Protection Quarterly Vol.26(2) 2011 57 recorded. In a survey of canola (Brassica napus L.) crops, Lemerle et al. (1999), found Association between environmental factors and the fumitory in 42% of the fields. Indeed, it is possible that the substantial increase in the occurrence of six fumitory species ( spp. L.) area sown to canola across this zone in the in southern-eastern last 20 years may be associated with the in- creasing incidence of fumitory. There are Gertraud M. NortonA,B, Deirdre LemerleC, James E. PratleyA,C and Mark R. two factors that may be related to this in- A,C,D creased fumitory incidence. Firstly, there Norton are no herbicides capable of selectively re- A School of Agricultural and Wine Sciences, Charles Sturt University, Wagga moving fumitory from canola, and second- Wagga, New South Wales 2678, Australia. ly, if fumitory seed is harvested together B Department of Agriculture, Forestry and Fisheries, GPO Box 858, Canberra, with canola, its similar size precludes the ACT 2601, Australia. possibility of decontaminating the canola C EH Graham Centre for Agricultural Innovation (an alliance between CSU seedlot. This scenario indicates that canola and NSW DPI), Private Mail Bag, Wagga Wagga, New South Wales 2650, seedlots might be a means whereby fumi- Australia. tory is spread further across the cropping D NSW Department of Primary Industries, GPO Box 1600, Canberra, ACT zone and indeed fumitory has been found in canola seedlots (Norton 2003). 2601, Australia. Email: [email protected] The work reported here is apparently the first attempt to survey exclusively the distribution of this genus anywhere in the world. Consequently there is little infor- Abstract species are recognized world-wide, some mation available about the natural distri- The occurrence of Fumaria as a weed in of which occur as weeds in agricultural or bution of the genus. No previous surveys south-eastern Australian cropping sys- horticultural crops (Sell 1964, Lidén 1986). within Australia, which have included tems is believed to have increased sub- However, Australian herbarium records fumitories within the suite of weeds un- stantially in recent decades. To study and associated state and nation-wide Flora der study, have identified it beyond the this, a survey was conducted in contrast- works suggest that only seven species are genus and none has attempted to associ- ing regions of this zone, viz. southern present: Bor., F. capreolata ate its occurrence with any environmental New South Wales, mid-north South Aus- L., F. densiflora DC., F. muralis Sonder ex descriptors. Anecdotal information indi- tralia. The survey analysed the pattern Koch, F. officinalis L., F. parviflora Lam. and cates there are differences in susceptibility of occurrence of each of the six natural- F. indica (Hauskn.) Pugsley (Toelken 1986, of the different fumitory species to herbi- ized species found (Fumaria bastardii, Harden 1990, Walsh 1994). While F. indica cides. Therefore, to begin to develop an F. densiflora, F. muralis, F. officinalis, appears to be restricted to inland areas of understanding of the distribution of fumi- F. parviflora and F. capreolata) and the Australia, all remaining species occur in tory across the cropping zone it is essen- natural environmental factors associated the southern winter cropping zone. tial for weed control purposes to identify with their distribution. While five spe- In considering the abiotic factors that to the species level. The research cies were primarily found in agricultural control the distribution of vegetation reported here is the first fumitory survey environments, F. capreolata occurred ex- communities and plant species, the works reporting at the species level in Australia. clusively in non-agricultural situations of Woodward and Williams (1987) and To begin to address this shortfall, a tar- characterized by the presence of high Brown (1984) were important in empha- geted survey was carried out in two con- soil organic matter. F. densiflora and sizing the influence of minimum tempera- trasting regions of the south-eastern crop- F. bastardii were the most widespread ture and moisture availability. Species that ping zone designed to: (i) identify natural and abundant species. F. officinalis was become weeds are a unique case however, environmental factors of soil, climate and the rarest. because human management of the eco- topography associated with species occur- Environmental factors were signifi- system or some other human interaction rence, and (ii) extend understanding of the cantly associated with the occurrence of with the plant itself can be instrumental in areas of distribution for each Fumaria spe- each species. Soil texture and/or rainfall facilitating the development of these spe- cies. during one of the autumn months were cies into weeds (Grime 1979). An histori- important. Factors associated with some cal overview of the weed surveys of the Materials and methods species overlapped with other species south-eastern Australian cropping zone Survey and it was common to find more than one suggests a substantial increase of Fumaria The high potential for plants of the Fumaria species at a site. Occurrence over a wide occurrence. In the 1960s, fumitory was genus to become serious crop weeds and range of soil types in some species sug- found in less than 4% of crops in southern the lack of knowledge of the environmen- gests the presence of substantial bio- and New South Wales (Taylor and Lill 1986), tal factors influencing its distribution pro- ecotypical heterogeneity. although occasional serious infestations vided the basis for the survey. As it was Keywords: Generalized linear model, were noted. By the early 1990s, the propor- not feasible to survey the entire southern multiple logistic regression, weed distri- tion of cereal crops infested with fumitory Australian cropping zone where fumitory bution. had risen to 36% in the same area (Lemerle has been known to occur, it was consid- et al. 1996). In Victoria, Wells and Lyons ered that an understanding of the envi- Introduction (1979) showed large regional variations in ronmental influences could be achieved The members of the genus Fumaria L., its occurrence with 8 to 36% of cereal crops by keeping latitude constant while vary- known as fumitory or carrot weed in infested. Velthuis and Amor (1982/83) ing rainfall distribution and soil type. Australia, are annual herbs of semi-erect, observed that, despite being present in The first region covered part of the South trailing or climbing habit, ranging in ori- only 4% of south-western Victorian cereal Australian wheat belt and experiences a gin from Western Europe, throughout the crops, the average magnitude of its infes- typical Mediterranean type climate with Mediterranean Basin and eastward to the tations meant that it ranged from the sixth distinctly dry summers and cool, moist Indian Subcontinent. About 50 different to tenth most abundant of the 38 weeds winters. Soils are predominately neutral to 58 Plant Protection Quarterly Vol.26(2) 2011 alkaline, and 20% of the sites in the survey Statistical analysis i. With each Fumaria species considered were in that region. The second region, A range of different statistical procedures as the dependent variable and each comprising 80% of the survey sites, was and models are available to analyse po- of the environmental factors the inde- characterized by intermittently storm- tential weed distributions (Kriticos and pendent variables, simple logistic re- prone summers, cool-moist winters and Randall 2001). In this survey a general- gressions were undertaken. Initially, acidic soils. This region extended from ized linear modelling approach employ- all regressions were calculated using northern Victoria to the northern bound- ing stepwise multiple logistic regression the complete data set of 161 sites al- ary of the winter cropping area in central was a central component of the statisti- though subsequently data points from New South Wales (Figure 1). cal analyses. This method is increasingly non-agricultural areas were excluded The survey was undertaken from early utilized and was preferred here because in the species that primarily occurred July to the end of September. To maximize of its ability to relate the occurrence of a in agricultural areas, i.e. all species ex- the number of samples found, areas of species (dependent variable) to multiple cept F. capreolata and F. muralis, as these known Fumaria presence were surveyed environmental parameters (independent sites exerted unacceptable leverage in predominantly. From a randomly cho- variables) (Saab 1999, Collingham et al. shaping the slope of the regression sen starting point in the designated area, 2000, Grundy and Mead 2000, Corbacho function. Indeed, as the main interest crops were inspected at intervals of ap- et al. 2003). of the survey was the exploration of Fu- proximately 10 km if Fumaria was present, Due to the categorical nature of several maria presence in agricultural areas, the or less, if absent. Regions with undulating survey variables, logistic regressions with data from only the 146 agricultural sites topography were sampled at greater in- binary response variables (0 = absent, 1 = were used for F. bastardii, F. densiflora tensity. At each point, latitude, longitude present) were used and their significance and F. parviflora. and altitude were recorded with a Trimble was subsequently assessed by analysis of ii. Correlations between independent Scoutmaster GPS™. deviance (McCullagh and Nelder 1986). variables were assessed using simple Where Fumaria was found, both plant Logit transformations ensured that the linear or logistic regressions of each and soil samples were collected whereas modelled probabilities could only take variable against each of the others to where it was not found, only soil samples values between 0 and 1. exclude correlated variables from the were taken. Topographical details includ- The probability P of the occurrence of next step. Variables were considered ing aspect and position in the landscape, the response y, where influenced by an correlated if the regression equation as well as the presence of limestone or independent predictor variable x of sig- was significant and explained at least rocks was recorded. In total 161 sites were moidal shape (with a monotonic increase 70% of the variance. surveyed, 146 of which were agricultural. or decrease) and determined by the pa- iii. Stepwise multiple logistic regressions

Where it was not possible to identify rameters b0 and b1, was expressed by the with all uncorrelated, significant inde- Fumaria seedlings, plants were collected to equation: pendent variables obtained in step (1) grow for later identification at flowering. were calculated in upward and down- (b + b x) e 0 1 ward progression, and significant vari- P(y) = (b + b x) Soil variables 1 + e 0 1 ables were determined by analysis of Soil texture was identified as either sand, deviance. sandy loam, loam, clay loam, light clay or Since y was a binary response variable medium-heavy clay (Northcote 1979). In that could only be 1 or 0, it was possible to Results addition, the following soil chemical pa- calculate the odds of event 1 (presence) by Variation in occurrence of fumitory rameters were determined: pH and elec- dividing its probability by that of event 0 species and environmental factors trical conductivity (EC)(1:5 soil:0.01 mol (absence) at a certain level of x. The change Most agricultural sites (AS) were in New −1 L CaCl2 extraction), K, Na, Ca, Mg, Al between the odds of event 1 at two differ- South Wales (83%) while the majority of and Mn (BaCl2/NH4Cl extraction) with ent levels of x was described by the odds non-agricultural sites (NAS), where all F. subsequent calculation of Effective Cation ratio OR (Neter et al. 996). These authors capreolata samples were found, occurred Exchange Capacity (ECEC) as described showed that, upon transformation, an in South Australia (67%) (Table 1). F. den- by Rayment and Higginson (1992), total OR of 2 between two odds of event 1 at siflora and F. bastardii were by far the most C, total N and total S (Dumas digestion; points with a difference of 10 units of x, common species throughout the survey LECO analysis) and available (Olsen) P means that the odds of event 1 will double area, occurring at slightly more than 50% (micro-Kjeldahl digestion). The ratios with each 10-unit increase of x. The cut-off of all sites in New South Wales and 30 and and proportions: K %ECEC, Na %ECEC, point of P for the occurrence of event 1 was 15%, respectively, of all sites in South Aus- Ca %ECEC, Mg %ECEC, Al %ECEC, also determined to identify prediction er- tralia (Table 1). Both species were mostly Mn %ECEC were calculated as they are rors. For this purpose, those values were found in AS but rarely in NAS (Figure 1, indicators for soil physical properties or obtained where the overall prediction er- Table 1). In contrast, F. capreolata never oc- plant nutritional problems (Edwards 1985, ror was minimized. curred at AS whereas 25% of sites with F. Shaw 1985, Smith 1985). Pre-processing of data determined muralis were NAS (Table 1). F. parviflora whether independent variables should was relatively frequent in South Australia, Climate variables be included in the multiple regressions to but was rare in New South Wales, while F. Climatic data were estimated using the reduce dimensionality and collinearity. officinalis was rare in both States (Table 1). MetAccess® database. Utilizing the GPS- This was necessary because, in some spe- The mean annual rainfall of AS and coordinates, the nearest weather stations cies, the number of independent variables NAS differed only slightly, although their provided rainfall and/or temperature exceeded the number of data points and ranges and monthly distributions varied data for each survey point. because high correlations between some substantially (Table 2). Among the AS, Average annual rainfall was used to x-variables would have led to an ill-con- there was a difference of 40 to 50 mm be- describe the zones where Fumaria was ditioned model with unstable parameter tween the minimum and maximum aver- found. Average monthly rainfall and tem- estimates. Therefore, the following prag- age monthly rainfall over autumn. Most perature values between April 1 and June matic, but widely used (e.g. Grundy and NAS were in South Australia where sev- 30 were also collected as these months de- Mead 2000), three-step approach was ap- eral were located at the foot of the Mt. lineate the period of peak germination of plied using GenStat for Windows (5th edi- Lofty Ranges where more rainfall was re- all Fumaria species. tion, Release 4.2, Payne et al. 1994): ceived later in the season than elsewhere. Plant Protection Quarterly Vol.26(2) 2011 59

Fumaria species F. densiflora F. officinalis F. parviflora 1 2

12

!P CONDOBOLIN !P CUDAL

SNOWTOWN UNGARIE CANOWINDRA !P !P !P

!P KADINA COWRA !P

!P ARDLETHAN BARELLAN !P TEMORA OWEN !P !P

!P HAMLEY BRIDGE MALLALA !P !P JUNEE

!P GAWLER !P LOCKHART WAGGA WAGGA !P CANBERRA P!

ADELAIDE BERRIGAN P! !P

CALLINGTON !P

DOOKIE !P

Figure 1 (a). Approximate areas of the survey together with the collection locations in southern New South Wales/ northern Victoria and South Australia of the narrow-leaved Fumaria densiflora, F. officinalis and F. parviflora.

Moreover, the range of annual rainfall matter contents and elevated levels of total some crops (Edwards 1985). Levels of N, P was much higher in the NAS than the N, available P, ECEC and associated mac- and S ranged from very low to very high AS. However, there was no difference in ro-nutrient cations. In addition, a larger (Table 2). temperatures between AS and NAS dur- proportion of the NAS had a higher aver- Due to their proximity to the sea, the ing the three autumn months when the age Na content due to proximity to the sea, mean altitude of the NAS was substan- temperature range was approximately 4°C high levels of calcium carbonate and the tially lower than that of the AS. Some NAS (Table 2). occurrence of limestone chips (a common were also found on steep banks along While about 50% of all AS were located feature in South Australian soils). This creeks and rivers. For the sites located on on clay soils and approximately 10% on explains the higher Ca and pH of NAS slopes, there was no association between sands, 50% of the NAS had sandy soils soils and the substantially lower Al levels. aspect and fumitory occurrence (Table 1). and 20% were on heavier soils (Table 1). The lower Mn levels found were associ- The differences in soil analyses between ated with the more frequent occurrence of Simple relationships between species and AS and NAS appeared to be related to lighter, well-drained soils among the NAS environmental variables the type of sites where F. capreolata was (Table 2). No toxic levels were observed in The occurrence of all species, except F. of- found, since these had higher amounts any soils, although the value of 1 cmol kg−1 ficinalis, was significantly associated with of leaf litter resulting in higher organic exchangeable Al can be yield-reducing in environmental factors. Because of the 60 Plant Protection Quarterly Vol.26(2) 2011 Fumaria spp.: Broad leaf 2

IRYMPLE !P

" RED CLIFFS !P 3 1 2

Fumaria species F. bastardii F. capreolata F. muralis 13

!P CONDOBOLIN !P CUDAL

SNOWTOWN UNGARIE CANOWINDRA !P !P !P

!P COWRA !P KADINA

BARELLAN !P !P TEMORA ARDLETHAN !P !P OWEN

!P HAMLEY BRIDGE !P MALLALA !P JUNEE

!P GAWLER !P !P WAGGA WAGGA LOCKHART CANBERRA P!

BERRIGAN ADELAIDE P! !P

CALLINGTON !P

DOOKIE !P

Figure 1 (b). Approximate areas of the survey together with the collection locations in southern New South Wales/northern Victoria and South Australia of the broad-leaved fumitory species F. bastardii, F. muralis and F. capreolata.

non-parametric nature of the distribution in June (Table 3). Conversely, F. parviflora In all species except F. bastardii, signifi- functions, all significance levels of χ2 are was found more often in those areas that cant preferences were found for certain given to a probability of 10% (Table 3). The receive higher rainfall later in the season. soil texture groups. F. capreolata preferred five samples of F. officinalis were from en- (χ2 levels + <0.001 in lighter soils (χ 2 level <0.001), F. muralis oc- vironments too heterogeneous for signifi- May and June) and F. capreolata (χ2 lev- curred most often in clay-loams (χ 2 level cant associations (Table 3). els + <0.001 in June) occurred more often 0.007), while F. parviflora was commonly Monthly rainfall during the three au- in the warmer parts of the survey area, adapted to heavy clay soils, although it tumn months was most consistently re- particularly in South Australian coastal or could also be present in light soils (χ 2 lated to the presence of all five Fumaria subcoastal areas. In contrast, F. bastardii level <0.001). F. densiflora was most often species. While F. bastardii, F. muralis and F. preferred cooler, more continental envi- found in either light, sandy loams or light capreolata showed strong positive relation- ronments (χ2 level − 0.053 in June, Table to heavy clays, but there were high prob- ships with rainfall throughout most of that 3) whereas the occurrence of F. densiflora abilities of it occurring in any soil, reflect- period, the presence of F. densiflora was and F. muralis was not correlated with ing the wide-spread distribution of this negatively associated with precipitation temperature. species (Table 3). Plant Protection Quarterly Vol.26(2) 2011 61 Among the five Fumaria species, F. Table 1. Relative occurrence of each of the categorical environmental parviflora and F. capreolata were most asso- descriptors and the frequency of occurrence of Fumaria spp. at all survey ciated with soil chemistry variables, indi- sites (n = 161), those where agriculture occurred (n=146) and those of a non- cating more specialized requirements than agricultural (n = 15) nature. the other three species. The occurrence of these species was positively correlated Variable All sites (%) Agricultural sites (%) Non-agricultural sites (%) with pH, Ca, the presence of limestone, Soil texture group ECEC and all or most macro-nutrient cati- Sands 6.2 4.8 20.0 ons, as well as with total C and N, but neg- Sandy loams 8.7 6.2 33.3 atively with Al and Mn (Table 3). How- ever, while higher levels of K in absolute Loams 13.7 13.7 13.3 terms seemed to be preferred, these had to Clay loams 14.9 15.7 6.7 be balanced by a simultaneous increase in Light clays 30.4 32.2 13.3 all other macro-nutrients, particularly Ca, Medium-heavy clays 26.1 27.4 13.3 as the correlation with K as a percentage Limestone gravel 17.4 12.3 66.7 of the ECEC was negative for both species (Table 3). The other three species showed Rocks 5.0 4.8 6.7 far fewer and usually w eaker associations Position in landscapeA with the soil chemical variables. Plain 55.8 53.8 73.3 The distributions of F. densiflora and F. Slope 40.8 43.9 13.3 muralis were significantly associated with latitude with F. muralis found more often Valley 2.0 2.3 0.0 in the southern part of the survey zone, River/creek bank 1.4 0.0 13.3 while frequency of F. densiflora increased Fumaria sp. toward the north (Table 3). absent 10.6 11.6 0.0 The presence of all species, except F. muralis, was significantly correlated with bastardii 45.3 50.0 6.7 altitude. F. densiflora and F. bastardii in- capreolata 6.8 0.0 73.3 creased in frequency as altitude increased, densiflora 49.7 54.1 6.7 whereas the opposite occurred with muralis 9.9 8.2 26.7 F. capreolata and F. parviflora, though only weakly in the case of F. parviflora (Table 3). officinalis 3.1 3.4 0.0 parviflora 8.7 8.9 6.7 A based on n = 147, 132 and 15 for all, agricultural and non-agricultural sites, respectively.

Table 2. Mean (Av.), minimum (Min.) and maximum (Max.) of the continuous environmental variables describing all (n = 161), agricultural (n = 146) and non-agricultural (n = 15) survey sites. Variable All sites Agricultural sites Non-agricultural sites Min. Av. Max. Min. Av. Max. Min. Av. Max. Latitude (°S) 32.95 34.49 36.37 32.95 34.48 36.37 33.92 34.63 35.67 Altitude (m N.N.) 15 208 565 21 216 565 15 128.2 400.0 Annual rainfall (mm) 300.5 514.9 1095 325.7 514.1 736.6 300.5 522.6 1095 April rainfall (mm) 22.1 41.4 84.5 23.6 41.4 60.3 22.1 41.0 84.5 May rainfall (mm) 27.4 49.6 117.0 27.4 49.0 69.6 29.8 54.9 117.0 June rainfall (mm) 24.1 41.4 143.3 26.9 39.7 75.1 24.1 58.4 143.3 April temp. (°C) 13.8 16.8 18.0 14.6 16.8 18.0 13.8 16.9 17.6 May temp. (°C) 10.9 12.7 14.5 10.9 12.6 14.5 10.9 13.3 14.5 June temp. (°C) 8.0 9.7 12.1 8.0 9.6 12.0 8.6 10.6 12.1 pH 4.10 5.73 8.19 4.10 5.58 7.89 4.52 7.18 8.19 EC (dS m−1) 0.04 0.15 0.89 0.04 0.13 0.89 0.12 0.31 0.66 C (%) 0.37 2.10 8.16 0.37 1.86 6.10 1.32 4.39 8.16 N (%) 0.03 0.16 0.41 0.03 0.15 0.39 0.09 0.25 0.41 S (%) 0.00 0.02 0.38 0.00 0.02 0.20 0.00 0.06 0.38 P (mg kg−1) 4 22 110 4 20 78 11 48 110 ECEC (cmol kg−1) 1.75 12.4 34.0 1.75 11.4 34.0 7.2 22.2 33.3 Ca (cmol kg−1) 1.28 8.92 29.2 1.28 8.11 29.2 4.61 16.8 26.6 Mg (cmol kg−1) 0.23 1.80 6.12 0.23 1.68 6.12 0.77 2.96 4.18 Na (cmol kg−1) 0.00 0.16 2.25 0.00 0.12 1.46 0.01 0.58 2.25 K (cmol kg−1) 0.16 1.38 5.89 0.17 1.34 3.45 0.16 1.80 5.89 Al (cmol kg−1) 0.00 0.06 1.00 0.00 0.06 1.00 0.00 0.01 0.07 Mn (cmol kg−1) 0.00 0.13 0.64 0.00 0.14 0.64 0.00 0.04 0.24 62 Plant Protection Quarterly Vol.26(2) 2011 Table 3. Summary of significant correlations (χ2) between single environmental descriptors and occurrence of Fumaria species in a survey in south-eastern Australia. Environment Agricultural sites (n = 146) All sites (n = 161) descriptor Fumaria bastardii Fumaria densiflora Fumaria parviflora Fumaria muralis Latitude − <0.001 + 0.043 + 0.009 Altitude + 0.047 + 0.037 − 0.082 −0.004 An. Rainfall + 0.004 − 0.006 + 0.006 + 0.012 Rainfall April + 0.001 − <0.001 + 0.016 + 0.010 Rainfall May + 0.043 + <0.001 + 0.003 + 0.037 Rainfall June −0.016 + 0.006 + 0.048 + 0.000 + <0.001 Temp. April − 0.025 + 0.013 Temp. May − 0.015 + <0.001 + 0.004 Temp. June − 0.053 + <0.001 + <0.001 pH + <0.001 + <0.001 EC − 0.087 + <0.001 C% + 0.008 + <0.001 N% −0.100 + 0.003 + <0.001 S% + 0.011 P ppm + 0.051 + <0.001 ECEC − 0.058 + <0.001 + <0.001 Ca + <0.001 + <0.001 Mg − 0.029 + 0.006 − 0.027 + 0.003 Na − 0.017 − 0.024 + 0.004 K + 0.005 + 0.043 Al − 0.002 − 0.002 Mn − 0.004 − <0.001 Ca %ECEC + <0.001 + 0.073 Na %ECEC − 0.089 − <0.001 + 0.014 K %ECEC − <0.001 − 0.005 Al %ECEC − <0.001 − 0.002 Mn %ECEC − <0.001 − <0.001 Ca:Na − 0.014 Ca:K + <0.001 + 0.002 Soil texture 0.010 <0.001 0.016 0.007 <0.001 Limestone − 0.012 + <0.001 −0.069 + <0.001 Semi-shade + <0.001 Position 0.004

Relationships between Fumaria species accumulations of prediction er- 1.0 P = exp(−0.001 + 0.079x)/ (1 + exp(−0.001 + 0.079x)) rors of the same type: Due to bast and multiple environmental variables 0.9 Fumaria bastardii Of all the environ- low April rainfall, F. bastardii 0.8 mental descriptors, April rainfall re- was predicted to be absent from 0.7 mained the only significant predictive all South Australian sites, ex- bast variable in the maximal model of the cept one. However, F. bastardii 0.6 stepwise multiple regressions. Figure 2 il- occurred at seven locations in 0.5 lustrates the probability Pbast of F. bastardii South Australia. In the Ootha/ 0.4 occurring in the survey area as affected Trundle area of New South 0.3 by average April rainfall. The mean of the Wales, F. bastardii was pre- Probability P 0.2 rainfall data is located at the origin and is dicted to occur at each location equivalent to 41.4 mm precipitation (Table sampled, but was only found 0.1 2). Calculation of the odds ratio for x = 10, on rocky hilltops. This regional 0.0 −20 −10 0 10 20 which is e10 × 0.079 = 2.20, showed that the clustering of prediction errors odds of F. bastardii presence in compari- suggests the possibility of sub- April rainfall (mm, centred) son with its absence more than doubled stantial heterogeneity of this with each 10 mm increase in average April species. Figure 2. Effect of April rainfall on the rainfall. probability of Fumaria bastardii occurring in When relating fitted values to ac- Fumaria densiflora Latitude the survey area. tual responses, it became apparent that and the soil parameters Na as in some instances there were regional % ECEC, % N and soil texture Plant Protection Quarterly Vol.26(2) 2011 63 were important in their association with Fumaria parviflora Only limestone with the occurrence of F. muralis irrespec- the occurrence of F. densiflora. The initial presence and soil texture were shown to tive of whether analysed with data just maximal model contained the covariates: have a significant effect on the likelihood from the agricultural sites or whether latitude, altitude, June rainfall, Na, Na as of occurrence of F. parviflora in the multi- analysed with data from all sites. These %ECEC, %N and soil texture. The final ple regression model. The partial deviance analyses were used to derive probability model only contained the first and last for limestone presence (1 d.f.) was 34.0591 curves for its occurrence (Figure 3) where three of these as significant covariates. The with P(χ2) <0.001, and the respective val- the origin represents the mean May rain- probabilities of their respective χ2-values ues for soil texture (5 d.f.) were 12.0317 fall of 49.0 mm in the AS, and the value of were <0.001, 0.007, 0.058 and 0.06. and P(χ2) = 0.034. 49.9 mm is the mean at all sites. Due to the categorical nature of the var- The presence of limestone dramatically The extension of the data set to all sites iable soil texture, the parameter btext takes increased the likelihood of occurrence of F. did not substantially affect the coefficient different values for the various soil texture parviflora (Table 4) with the odds increas- of May rainfall (bAS + NAS = 0.1574 and bAS = groups: ing by a factor of almost 110. In addition, 0.1391) and therefore the shape of the re-

bsands = 0.00 while F. parviflora seemed to occur only gression curves. However, while the curve

bsandy loams = 1.70 rarely on medium-textured soils, it was for AS is only valid in the range between

bloams = −0.18 found as commonly in light, sandy soils about 25 and 70 mm of May rainfall, the

bclay loams = −0.48 as on clay soils. This pattern may indicate curve for all sites covers a range of up to

blight clays = 1.08 the existence of different ecotypes within approximately 120 mm.

bm/h clays = 1.58 the species. Irrespective of whether the site was ag-

If a Pparv cut-off point of 0.04 for the ricultural, F. muralis did not occur much in From these coefficients, the probabil- prediction of presence of F. parviflora was either very light or very heavy soils (Fig- ity and odds of F. densiflora occurring in applied, the overall prediction error was ure 3). Thus, the resulting probabilities of a certain soil type were calculated for the 3.4%. The low cut-off value reflects the low occurrence were in both cases close to zero point where all other variables were held overall occurrence of F. parviflora. across the whole rainfall range (Figure 3). at their mean: The differences in occurrence between

Pdens in sands = 0.310, odds = 0.45 Fumaria muralis Soil texture (Pχ <0.001) the heavier medium-textured soils were

Pdens in sandy loams = 0.711, odds = 2.46 and May rainfall (Pχ <0.001) were left as only small in both data sets. However, due

Pdens in loams = 0.273, odds = 0.38 the only covariates significantly correlated to the different distribution of F. muralis

Pdens in clay loams = 0.218, odds = 0.28

Pdens in light clays = 0.570, odds = 1.33 P in m/h clays = 0.686, odds = 2.18 dens Table 4. The probability of occurrence of Fumaria parviflora in different The probability of occurrence of F. den- soil texture groups with and without the presence of limestone. siflora at the mean of all other significant Soil texture group Soil group − limestone Soil group + limestone covariates was about 2.5 times higher in Sands 0.03 0.79 sandy loams, light clays and medium/ heavy clays than in sands, loams and clay Sandy loams 0.03 0.78 loams. The odds of the presence of F. den- Loams <0.001 <0.001 siflora in sandy loams and clays in com- Clay loams <0.001 <0.001 parison to its absence in those soils were more than 2:1. In light clays the chances of Light clays 0.01 0.53 presence to absence were about equal, and Medium/heavy clays 0.04 0.81 in all other soils less than 0.5:1. The following parameter estimates and odds ratios were obtained for the continu- ous variables: 1.0 (a) (b) blat = −0.678 (P of t = 0.019), clay loams OR = 0.51 0.9 clay loams bNa%ECEC = −0.434 (P of t = 0.043), 0.8 OR = 0.65

bN% = −8.39 (P of t = 0.026), 0.7 sandy loams OR = 0.00023. loams loams

mur 0.6 light clays The latitude covariate shows that, inde- 0.5 light clays pendent of soil type, a geographical move south by 1 degree reduced the odds of 0.4

F. densiflora occurring by a factor of 0.51. Probability P Similarly, an increase of 1% Na expressed 0.3 as %ECEC reduced the odds of occur- 0.2 rence by a factor of 0.65, if all other vari- ables were held constant. However, the 0.1 reduction of odds for each increase of 1% 0.0 in soil total N% was only by a negligible −30 −20 −10 0 10 20 30 40 50 60 −20 −10 0 10 20 30 40 50 60 70 0.00023. May rainfall (mm, centred) May rainfall (mm, centred) There was no regional accumulation of false predictions although half of all occur- rences of F. densiflora in sandy loam soils Figure 3. Probability of the presence of Fumaria muralis as affected by soil came from the Barellan district of New texture and May rainfall at (a) agricultural sites only and (b) at all survey South Wales. sites. 64 Plant Protection Quarterly Vol.26(2) 2011 across the soil texture groups in AS and substantially increased by a factor of al- rates for this species in parts of the present NAS, a substantial change was seen in the most 40 with each unit rise in pH or by 4 survey. estimate for sandy loams in AS and in that for each 10 mm increase in June rainfall, The occurrence of F. muralis, was sig- for all sites: While the bsandy loams coefficient given that all other covariates were held nificantly affected by both soil texture in agricultural sites is very low (−0.6) and at the same level. F. capreolata was usu- and rainfall during autumn, most notably the resulting probabilities predict a near ally found in locations with substantially during May when it was often flowering. absence of F. muralis from those soils in ag- above site-mean values for pH and June In agricultural sites it preferred the me- ricultural situations, the probabilities for rainfall. The shape of the rainfall curve at dium to heavier-textured soils with higher all sites suggest that it occurs in the same the mean of all other variables is similar to rainfall, while in non-agricultural environ- proportions on these soils as on any other those presented for F. muralis. ments it also occurred on lighter soils, pro- of the medium-textured soils (Figure 3). vided they were high in organic matter This was entirely due to the lighter soils in Discussion and in higher rainfall areas. This plant is the NAS, where F. muralis occurred, being The six Fumaria species studied were not relatively uniform and has more clearly characterized by high organic matter. uniformly distributed throughout the sur- defined moisture requirements than the Calculation of OR shows that for each vey zone. Indeed the survey has shown previous two species. Very high rainfall 10 mm increase in May rainfall, the proba- their occurrence to be often only regional and very heavy soils do not seem to suit bility of F. muralis being present increased and influenced by specific edaphic and the plant, possibly due to susceptibil- by a factor of 4 in the AS, and by a factor climatic variables. Soil texture and rainfall ity to waterlogging as reported by Fedde of 4.8 when taking all sites into account. during autumn, the germination and early (1936) and Ellenberg et al. (1992) who, with

The cut-off point of Pmur predicting actual growth period of these species, were the Valdés (1987), state that it mainly occurs presence of F. muralis was determined as most important variables influencing dis- on acidic soils, an observation confirmed being Pmur ±0.22 in all sites and ±0.21 in tribution. Most of these Fumaria species, in the present survey. The soil texture pref- agricultural sites with prediction errors like true agrestal weeds, occurred almost erences of F. muralis make it less likely to being 10.6% and 9.6%, respectively. exclusively in regularly disturbed sites. occur together with F. densiflora, while in However, others, in particular F. capreo- higher rainfall areas it is commonly found Fumaria capreolata The distribution of lata, did not depend on soil disturbance with F. bastardii. this species was significantly correlated and behaved like true wild plants. Inter- A similar habitat has been described with the highest number of covariates, estingly, Pugsley (1919) concluded that all for F. parviflora, which suggests it has very suggesting that it might be more special- Fumaria species could occur as wild plants specific requirements. For example, Fenni ized in its environmental requirements, in certain natural habitats. Similar to our (1993) reports it to be common, some- than the other species. Fifteen of the more observations he specifically mentioned times in mass infestations, on calcareous significant, uncorrelated covariates were F. capreolata (1919, 1927, 1932 and 1934) soils in the 300–800 mm rainfall area of included in the maximal model: altitude, as the species present most frequently in northern Algeria. Pugsley (1912) and Stace June rainfall, June temperature, pH, EC, undisturbed sites throughout Europe and (1997) found it on the heavy soils of the ECEC, %C, mg L−1 P, Na, Mg, Mn, Ca:K, Northern Africa. However, F. capreolata is ‘Chalk Districts’ of Britain whilst Rauh K as %ECEC, limestone presence and soil also reported to occur in cultivated land in and Senghas (1976) noted it in the warm, texture. However, due to the low occur- its native range of distribution (e.g. Pugs- limestone-derived soils of the Rhine and rence of F. capreolata and the strong asso- ley 1912, Susplugas et al. 1975, Soler 1983). Main valleys in . In Pakistan and ciation of its presence with specific levels Fumaria bastardii and F. densiflora were India, it is a serious weed of the heavy, of several covariates, it was not possible the least exacting in their soil texture and calcareous soils of the Punjab (PSSD 1964, to carry out stepwise downward multiple rainfall requirement in this survey ex- Majid and Sandhu 1984, Singh and Singh regressions and only the following sig- plaining their widespread and overlap- 1996). In our survey, F. parviflora was nificant covariates could be fitted in for- ping occurrence. As there are few herbi- commonly found in the lime-rich envi- ward multiple regressions: June rainfall cides capable of controlling Fumaria for ronments, particularly in South Australia (Pχ <0.001), pH (Pχ <0.001), K as %ECEC use in broadleaved crops such as canola, and north-western Victoria. The scarcity (Pχ = 0.003) and %C (Pχ = 0.004). Despite it probably explains why these two species of these soils in New South Wales explains being potentially incomplete, the resulting have become important weeds. F. densi- the rarity of the species in that State. The multiple response curve was highly signif- flora presence was largely unaffected by presence of limestone suggested that the icant (Pχ <0.001) and predicted presence autumn rainfall amount but more influ- associated pH level is not crucial to spe- and absence of F. capreolata with an overall enced by soil texture, consistent with the cies presence, since this varied between 5.1 error of only 4.3%, where Pcapr ≥0.19 meant report by Pugsley (1912) in Britain, where and 7.8 at F. parviflora sites, but rather the presence and Pcapr <0.19 meant absence of it occurs either on sandy loams or calcare- high level of Ca as per cent of ECEC, which the species. ous clays. Large variability was observed typically varied between 70 and 85%. This The estimated parameters and odds ra- in corolla and size, indicating a het- preference for limestone-rich soils makes tios for a one-unit change in the respective erogeneous species with potentially dif- F. parviflora unlikely to accompany F. bas- x value were: ferent bio- and ecotypes. F. bastardii was tardii and F. muralis at the same site. If

b0 = −7.21 present equally on all soil types, although limestone is present, the predilection of F.

brainf6 = 0.1387 (Pt = 0.008) OR = 1.15 it occurred more frequently in areas of parviflora for sandy as well as heavy soils

bpH = 3.69 (Pt = 0.02) OR = 39.9 higher April rainfall. Pugsley (1912) and suggests that it may be a heterogeneous

bK%ECEC = 0.698 (Pt = 0.024) OR = 2.0 Stace (1997) report it to occur exclusively species which is likely to occur together

bC% = 1.127 (Pt = 0.018) OR = 3.1 in the wetter western part of Great Britain with F. densiflora . Indeed, several varieties and Ireland, but in it has a much of F. parviflora were described by Pugs- The probability of F. capreolata occur- larger rainfall range (Soler 1983, Valdés ley (1919) and, although controversial, the ring at the mean of all four variables (i.e. at 1987). Most authors (Pugsley 1912, 1919, most recent classification of the genus also 41.4 mm June rainfall, pH = 5.73, 2.1% total Soler 1983) report large morphological indicates high variability within this spe- C and K representing 12.8% of the ECEC) variability within F. bastardii, which is so cies (Lidén 1986). was only 0.07%. Consequently, the odds of marked that some support its division into In the survey, F. capreolata preferred neu- its presence in comparison to its absence varieties. This high intra-specific variabil- tral to alkaline, moist and well-draining were 0.00073:1. However, these odds were ity may explain the high prediction error soils with high levels of organic matter, Plant Protection Quarterly Vol.26(2) 2011 65 although there was a scattered, minor oc- two species may indicate less specificity Grundy, A.C. and Mead, A. (2000). Mod- currence on some slightly acidic soils in in natural environmental requirements eling weed emergence as function of New South Wales. Previous descriptions and the existence of different ecotypes meteorological records. Weed Science of its habitat refer to neutral to alkaline with as yet incomplete distribution across 48, 594-603. soils in Mediterranean areas (Susplugas all suitable habitats. Therefore, further Harden, G.J. (1990). Fumariaceae. In ‘Flora et al. 1975, Valdés 1987), whereas popula- spread into areas presently unaffected by of New South Wales, Vol. 1’, ed. G.J. tions in Middle Europe seem to be more Fumaria might be expected. However, the Harden, pp. 172-5. (Royal Botanic Gar- associated with acidic soils (Ellenberg et higher prediction error could also mean dens Sydney, Sydney, Australia). al. 1992). It was also reported to be present that factors other than those relating to the Haussknecht, C. (1873). Beitrag zur Kennt- in ‘fresh’, presumably young or alluvial natural environment play a role in spe- niss der Arten von Fumaria sect. Sphae- soils (Valdés 1987, Ellenberg et al. 1992) in cies distribution. As could be expected rocapnos DC. Flora 56, 401-14, 416-25. Europe, or among shady, rocky outcrops in all introduced species, these are likely Kriticos, D.J., Randall, R.P. (2001). A com- in Northern Algeria (Pugsley 1927). Al- to include historical chance events which parison of systems to analyse potential though these descriptions have some com- allowed accidental introduction into one weed distributions. In ‘Weed risk as- monality with those of the sites where F. area but not another. sessment’, eds R.H. Groves, F.D. Pan- capreolata was found in Australia, they etta and J.G. Virtue, pp. 61-82. (CSIRO are not specific enough to explain why F. Acknowledgments Publishing, Melbourne, Australia). capreolata was not detected in cultivated This project was supported by the Grains Lemerle, D., Blackshaw, R., Potter, T., fields. Research and Development Corporation Marcroft, S. and Barrett-Lennard, R. was found at only five of Australia. The extensive advice on Fu- (1999). Incidence of weeds in canola locations. The populations belonged to maria occurrence given by many advisory crops across southern Australia. Pro- three distinct types and it was not possible personnel from both the public and private ceedings of the 10th International Rape- to identify factors affecting its occurrence. sector in New South Wales, South Aus- seed Congress: New horizons for an old However, the heterogeneity of the species tralia and Victoria is gratefully acknowl- crop’. (Groupe Consultatif Internation- here reflects its widespread distribution in edged. Drs Alison Smith and Brian Cullis al de Recherche sur le Colza, Canberra, Europe (e.g. Haussknecht 1873, Pugsley provided invaluable statistical assistance. Australia). 1912, 1919) where it is the most common Lemerle, D., Yuan, T.H., Murray, G.M. and species, occurring in a very wide range References Morris, S. (1996). Survey of weeds and of different climatic environments (Lidén Brown, J.H. (1984). On the relationship diseases in cereal crops in the southern 1986). Closer analysis of its European dis- between abundance and distribution wheat belt of New South Wales. Aus- tribution shows that it primarily occurs in of species. The American Naturalist 124 tralian Journal of Experimental Agricul- areas with young, relatively unweathered (2), 255-79. ture 36, 545-54. and therefore base-rich soils. Although Collingham, Y.C., Wadsworth, R.A., Lidén, M. (1986). Synopsis of Fumarioide- not clear why it is rare in Australia it is Huntley, B. and Hilme, P.E. (2000). Pre- ae () with a monograph possible that the mostly old, weathered dicting the spatial distribution of non of the tribe Fumarieae. Opera Botanica and comparatively infertile Australian indigenous riparian weeds: issues of (Lund) 88, 1-133. soils do not meet its requirements. spatial scale and extent. Journal of Ap- Majid, A. and Sandhu, G.R. (1984). Effects Although soil texture and rainfall, as plied Ecology 37 (Suppl. 1), 13-27. of Fumaria parviflora on yield and yield well as pH in some cases, were the most Corbacho, C., Sánchez, J.M. and Costillo, E. components of wheat. Pakistan Journal important variables affecting the occur- (2003). Patterns of structural complex- of Agricultural Research 5, 141-3. rence of the different species, there were ity and human disturbance of riparian McCullagh, P. and Nelder, J.A. (1986) also several ‘sporadic’ variables. Soil Na vegetation in agricultural landscapes of ‘Generalized linear models’. (Chapman content remained a significant, negatively a Mediterranean area. Agriculture, Eco- and Hall, London). correlated predictor in the multiple re- systems and Environment 95, 495-507. Neter, J., Kutner, M.H., Nachtsheim, C.J. gression model for F. densiflora and F. bas- Edwards, D.G. (1985). Diagnosing toxicity. and Wasserman, W. (1996). ‘Applied tardii, but was positively correlated with In ‘Identification of soils and interpreta- linear statistical models, 4th edn.’ (Ir- the presence of F. capreolata. While not tion of soil data’, pp. 129-44. (Austral- win, Chicago, IL, USA). explicitly reporting Na levels, Ellenberg ian Society of Soil Science (Queensland Northcote, K.H. (1979). ‘A factual key for et al. (1992) support the present findings Branch), Brisbane). the recognition of Australian soils, 4th for F. bastardii and F. densiflora by catego- Ellenberg, H., Weber, H.E., Duell, R., edn.’ (Rellim Technical Publications, rizing all Fumaria species as salt sensitive. Wirth, V., Werner, W. and Paulissen, Adelaide, Australia). Notwithstanding, the range of Na levels D. (1992). ‘Zeigerwerte von Pflanzen in Norton, G.M. (2003). Understanding the detected in the survey sites did not reach Mitteleuropa, 2nd edn.’ (Verlag Erich success of fumitory as a weed in Austral- levels that are normally considered toxic Goltze, Goettingen, Germany). ia. PhD thesis, Charles Sturt University, or growth-inhibiting to crops (Shaw 1985, Fedde, F. (1936). Papaveraceae. In ‘Die Wagga Wagga, Australia. Available at: 1999). natürlichen Pflanzenfamilien (Band http://www.users.on.net/~nortonm/ Latitude was an important variable in 17b)’, eds A. Engler and K. Prantl, pp. (verified 2 December 2010). the model for F. densiflora. The attempt to 5-145. (Duncker und Humblot, Berlin, Payne, R.W., Lane, P.W., Digby, P.G.N., clarify the reason for the importance of Germany). Harding, S.A., Leech, P.K., Morgan, latitude, by correlating this with several Fenni, M. (1993). Inventaire et étude de la G.W., Todd, A.D., Thompson, R., Tun- other environmental variables, was unsuc- flore messicole des haute plaines céréal- nicliffe, W.G., Welham, S.J. and White, cessful. ières sétifiennes (Nord-Est Algerie). R.P. (1994). ‘Genstat™ 5 Release 3 – ref- It could be hypothesized that the re- Mededelingen – Faculteit Landbouwkun- erence manual.’ (Clarendon Press, Ox- sults of this survey indicate that many dige en Toegepaste Biologische Weten- ford, UK). parts of the south-eastern cropping zone schappen, Universiteit Gent 58, 1003- PSSD (1964). ‘Soil survey report on the are suitable for colonization by at least 11. Lahore District.’ (Pakistan Soil Survey one Fumaria species, most likely F. bastardii Grime, J.P. (1979). ‘Plant strategies and Department, Lahore, Pakistan). and F. densiflora. The substantially greater vegetation processes’, 2nd edition. (Wi- Pugsley, H.W. (1912). The genus Fumaria error in predicting the presence of these ley, Chichester, UK). L. in Britain. Journal of Botany 50, 1-76. 66 Plant Protection Quarterly Vol.26(2) 2011 Pugsley, H.W. (1919). A revision of the Taylor, A.C. and Lill, W.J. (1986). Wheat genera Fumaria and Rupicapnos. Journal crop surveys in southern New South of the Linnean Society (Botany) 44, 233- Wales. 4. The response by grain yield 353. and other wheat attributes to weeds. Pugsley, H.W. (1927). Further notes on the Australian Journal of Experimental Agri- genera Fumaria and Rupicapnos. Journal culture 26, 709-15. of the Linnean Society (Botany) 47, 427- Toelken, H.R. (1986) Fumariaceae. In ‘Flo- 69. ra of South Australia, Vol. 1’, eds J.P. Pugsley, H.W. (1932). Further notes on Jessop and H.R. Toelken, pp. 369-73. the genera Fumaria and Rupicapnos. – II. (South Australian Government Print- Journal of the Linnean Society (Botany) 49, ing Division, Adelaide, Australia). 93-113. Valdés, B. (1987.) Fumaria L. In ‘Flora vas- Pugsley, H.W. (1934). Further notes on cular de Andalucía Occidental’, eds B. the genera Fumaria and Rupicapnos. – Valdés, S. Talavera and E. Fernández- III. Journal of the Linnean Society (Botany) Galiano, pp. 140-7. (Ketres Editora S.A., 49, 517-29. Barcelona, Spain). Rauh, W. and Senghas, K. (1976). ‘Schmeil Velthuis, R.G. and Amor, R.L. (1982/83). – Fitschen Flora von Deutschland und Weed survey of cereal crops in south- seinen angrenzenden Gebieten, 86th west Victoria. Australian Weeds 2, 50-2. edn.’ (Quelle and Meyer, Heidelberg, Walsh, N.G. (1994.) Fumariaceae. In ‘Flora Germany). of Victoria, Vol. 3 Dicotyledons – Win- Rayment, G.E. and Higginson, F.R. (1992). teraceae to Myrtaceae’, pp. 73-8. (Inkata ‘Australian handbook of soil and water Press, Melbourne). chemical methods.’ (Inkata Press, Mel- Wells, G.J. and Lyons, F.J. (1979). ‘Weed bourne). survey of Victorian cereal crops, 1977.’ Saab, V. (1999). Importance of spatial scale (Wheat Research Institute, Horsham, to habitat use by breeding birds in Australia). riparian forests: a hierarchical analysis. Woodward, F.I. and Williams, B.G. (1987). Ecological Applications 9, 135-51. Climate and plant distribution at global Sell, P.D. (1964). Fumaria. In ‘Flora Euro- and local scales. Vegetatio 69, 189-97. paea, Vol. 1’, eds. T.G. Tutin, V.H. Hey- wood, N.A .Burges, A.O. Charter, J.K. Edmondson, D.H. Valentine, S.M. Wal- ters and D.A. Webb, pp. 255-8. (Cam- bridge University Press, Cambridge, UK). Shaw, R.J. (1985). Soil salinity. In ‘Identifi- cation of soils and interpretation of soil data’, pp. 145-74. (Australian Society of Soil Science (Queensland Branch), Bris- bane, Australia). Shaw, R.J. (1999). Soil salinity – electri- cal conductivity and chloride. In ‘Soil analysis – an interpretation manual’, eds K.I. Peverill, L.A. Sparrow and D.J. Reuter, pp. 129-44. (CSIRO Publishing, Collingwood, Australia). Singh, G. and Singh, O.P. (1996). Response of late-sown wheat (Triticum aestivum) to seeding methods and weed-control measures in flood-prone areas. Indian Journal of Agronomy, 41, 237-42. Smith, G.D. (1985). Soil factors affecting soil physical properties. In ‘Identifica- tion of soils and interpretation of soil data’, pp. 175-96. (Australian Society of Soil Science (Queensland Branch), Bris- bane, Australia). Soler, A. (1983). Revision de las especies de Fumaria de la Peninsula Iberica es Islas Baleares. Lagascalia 11, 141-228. Stace, C. (1997). ‘New Flora of the British Isles, 2nd edn.’ (Cambridge University Press, Cambridge, UK). Susplugas, J., Massa, V., Susplugas, P., Taillade, R., Susplugas, C. and Salabert, J. (1975). Fumeterres en Languedoc– Roussillon. Anales d’Institute Botanique des Cavanilles 32, 233-9.