Soil types will alter the response of arable agroecosystems to future rainfall patterns

Authors


Abstract

Climate change scenarios for central Europe predict fewer but heavier rains during the vegetation period without substantial changes in the total amount of annual rainfall. To investigate the impact of rainfall patterns derived from regionalised IPCC scenarios on agroecosystems in Austria, we conducted an experiment using 3 m2 lysimeters where prognosticated (progn.) rainfall patterns were compared with long-term current rainfall patterns on three agriculturally important soil types (sandy calcaric phaeozem, gleyic phaeozem and calcic chernozem). Lysimeters were cultivated with field peas (Pisum sativum) according to good farming practice. Prognosticated rainfall patterns decreased crop cover, net primary production (NPP) and crop yields, but increased root production and tended to decrease mycorrhization. Soil types affected the NPP, crop density and yields, weed biomass and composition, as well as the root production with lowest values commonly found in sandy soils, while other soil types showed almost similar effects. Significant interactions between rainfall patterns and soil types were observed for the harvest index (ratio crop yield versus straw), yield per crop plant, weed density and weed community composition. Abundance of the insect pest pea moth (Cydia nigricana) tended to be higher under progn. rainfall, but was unaffected by soil types. These results show that (a) future rainfall patterns will substantially affect various agroecosystem processes and crop production in the studied region, and (b) the influence of different soil types in altering ecosystem responses to climate change should be considered when attempting to scale-up experimental results derived at the plot level to the landscape level.

Introduction

Human-induced global warming is expected to affect the frequency, timing and intensity of precipitation (IPCC, 2007). In Europe, this will lead to higher winter precipitation in northern regions with drier and hotter summers in central Europe (Viner et al., 2006; Eitzinger, 2010) Accordingly, the regional climate scenario for eastern Austria (pannonian region), an important arable crop region, predicts fewer but heavier rains during the vegetation periods without substantial changes in the annual amount of rainfall (Eitzinger, 2010). We therefore conducted a lysimeter experiment to see how reduction in precipitation on different soil types will affect agroecosystem parameters.

Agroecosystem parameters affected by reduction in precipitation and associated drought from past studies include crop production, weed, arbuscular mycorrhizal fungi (AMF) and insect abundance. It has been demonstrated that reduction in precipitation reduces crop production (Martin & Jamieson, 1996; Sanchez et al., 2001), increases weed invasibility (Kreyling et al., 2008), reduces the mycorrhization rates (Augé, 2001; Porcel et al., 2003, Smith & Read, 2008) and increases insects and invertebrate population (Ziska & Dukes, 2011). Weeds are expected to be less reduced by droughts than cultivated plants owing to their wide climatic or environment tolerance, short generation time, small seed size, uniparental reproduction capacity, high competitive ability, high growth rate and phenotypic plasticity (Whitney & Gabler, 2008; Clements & Ditommaso, 2011). The AMF has also been shown to alleviate drought (Augé, 2001; Porcel et al., 2003), improving the crop yield and water use efficiency (Bolandnazar et al., 2007), while making plants less vulnerable to withstand various abiotic stresses (Koltai & Kapulnik, 2010). As AMF foster plants nutrient uptake it equally increases the soluble protein content improving plants quality for herbivores (Subramanian & Charest, 1998). Reduction in precipitation increases insect mortality, affecting its abundance, morphology and physiology (Moran et al., 1987; Robinson et al., 2012).

Reduction in precipitation was tested on the soil types calcaric phaeozem (S-soils), gleyic phaeozem (F-soils) and calcic chernozem (T-soils), representing 80% of the agricultural soil in Austria's most fertile region (region of Marchfeld). S-soils are highly sandy, with very low profile water and evaporation; F-soils have very high clay content, highly mottled subsoils, with high profile water and evaporation; while T-soils are highly silty with the highest profile water content (Table 1). Soil types and its characteristics have been demonstrated to affect several processes in agroecosystems, such as the availability and supply of water to plants (Passioura, 1991), respiration and soil temperature (Koizumi et al., 1999), plant growth, vegetation cover and yield (Mako et al., 2008; Bestland et al., 2009; Genxu et al., 2009), the transfer and interaction of mineral nutrients (Echevarria et al., 2003; Matias et al., 2011), and the physical, chemical and biochemical properties of soils (Rhoton et al., 1993; Paz-Ferreiro et al., 2011), while higher soil sand content has been shown to improve AMF colonisation (Zaller et al., 2011).

Table 1. Characteristics of the experimental soil types in the lysimeters (from soil analysis and partly from Steinitzer & Hoesch, 2005)
ParametersSandy Calcaric Phaeozem (S-soils)Gleyic Phaeozem (F-soils)Calcic Chernozem (T-soils)
  1. CAL, calcium acetate-lactate method; EDTA, Ethylenediaminetetraacetic - acid.
Profile water content (mm)250–500400–700460–730
Infiltration (mm)430250
Evaporation (mm)280031503150
pH value: CaCl27.47.67.6
Calcium carbonate (%)0.1430.2600.106
Phosphor, CAL (mg kg−1)1437376
Potassium, CAL (mg kg−1)187246286
Magnesium, available (mg kg−1)83273277
Humus content2.14.94.9
Nitrogen, mineralisation (mg kg−1 7 days−1)565768
Boron, available (mg kg−1)1.32.72.9
Iron, EDTA (mg kg−1)694439
Manganese, EDTA (mg kg−1)813433
Copper, EDTA (mg kg−1)3.33.43.2
Zinc, EDTA (mg kg−1)4.64.64.7
Sand (%)67.921.522
Silt (%)19 506755
Clay (%)9.927.8323
Cation exchange capacity, mmol/10011.2925.1326.00

Both the effect of precipitation and soil types on agroecosystem processes have been studied in isolation; however, it is unclear how these important factors interact. On the basis of previous findings we hypothesize that soil types with lower water holding capacities and/or nutrient availability like S-soils will interact with lower precipitation disrupting nutrient and water transportation into and within the plants, slowing down plant's physiological activities like photosynthesis, subsequently reducing plants biomass, crop yields and weeds abundance more than F-soils and T-soils with higher soil water content. Thus, S-soils are expected to have much lower groundcover and root biomasses, which are indicators for soil water und nutrient availability. In this study, we tested the effects of current long-term average rainfall patterns versus future prognosticated rainfall patterns based on regionalised global climate change models simultaneously on three different soil types in a large-scale lysimeter facility.

Materials and methods

Experimental site

This experiment was carried out in 2011 using 18 cylindrical steel (Cr/Ni 18/9) lysimeters each with a surface area of 3.02 m2 and 2.45 m soil depth. Lysimeters were located in Vienna, Austria and situated under a 10 m × 46 m tunnel covered with transparent polyethylene film (Fig. 1). Tunnels were open at the front and back and had 2-m-high openings at both length sides to allow proper ventilation.

Figure 1.

Above ground (A) and below ground (B) view of the lysimeter station where this study was conducted.

The soil types (S-soils, F-soils and T-soils) were filled each into six lysimeters. These soil profiles were carefully excavated from field sites and filled into the lysimeters with their natural bulk density of 1.4 g cm−3. Each soil type was analysed in the laboratory and their different characteristics are reported in Table 1.

Half of the lysimeters were subjected to the current rainfall pattern (treatment ‘curr.’) and the other half to the prognosticated rainfall pattern (treatment ‘progn.’). The current rainfall pattern was calculated by averaging the amount and frequency of precipitation between the years 1971 and 2000 from a weather station located 10 km from the experimental site. The prognosticated rainfall pattern was based on the regionalisation of the IPCC, 2007 climate change scenario for the period 2071–2100, gotten from local climatology and climate change signal from the ensemble mean of the regional climate model scenarios from the EC-project ENSEMBLES (Christensen & Christensen, 2007). The weather generator LARS-WG version 3.0 (Semenov & Barrow, 2002) was used to transfer the derived local climate change signals to daily precipitation rates. Rainfall amounts were applied early in the morning by a hand sprinkler with an attached gauge using tap water. Rainfall treatment started 63 days after seeding whereby the prognosticated treatment received 1/3 reduced precipitation in longer duration (Fig. 2). At the bottom of each lysimeter containers collected the leachate, of which the chemical data are presented in Table 2.

Figure 2.

Current and prognosticated rainfall amounts applied onto field pea stands during the vegetative period from May to July 2011.

Table 2. Chemistry of the flow through water (leachate) on S-soils (except sample 25 on T-soil)
SamplesRainfallSoil TypePermeability (µS cm−1)Phosphorus (mg L−1)NH4 (mg L−1)NO3 (mg L−1)
  1. Curr., current; Progn., prognosticated.
1Curr.S5600.120.2017
2Curr.S6200.070.0026
3Curr.S6000.050.2321
4Curr.S4900.060.1414
5Curr.S5100.070.3116
6Curr.S5700.070.0022
7Curr.S4800.050.0013
8Curr.S5000.060.0015
9Curr.S6000.050.0026
10Curr.S4300.020.1412
11Curr.S4500.010.3916
12Curr.S5700.020.3229
13Progn.S6200.060.4923
14Progn.S6200.040.0023
15Progn.S8200.080.0449
16Progn.S6900.060.0635
17Progn.S6300.050.0127
18Progn.S7500.100.0047
19Progn.S6700.050.0035
20Progn.S5900.050.0924
21Progn.S8100.070.0053
22Progn.S6000.020.2433
23Progn.S5200.020.0723
24Progn.S8400.050.4559
25Progn.T25000.060.0025

Lysimeters were sown with field peas (Pisum sativum cf. Jetset) on 23 March 2011. All lysimeters received 49 mm of natural rainfall before the simulation treatments started. Until the harvest of the crops on 3 July 2011, curr. treatments received 131 mm and progn. treatments 93 mm rainfall. As our aim was to mimic the real-world situation for farmers, we analysed soil P and K nutrient concentrations in all lysimeters and fertilised the lysimeters according to official recommendations. Therefore, all F-soils and T-soils received 65 kg ha−1 P2O5 and all S-soils 100 kg ha−1 K2O, all F-soils and T-soils 50 kg ha−1 K2O. No N fertiliser was applied as lysimeters were planted with the nitrogen-fixing legume alfalfa (Medicago sativa L.) in the year prior to this experiment. No chemical weed control was applied during the course of the experiment.

Measurements

The groundcover was measured from images taken with a digital camera on a tripod located 1.6 m above a marked area (1.2 m × 1.2 m) per lysimeter. We took images every week between days 13 and 70 after sowing and calculated percent ground cover using the freely available software ImageJ (http://rsbweb.nih.gov/ij/). The leaf area index (LAI) was measured using a ceptometer (SunScan type SS1, DELTA-T Device, Cambridge, UK), inserted eight times horizontally (2 cm above the soil surface) from outside to the centre in sections of 45°; LAI was calculated by averaging the eight readings per lysimeter.

Root production was measured by using five randomly located ingrowth cores per lysimeter (diameter 5 cm and depth 20 cm). First, roots present in the soil cores were sorted out and the rootless soils refilled back into the bored holes. Then, 49 days later the same positions were resampled and all roots growing into these cores were washed out in a sieve (mesh size 0.5 mm) under a jet of tap water. Root-free soil was refilled and ingrowth cores were resampled after another 30 days and processed as described above. Of these roots one half was used to determine dry mass after oven drying at 50°C for 48 h. The other half of the root mass was stored in 50% ethanol and their colonisation with vesicular-AMF was measured after ink staining (Vierheilig et al., 1998) using a modified gridline intersection method under the dissecting microscope by counting at least 100 sections (Giovannetti & Mosse, 1980).

Weed infestation was measured on permanently marked 50 cm × 50 cm plot per lysimeter. Weeds growing within this area were successively removed, identified to plant family level, counted and their mass weighed after drying at 50°C for 48 h; weeds growing on the remaining lysimeter area were pulled by hand and weighed. Total weed biomass of each lysimeter was calculated by adding the biomass of the permanent and the remaining plot area.

Field pea plants and weeds were harvested by hand cutting them 5 cm above the soil surface. Pea yield was obtained by threshing the sheets in the laboratory. Field peas and straw were ground and N content was determined using an elemental analyser (LECO TruMac, St. Joseph, MI, USA). Crop P, K and Mg contents were determined by inductively coupled plasma atomic emission spectroscopy (ICP-AES, Thermo Scientific, iCAP 6000 series, Waltham, MA, USA).

No insecticide was used and insect pest population was determined by direct sampling. At harvest abundance of the Pea moth Cydia nigricana Fabricius (Lepidoptera: Tortricidae) was counted in pea sheets on 10 randomly chosen crop individuals per lysimeter. Pea moth abundance per m2 was calculated by multiplying the abundance plant−1 with crop density.

Statistical analyses

First, we tested the normal distribution and variance homogeneity of each parameter using the Shapiro test and Kolmogorov–Smirnov, respectively. Parameters that did not meet criteria for parametric tests were transformed using Boxcox transformations. Afterwards, all parameters (total biomass, pea, straw, weed, harvest index, plant density, root production, mycorrhization, root–shoot ratio, weed families and number of pea moth) were analysed using a two factorial analysis of variance (ANOVA) with precipitation (two levels: curr. rainfall versus progn. rainfall) and soil types (three levels: F-soils, S-soils and T-soils) as factors. We also performed correlations between LAI and biomass (Spearman's rank correlation coefficient) and pea yield and root production (Pearson's correlation coefficient). All statistical analyses were performed using the freely available software R (Free Software Foundation, Inc., Boston, MA, USA; www.r-project.org).

Results

Future rainfall pattern reduced the net primary production (NPP), weed abundance, pea biomass and yield more on S-soils than on T-soils and F-soils. The NPP, harvest index, pea biomass plant−1, pea yield plant−1, root production and root-to-shoot ratio were significantly affected by both rainfall and soil types (Table 3). The NPP under progn. rainfall was 29% lower than under curr. rainfall patterns. The NPP of S-soils was 36% lower compared with F-soils and 43% lower than T-soils, but insignificant between F-soils and T-soils biomass (Fig. 3A). Groundcover was significantly reduced under progn. rainfall, however unaffected by soil types; LAI was marginally significantly affected by rainfall and soil type (Table 3).

Table 3. Analysis of variance results on effects of three different soil types (gleyic phaeozem—F-soils, sandy calcaric phaeozem—S-soils and calcic chernozem—T-soils) and rainfall patterns (curr. rainfall versus progn. rainfall) on agroecosystem variables in field peas
VariableSoil TypeRainfallSoil Type × Rainfall
FPFPFP
  1. DAS, days after seeding; LAI, leaf area index.
Ground cover (70 DAS, %)1.3590.29416.4740.0020.5560.588
LAI (90 DAS)3.3910.0683.2860.0950.9300.421
Net primary production (g m−2)28.676<0.00113.30.0030.9370.419
Pea + straw (g m−2)4.4920.0352.9510.1110.7540.491
Pea (g m−2)12.4860.0013.9790.0690.0810.922
Weed (g m−2)9.6020.0032.6920.1270.2280.800
Harvest index119.093<0.0015.8370.03312.2850.001
Plant density (ind. m−2)8.5740.005    
Biomass per plant (g)31.522<0.00111.0930.0061.6160.239
Pea per plant (g)71.842<0.00117.5670.0014.4590.036
Root production, pretreatment (g m−2)16.590<0.001    
Root production, treatment (g m−2)11.7690.0018.4380.0130.4490.648
Mycorrhization, pretreatment (%)1.5830.238    
Mycorrhization treatment (%)0.1210.8873.7360.0770.1930.827
Root/shoot ratio23.508<0.00120.4270.0011.2650.317
Weed density (ind. m−2)0.8630.4470.0000.9873.7750.053
Pea moth infestation (ind. m−2)0.0790.9251.7360.2120.0770.926
Figure 3.

Net primary production (A), biomass of field pea + straw (B), pea yield (C) and root production (D) in field peas at different soil types (gleyic phaeozem—F-soils, sandy calcaric phaeozem—S-soils and calcic chernozem—T-soils), under current and prognosticated rainfall patterns. Means ± SD, n = 3.

Pea biomass, yield and weed biomass were significantly affected by soil types, but not by rainfall patterns (Table 3). Harvest index, pea yield per plant and weed density showed significant interactions between soil types and rainfall (Table 3). The pea and straw biomass production in S-soils was 37% lower than F-soils and 35% lower than T-soils (Fig. 3B). The pea yield per m2 was significantly affected by soil type and marginally significantly affected by rainfall; S-soils produced the lowest pea yield being 63% lower compared with F-soils and 59% lower than T-soils; F-soils and T- soils had similar yields (Table 3, Fig. 3C). Root production before implementing rainfall treatments was significantly different between soil types (Table 3): S-soils showed higher root production than F-soils and T-soils, whereas the root production between T-soils and F-soils was similar (data not shown). One month after implementing the rainfall treatments, root production was significantly affected by rainfall and soil types (Table 3; Fig. 3D). Across all soil types the root growth under progn. rainfall was on average 53% higher than under curr. rainfall patterns. Root growth was significantly different among the various soil types. Between all soil types the root-to-shoot ratio differentiated significantly (data not shown).

The harvest index was 9% lower under progn. rainfall patterns than under current rainfall patterns. S-soils had the lowest harvest index and differentiated from the T-soils and F-soils by 39% and 42%, respectively, whereas T-soils had 6% lower harvest index than F-soils (Fig. 4). Root mycorrhizal colonisation rate was on average 22%; however, it was not affected by soil types; progn. rainfall showed a trend towards lower mycorrhization rates compared with curr. rainfall (Table 3; Fig. 5).

Figure 4.

Harvest index in field peas at different soil types (gleyic phaeozem—F-soils, sandy calcaric phaeozem—S-soils and calcic chernozem—T-soils), under current and prognosticated rainfall patterns. Means, n = 3.

Figure 5.

Mycorrhization of field pea roots at different soil types (gleyic phaeozem—F-soils, sandy calcaric phaeozem—S-soils and calcic chernozem—T-soils), under current and prognosticated rainfall patterns. Means ± SD, n = 3.

Weed production was significantly affected by soil types, with a trend towards decreasing weed production under progn. rainfall (Table 3). S-soils had 50% less weed biomass than T-soils and 34% less weed biomass than F-soils. Weed density was unaffected by rainfall or soil types (Fig. 6A). Weed communities consisted of the families Asteraceae, Chenopodiace, Polygonaceae and Poaceae. The relative contribution of these families to the weed community was unaffected by rainfall or soil types (Fig. 6B), although there were considerable changes in the contribution of these families to the weed communities.

Figure 6.

Absolute (A) and relative (B) abundance of weed families per m2 in field peas at different soil types (gleyic phaeozem—F-soils, sandy calcaric phaeozem—S-soils and calcic chernozem—T-soils), under current and prognosticated rainfall patterns. Means, n = 3.

Across soil types, the abundance of pea moth (C. nigricana) was on average 105% higher under progn. rainfall than under curr. rainfall; however, this was not statistically significant; soil types had no influence on C. nigricana (Fig. 7).

Figure 7.

The abundance of pea month per m2 in field peas at different soil types (gleyic phaeozem—F-soils, sandy calcaric phaeozem—S-soils and calcic chernozem—T-soils), under current and prognosticated rainfall patterns. Means, n = 3.

Leaf area index significantly correlated with pea biomass (r = 0.724, P = 0.024). There was no correlation between mycorrhization rate and the pea yield, root production or NPP (data not shown).

Analysing the soil NH4 and NO3 contents from 0.1 M KCl soil extract showed strong increase in average NO3 content from 0.416 to 2.225 µg g−1 on S-soil under prognosticated climate (Table 4). By the end of the experiment no leachate was collected on F-soils and barely one sample on T-soils, whereas on S-soils the leachate average NO3 content was almost twice as much with progn. treatment, while NH4 and P contents were almost the same for both treatments (Table 2).

Table 4. The average and standard deviation (SD) values of NH4 and NO3 content in the soil types (gleyic phaeozem—F-soils, sandy calcaric phaeozem—S-soils and calcic chernozem—T-soils), under current and prognosticated rainfall patterns; means ± SD, n = 3
RainfallSoil TypeNH4 (µg g−1)SD—NH4 (µg g−1)NO3 (µg g−1)SD—NO3 (µg g−1)
CF1.5970.1472.3142.543
S1.4440.1140.4160.271
T1.7280.1071.5641.715
DF1.4770.4112.7701.780
S1.3310.3492.2250.746
T1.3910.3061.9811.047

Discussion

Effects of rainfall patterns

Simulated future rainfall patterns with 30% decreased rainfall amount during the vegetation period and 36% longer dry periods between rainfall events than the current long-term rainfall patterns affected several important processes within this agroecosystem. It was very interesting to observe most changes just about 4 weeks after implementing treatments, which differ by only 38 mm rainfall. Field pea stands responded to future rainfall patterns with a reduced ground cover and aboveground production but increased root production. The allocation of production into roots is probably a stress reaction counteracting the induced drought by increasing the root surface area of water absorption likewise extending deeper to meet the underground available water (Masilionyte & Maiksteniene, 2011).

We attribute the reduction of NPP under progn. rainfall patterns to differences in the soil profile water content, infiltration and evaporation rates (Steinitzer & Hoesch, 2005). In a lysimeter experiment with seven different crops including field pea, it was shown that the straw yield responded positively to moisture with a 21% increase for pea straw biomass under irrigation (Gan et al., 2009). The positive correlation between NPP and the LAI showed a stronger effect of the climate on the vegetative growth and confirms findings that induced drought being responsible also for the reduction in crop cover rate (Cui & Nobel, 1992; Augé, 2001; Echevarria et al., 2003; Porcel et al., 2003; Matias et al., 2011). Decrease in harvest index under progn. rainfall was in contradiction to the findings of Martin & Jamieson (1996) associating increase in field pea harvest index to sensitivity in reproductive growth, but is in conformity with the findings of others attributing it to photosynthetic changes (Sanchez et al., 2001).

The observed increased root growth on all soil types under progn. rainfall indicates that soil conditions in the three soil types were still suitable for root extension (Passioura, 1991; Feiziene et al., 2011). Overall root AMF colonisation was low, suggesting that AMF is not very important in this leguminous crop. Nevertheless, a trend towards reduced AMF colonisation under progn. rainfall could be attributed to the fact that water stress causes plants to be more metabolically perturbed. According to Augé (2001), the fungus strongly competes for root allocates with the onset of stress, leading to reduced mycorrhization rates in response to resist drought stress (Stahl & Christensen, 1982; Cui & Nobel, 1992; Subramanian & Charest, 1998; Augé, 2001; Bolandnazar et al., 2007). The reduced AMF trend observed on all soil types with reduced progn. rainfall could be attributed to reduced soil water content (Stahl & Christensen, 1982), contradicting the findings of Cui & Nobel (1992) who associated higher colonisation with improved water availability.

Overall, there were very few insect pests on the crops in the experimental year. Nevertheless, considerably more C. nigricana were collected under progn. rainfall than under curr. rainfall. Although this difference was not statistically significant owing to high variation between lysimeters, this indicates that pest species living in sheets benefitted from future rainfall patterns. It has long been known that pea moth is more abundant on pea varieties with later flowering dates and longer flowering duration (Nolte & Adam, 1962) and it could also be shown that the abundance of this pest species also correlates with the pea cropping area in the surroundings (Thoeming et al., 2011) as known for other crops (Zaller et al., 2008).

Effects of soil types

The three soil types differed mainly in sand, silt, clay and humus contents, soil water capacity and cation exchange capacities. Overall, crops and weeds in sandy soils were most sensitive to rainfall manipulations. A general decrease in crop production, mycorrhization and insect pests on soil types with progn. rainfall was observed, while the response of weeds varied among soil types. Weed biomass production was unaffected by rainfall patterns, confirming the findings of Gan et al. (2009). Weed abundance decreased in S-soils and T-soils under progn. rainfall but was unaffected on F-soils, indicating that soil types with higher sand and silt content are more prone to reduced rainfall than those with higher clay content. On plots with progn. rainfall weed density increased on F-soils, decreased on S-soils, while T-soils were almost unaffected. This could be attributed to the soil properties, the types of weed species and a better water use efficiency of individual weed families (Bolandnazar et al., 2007). Soil types influencing the abundance of plant communities are in conformity with most earlier research (Koizumi et al., 1999; Echevarria et al., 2003; Mako et al., 2008; Bestland et al., 2009; Clark et al., 2011; Feiziene et al., 2011; Matias et al., 2011), but differ from the finding of Kreyling et al. (2008) that changes in the physical environment had the same effect on vegetation type and diversity level.

In this experiment, S-soils had the lowest soil moisture and highest sand content, but here the smallest AMF reduction was observed, implying that soil moisture alone cannot be the reason for the reduced AM trend. This is somewhat unexpected as a higher sand content has been shown to increase root colonisation with AMF (Zaller et al., 2011).

Weed production was significantly different between soil types although the contribution of different weed families to the weed community was not different. There was a trend towards more abundance of Asteraceae, Chenopodiace and Polygonaceae under progn. rainfall on F-soils and T-soils; however, on S-soils they were all reduced. It appears that weed families with a broader root system are more competitive than the cultivated P. sativum; this could also be reflected in the shift of the ratio of weed-to-crop biomass towards weeds. The abundance of the Poaceae family differed from the other weed families. It increased on F-soils, decreased on T-soils and was almost unchanged on S-soils, under progn. rainfall. This could be attributed to its C4-photosynthetic pathway, compared with C3 pathway of other species.

No leachate on F-soils and barely one sample from T-soils could be attributed to their higher water holding capacities and lower sandy contents than S-soils, indicating S-soils vulnerability to climate change.

Conclusions

To our knowledge, the results of this study demonstrate for the first time that different soil types can alter the impact of rainfall patterns on agroecosystem processes. The influence of different soil types in altering ecosystem responses should be considered when trying to scale-up experimental results derived at the plot level to the landscape level. These results also indicate that crops such as field peas where irrigation as an adaptation to climate change is economically not feasible may be especially prone to future rainfall patterns.

Acknowledgements

We are grateful to Lina Weissengruber and Nadja Santer for their help in the field, to Wolfgang Holzner and Gerhard Karrer for their advice on weeds and to Wolfgang Wanek for lending us the ceptometer. This research was funded by the Austrian Climate and Energy Fund as part of the call ‘ACRP’.

Ancillary