.890.020.842.030.060.751.110.080.710.060.84

The relationship between relative photosynthetic rate of infected leaves and the proportion of sporulating area varied greatly as the lesion developed (Fig. 4a). The estimated value of βS from Eqn 2.1 ranged between 1.5 and 21.5 (Fig. 5) and was strongly dependent on lesion age (F = 71, ν1 = 4 and ν2 = 286), but was independent of the fertilisation treatment (F = 0.86, ν1 = 2 and ν2 = 286). On average, over the three fertilisation treatments, βS was estimated to be 1.6 at day 11 of sporulation; it remained stable at around 2.1 between days 16 and 21 of sporulation, and increased rapidly towards high values after day 21. After 36 d of sporulation, the infected leaves were almost fully necrosed (Fig. 2) and the sporulating area was no longer related to the photosynthetic capacity (Fig. 4a). The only difference between the three fertilisation treatments was that βS was higher for N2 (21.5) than for N1 (8.5) and N0 (10.8) at the end of the sporulation period. With pooled data (three fertilisation treatments and five assessment dates), βS was estimated to be 2.68 ± 0.27.

Figure 5.

Evolution of the parameter β during the sporulation period according to the model used (Eqns 2.1, 2.2, 2.3, 3.1, and 3.2): (a) three fertilsation treatments (N2, squares; N1, triangles; N0, circles); (b) average values for the three fertilisation treatments. Mean values are shown with confidence interval bars (P = 0.05). Y-axis scale is different in (a) and (b) and for Eqn 2.1.

The relationship between relative photosynthetic rate of infected leaves and the proportion of sporulating and necrotic area varied only a little as the lesion developed (Fig. 4b). The parameter βSN estimated from Eqn 2.2 ranged between 1.13 and 2.34 (Table 3, Fig. 5) and was much more stable than βS, but again showed a significant effect of lesion age (F = 6.96, ν1 = 4 and ν2 = 286). The fertilisation treatment had no effect on βSN (F = 0.85, ν1 = 2 and ν2 = 286). Up until 16 d after the onset of sporulation, almost no necrosis was visible on the leaves and βSN was close or equal to βS. Afterwards, the average βSN value for the three fertilisation treatments decreased from 2.04 at day 16 to 1.48 at day 36. When estimated with the pooled data set (three fertilisation treatments and five assessment dates), βSN = 1.76 ± 0.10.

The relationship between relative photosynthetic rate of infected leaves and the proportion of total diseased area (sporulating + necrotic + chltorotic) did not vary as the lesion developed (Fig. 4c). The estimated values of βSNC (Eqn 2.3) ranged between 0.63 and 1.13 (Table 3, Fig. 5). No fertilisation treatment effect was found (F = 1.55, ν1 = 2 and ν2 = 286) and βSNC was more stable than βS with respect to lesion age. The effect of lesion age was significant at P = 0.95 but not at P = 0.99 (F = 3.08, ν1 = 4 and ν2 = 286). On average for the three fertilisation treatments, βSNC was slightly lower than 1, and tended towards 1 at the end of the sporulation period: it increased from 0.79 to 0.90 between days 11 and 36 of the sporulation period. When estimated with the pooled data set (three fertilisation treatments and five assessment dates), βSNC = 0.89 ± 0.04.

The value of βS1 estimated from Eqn 3.1 ranged between 1.48 and 3.52 (Table 3, Fig. 5) and was dependent on lesion age (F = 4.65, ν1 = 4 and ν2 = 286) but not on the fertilisation treatment (F = 0.60, ν1 = 2 and ν2 = 286). On average over the three fertilisation treatments, βS1 was estimated to be 1.6 at day 11 of sporulation and remained stable around 2.05 between days 16 and 21 of sporulation. After day 21, βS1 increased slightly and reached 2.5 after 36 d of sporulation. When fitting Eqn 3.1 to the pooled data set (three fertilisation treatments and five assessment dates), we found βS1 = 2.03 ± 0.12.

The values of βS2 and βC estimated from Eqn 3.2 (Table 3, Fig. 5) were independent of lesion age (F = 1.97, ν1 = 8 and ν2 = 281) and fertilisation treatment (F = 1.33, ν1 = 7 and ν2 = 281). When Eqn 3.2 was fitted to the pooled data set, βS2 and βC were estimated to be 1.11 ± 0.16 and 0.71 ± 0.11, respectively.

Model comparison

We compared the models in order to select the most appropriate equation for estimating photosynthesis of a diseased leaf relative to a healthy leaf in field conditions. In the field, lesions of different ages are simultaneously present on leaves. Thus, our aim was to select a model which is robust with respect to lesion age. With this in mind, we compared the five following equations (4.1, 4.2, 4.3, 4.4 and 4.5) that result from models 2.1, 2.2, 2.3, 3.1 and 3.2 when using the parameters estimated with the pooled data set:

image(Eqn 4.1)
image(Eqn 4.2)
image(Eqn 4.3)
image(Eqn 4.4)
image(Eqn 4.5)

Estimated and observed values are compared in Fig. 6. Table 4 indicates the corresponding R2, slope and intercept values for each assessment date and fertilisation treatment, as well as for the pooled data set. The residues (predicted value minus observed value) were plotted against observed values (Fig. 7). Table 4 indicates the corresponding RMSE obtained for the different treatments.

Figure 6.

Predicted values plotted against observed values for Px/P0 calculated with Eqns 4.1, 4.2, 4.3, 4.4, and 4.5: (a) for the three fertilisation treatments (N2, squares; N1, triangles; N0, circles, data are pooled for all assessment dates); (b) for the five assessment dates in days after onset of sporulation (open circle, 11 d; diamond, 16 d; open square, 21 d; closed circle, 28 d; closed triangle, 36 d; data are pooled for all fertilisation treatments).

Table 4.  Model evaluation: simulated Px/P0 fitted against observed Px/P0
 Eqn 4.1βS = 2.68Eqn 4.2βSN = 1.76Eqn 4.3βSNC = 0.89Eqn 4.4βS1 = 2.03Eqn 4.5βS2 = 1.11 and βC = 0.71
abR2RMSEabR2RMSEabR2RMSEabR2RMSEabR2RMSE
  • a, slope; b, intercept; R2, coefficient of determination; RMSE, root mean square errors.

  • *

    intercept significantly different from 0.

  • a

    data are pooled for all nitrogen treatments and assessment dates.

  • b

    for each fertilisation treatment, data are pooled for all assessment dates.

  • c

    for each assessment date, data are pooled for all nitrogen treatments.

pooled dataa
 0.530.27*0.150.240.92 0.050.730.130.96 0.030.840.101.06−0.040.750.120.99 0.010.840.09
fertilisation treatmentb
N00.690.20*0.230.210.96 0.02*0.810.100.95 0.040.840.091.04−0.020.770.110.96 0.030.850.09
N10.670.19*0.240.210.90 0.06*0.690.131.00−0.010.840.091.08−0.050.750.121.02−0.020.840.09
N20.240.40*0.040.300.91 0.06*0.670.150.95 0.050.840.111.07−0.040.710.141.00 0.020.840.10
assessments datec
110.900.19*0.740.151.17−0.100.750.090.96 0.060.740.091.07 0.010.750.101.00 0.030.770.09
160.960.09*0.540.161.29−0.240.610.140.97 0.040.780.101.17−0.110.590.131.03 0.000.770.10
210.960.080.620.141.13−0.120.750.111.01−0.030.810.091.08−0.050.720.111.05−0.050.820.09
280.690.060.270.250.75 0.130.520.170.96 0.000.790.110.91 0.020.620.150.98 0.000.780.11
360.300.16*0.030.440.78 0.130.720.130.73 0.11*0.840.100.78 0.060.740.120.75 0.11*0.820.10
Figure 7.

Residues (predicted minus observed values for Px/P0) against observed values for Eqns 4.1, 4.2, 4.3, 4.4, and 4.5: (a) three fertilisation treatments (N2, squares; N1, triangles; N0, circles; data are pooled for all assessment dates); (b) five assessment dates in days after onset of sporulation (open circle, 11 d; diamond, 16 d; open square, 21 d; closed circle, 28 d; closed triangle, 36 d; data are pooled for all fertilisation treatments).

The model quality was similar for the three fertilisation treatments (Table 4, Fig. 6a, Fig. 7a) but differed according to the equation used. When plotting predicted vs observed values, R2 varied from 0.15 for Eqn 4.1 to 0.84 for Eqns 4.3 and 4.5 and the intercept was significantly different from 0 for Eqn 4.1. Moreover, a strong bias due to lesion age was observed for Eqn 4.1 (Fig. 6b, Fig. 7b), with predicted values overestimated at the beginning of the sporulation period (bias =+0.122 at day 11) and strongly underestimated at the end (bias = −0.400 at day 36).

Predicted values obtained with Eqns 4.2, 4.3, 4.4 and 4.5 were more consistent with the experimental data for the three fertilisation treatments and the five assessment dates (Table 4). Eqn 4.2 was however, less appropriate than the others as it slightly overestimated Px/P0 for high observed values (Px/P0 > 0.8) and underestimated Px/P0 for low observed values (Px/P0 < 0.3) at dates 11, 16 and 21 (Fig. 7). In addition, this model overestimated Px/P0 for the observations at 36 d after onset of sporulation (bias =+0.06). This bias was also observed for Eqn 4.4. These observations are consistent with the fact that models 4.3 and 4.5 were the only ones found to be independent of lesion age (and thus to the change in proportion in the different types of symptoms). Eqn 4.5 did not improve the model quality relative to Eqn 4.3, even though two parameters were used instead of one, in order to take into account a specific effect of each type of symptom.

Discussion

Effect of plant fertilisation on rust development and damage

Despite the large differences in the amount of N applied to the plants, there were small (but significant) differences in the leaf N status (Table 1). A likely explanation is that plants given little fertilizer produced fewer tillers and had much smaller leaves. Plants react to early N deficiency first by changing their structure and then by decreasing their leaf N content. As a consequence, the photosynthesis in control leaves was not significantly different in the three treatments. The range of the leaf N contents of our glasshouse-grown plants was close to that obtained in standard field conditions; in Northern France it varies between 0.23 and 0.30 mg per cm2 of leaf at flowering, depending on the N fertilization rate and timing (Girard, 1997) and is in the same range in the UK (N. Paveley, 2004, pers. comm.).

In our experiment, fertilisation treatment had only a marginal effect on rust symptom development. The only significant difference between the treatments was in the proportion of sporulating tissues, being slightly higher in the low fertilisation treatment at the end of the sporulating period. At this time however, the necrotic tissues were already larger than the sporulating tissues and image analysis precision was lower. This does not mean, however, that the plant N content did not affect the pathogen. First, in the same experiment, we found (Robert et al., 2004b) that the low fertilisation treatment induced a lower spore production by the lesions. Second, although the plants of the three treatments were inoculated by the same method, the highest lesion density was 33 lesions cm−2 in N0, compared with 59 lesions cm−2 in N1 and 100 lesions cm−2 in N2. This suggests that leaf N content might have affected the rust infection process, as already mentioned by Tiedemann (1990). Pathogen development might also have been influenced by phosphate (P) and potassium (K) contents as our fertilization treatments altered the amount of N available for the plant but certainly also changed the amount of K and P.

Despite these differences in the pathogen development, we did not find any significant difference between the β-values obtained for the three fertilisation treatments. This suggests that the pathogen effect on the photosynthetic competence of the leaves was unaffected by leaf N content. Only at the end of the sporulation period was βS calculated with Eqn 2.1 higher (although not significantly) for N2 relative to N1 and N0. This, however, was probably an artefact that could be attributed to differences in the range of lesion densities among the three N treatments. Indeed, when βS was calculated for lesion densities < 40 lesions cm−2, such differences were no longer found.

Rust symptoms and damages

Net photosynthesis in control leaves slightly decreased toward the leaf apex. This was expected as it is well known that leaf N is strongly correlated with amount and quality of light received (Dreccer et al., 2000). In our experiment however, this effect was small and the photosynthesis was not significantly different when considering the 10–75% part of the leaf length. This was probably linked to our experimental design as flag leaves were maintained horizontally and homogeneously exposed to light.

Bastiaans’ model (Eqn 2.0) is a global and robust (Robert et al., 2004a) approach to evaluate the photosynthesis of a diseased leaf relative to a healthy leaf. It allows quantifying globally the photosynthetic competence of diseased leaves. This model has been widely used (Garry et al., 1998; Bassanezi et al., 2001; Lopes & Berger, 2001; Erickson et al., 2003) to simulate disease effects on leaf photosynthesis. The value of β indicates whether this disease reduces (β > 1) or not (β = 1) photosynthesis in the remaining green leaf tissue (deep effect), but it does not give any information about the mechanisms responsible for and the precise location of this reduction.

In our study, the photosynthesis assessments were done between the 11th and the 36th day after onset of sporulation. When sporulating tissues alone were taken into account (Eqn 2.1), the measured βS values were close to 2 until the 21st day of sporulation but then increased up to 11 (on average over the three N treatments) at the end of the sporulation period. Including the necrosed tissue in the calculation (Eqn 2.2) removed huge variations in β, resulting in a range from 1.4 to 2. In Eqn 3.1, we hypothesised β = 1 for necrosed tissue and we found that 1.4 < β < 2.5 for sporulating tissue. Moreover, from days 11–21, necrosed tissue was either absent or else occupied only a small fraction of the diseased area, and the β-values from Eqns 2.2 or 3.1 were similar and varied between 1.6 and 2.0. This suggests, first, that it is reasonable to assume that necrosed tissue has no effect on the photosynthetic capacity of the surrounding green tissue and, second, that the effect of sporulating lesions on host photosynthesis is greater than that which could be accounted for by the surface area occupied by the lesion itself (sporulating area).

When chlorotic tissues were included in the assessment of the disease severity, along with necrosed and sporulating areas, calculated values of β decreased further, becoming < 1 (Eqn 2.3). When the lesions were fully developed and before the occurrence of necrosis (days 11, 16 and 21), inclusion of the chlorotic halo surrounding the lesions decreased β from values between 1.4 and 2 (Eqn 2.2) to values between 0.8 and 1.0 (Eqn 2.3). Moreover, when we partitioned the disease effect between sporulating surface and chlorotic surface (Eqn 3.2), we found that 0.85 < β < 1.24 for sporulating tissue and 0.43 < β < 0.79 for chlorotic tissue. This suggests that the ‘deep effect’ of the sporulating tissues on the host photosynthesis (β > 1) can be accounted by the photosynthesis reduction observed in the chlorotic halos surrounding the lesions.

The β-values found for chlorotic halos, lower than 1, suggest that chlorotic tissues are still partly photosynthetically active. This would be in agreement with previous studies that have shown that the chlorotic regions surrounding sporulating areas are still photosynthetically active for crown rust of oat leaves (Scholes & Rolfe, 1996), for brown rust of barley (Scholes & Farrar, 1986), and rust of leek (Roberts & Walters, 1988).

When considering sporulating lesion plus chlorotic halo plus necrosis, β-values were very close to 1. It can thus be concluded that at the leaf scale the ‘virtual lesion’ (leaf area where photosynthesis is nil), was almost equal to the total visually affected tissue and this in turn suggests that at the leaf scale, leaf rust has no global effect on the net photosynthesis of the remaining green parts of the leaves.

Our results are consistent with studies carried out at the leaf scale for other rusts. Using Bastiaans’ model, Bassanezi et al. (2001) and Lopes & Berger (2001) found β-values near 2 and 1, respectively, for bean rust (Uromyces appendiculoatus) and they concluded that bean rust causes a reduction in photosynthesis only in a limited area near the pustules. Similarly, reduction in photosynthesis rate was only slightly higher than would be expected when only considering the reduction in the green leaf tissues in corn infected by Puccinia sorghi (Shtienberg, 1992). Bassanezi et al. (2002) has suggested that the damage intensity caused by the diseases is linked to their trophic relationship with the host and that for rusts, the effect of the pathogen on the remaining green leaf area is minimal.

At the tissue scale, work has also been performed on photosynthetic disruption following rust infection. These results are partly consistent with ours as in rust-infected leek (Robert & Walters, 1988), bluebell (Scholes & Farrar, 1985), oat (Scholes & Rolfe, 1996), and bean leaves (Bassanezi et al., 2002) photosynthesis was most severely inhibited within fungal pustules than in the rest of the leaf. However, using quantitative chlorophyll image analysis, for rust-infected oat leaves, Scholes & Rolfe (1996) observed a reduction in the photosynthesis of the green part of the leaf during the late stage of disease development. Moreover, in barley leaves infected with brown rust (Scholes & Farrar, 1986), photosynthesis was inhibited preferentially in regions between fungal pustules, whereas within infected regions it was similar to or even higher than uninfected leaves. Direct photosynthetic rate assessments in the green parts of the diseased leaves or quantitative chlorophyll image analysis in the diseased leaves would test our hypothesis that leaf rust has no effect on the net photosynthesis of the remaining green parts of the leaves.

Variability in damage prediction

Our study proposes an explanation for the high variability in the quantification of leaf photosynthesis reduction caused by leaf rust. Depending on the type of diseased tissue taken into account and the lesion age, the calculated effect of the disease on the host's photosynthetic competence can vary from β = 0.7 to > 10. Such a range of variation is consistent with the data found in the literature (MacGrath & Pennypacker, 1990; Spitters et al., 1990; Shtienberg, 1992).

We found that the relationship between photosynthetic rate of infected leaves and the proportion of sporulating area varied greatly as the lesion developed (β from 2 to 11). When considering the proportion of both sporulating and necrotic area, this relationship varied much less (β from 1.4 to 2). Finally, when considering the total diseased area, this relationship was almost constant as the lesions develop (β from 0.8 to 1). Similarly, Bassanezi et al. (2001) showed that when all visible symptoms were assessed, no differences were found among β-values at the different stages of bean rust development. Madden & Nutter (1995) already stated that the difficulties in accurately measuring disease severity were mainly responsible for the absence of robust relationships between symptoms and damage. In particular, in field observations, lesions of different generations are present on the same leaves and, depending on how the disease is assessed, values of β may vary greatly. Our study confirmed this by illustrating the consequences of changing the method of disease assessment on the damage model parameter.

Because the proportion of the different types of symptoms vary with lesion age (especially sporulating and necrosed tissues), we hypothesised that partitioning the disease effect between sporulating and chlorotic surface, and evaluating the specific effect of each of these symptoms (Eqn 3.1, 3.2), rather than considering a global effect (Eqn 2.0, 2.1, 2.3), would reduce the parameter variability accounted for by lesion age and thus improve the model robustness. In particular, Eqn 3.2 takes into account the changes in the proportion of each symptom type during the sporulation cycle. However, Eqn 3.2 quite surprisingly explained our data no better than Eqn 2.3. An explanation might be that the effects of the different types of diseased tissues were not so different: as it is shown by Eqn 3.2, β = 1.1 for sporulating tissue, β = 1.0 for necroses and β = 0.7 for chlorosis.

We thus conclude that, under the conditions of our experiment, the variability in the relationship between photosynthetic rate loss and the proportion of sporulating area was mainly accounted for by the development of necrosis. Taking into account that chlorotic tissue slightly reduced the variability of the prediction but separated chlorotic and sporulating tissue did not improve predictions further. This, however, may be different for other pathogens.

Disease assessment and damage prediction

In the field, lesions of different ages are present on the same leaves. As shown by the huge variation of β with lesion age and poor predictive quality of Eqn 4.1, considering only sporulating tissue could lead to substantially biased estimates of the disease impact on host photosynthesis in the field. Considering the necrosed tissue in the disease assessments (Eqn 2.2) considerably increased the predictive quality but still maintained a slight bias linked to the lesion age. Finally, considering the total visible diseased tissue gave the best prediction. According to our results, a β-value of 1.8 (Eqn 4.2) or of 0.9 (Eqn 4.3) could be used to predict the decrease in photosynthesis through the whole sporulation period when assessing sporulating and necrosed symptoms together or all the visible diseased symptoms, respectively.

In field studies, there is a trade-off between assessment feasibility and the precision needed for damage estimation. Although we can conclude from the evaluation of the different forms of our model that taking into account chlorotic along with necrosed and sporulating tissues would lead to the best estimation of photosynthesis loss due to leaf rust, the feasibility of such a disease assessment in the field is open to discussion. In addition, most pathogens induce an acceleration of the leaf's natural senescence, which can be taken into account in our approach by considering the necrosed tissue as a whole and using Eqn 2.2 or 2.3. Eqn 3.2 requires the separate assessment of sporulating, necrotic and chlorotic areas, which is unrealistic for field assessments and could only be achieved through image analysis.

In conclusion, this study, by quantifying the relationships between rust symptoms and leaf damage, is the first step in modelling crop loss due to leaf rust. Coupling epidemic and crop growth models is a promising way to improve the understanding and prediction of yield reduction by diseases (Boote et al., 1983; Johnson & Teng, 1990; Pinnschmidt et al., 1995). Epidemic kinetics is based on the pathogen multiplication rate and thus sporulating area is a key variable in epidemic models as it allows the estimation of spore production (Robert et al., 2002, 2004b). By contrast, our study has demonstrated that, for estimating damages with crop growth models, the key variable is the total visible diseased area. Therefore in order to formulate an accurate model that combines both epidemiology and crop loss, it is necessary to consider these two key variables. This provides motivation for further investigation of symptom development in the infected leaves and modelling lesion growth including the different types of symptoms.

Acknowledgements

We thank E. Cachet, H. Autret, and J. Rodrigues for their help with photosynthesis measurements and image analysis. We thank P. Belluomo for image analysis programming. We are very grateful to P. Huet and S. Power for their valuable help in statistical analyses.

Ancillary