12.A landscape model of arable farmland: implications for the impacts of agricultural management on farmland birds
12.1Introduction
Through a public service agreement, the UK government has a commitment to reversing the long-term decline in the number of farmland birds by 2020, as measured annually against underlying trends. As discussed (Objective 2), the decline in farmland birds is associated with a variety of past agricultural developments. Further agricultural developments that might impact on farmland bird populations are also envisaged, notably the introduction of genetically-modified, herbicide tolerant (GMHT) crops and, potentially, reforms of agricultural policy. Ideally, any such changes should be compatible with the government’s public service agreement, or should be ameliorated by other changes that are.
Potential changes in agricultural management are likely to have effects on the environment over large spatial and temporal scales. Consequently, field trials or experiments conducted at the scale of single fields or even single farms are unlikely to provide accurate indications of what those effects might be. In Objective 2, we discussed the potential of simulation modelling to contribute to predicting the response of bird populations to agricultural developments on a landscape scale. We concluded that data required to parameterise simulation models of birds foraging in farmland are sparse and that behavioural depletion models represent the best available methodology for such simulations.
Here, we present a landscape model of arable farmland. It is not possible to model every component of the rural landscape, or every potential source of food for farmland birds. Instead, we restrict our model to the arable field component of the landscape, and our conclusions to the consequences of different management of those fields for the weeds or birds utilising them. The parameters underlying weed populations within crops and stubbles are based on a comprehensive review of arable weed dynamics (see Objective 4). Owing to problems with modelling bird foraging already discussed (Objective 2), we present results in terms of the consequences of agricultural practices for both weed populations and bird populations. The dynamics of arable weeds are well understood and, as such, the consequences of different management for weed populations may be predicted with greater confidence than the consequences for bird populations. We present the main model in relation to data on dynamics of weeds within crops. To illustrate the utility of the model, we also present the results of altering three different aspects of management. These are: the proportions of land under different types of rotation; the proportions of land left as stubble over winter; and the use of GMHT winter rape and sugar beet.
12.2Methods
12.2.1crop rotations and weed dynamics
A model of an arable landscape was constructed, using the base unit of a single field. The number of fields and size of each field was flexible and, hence, the total area of the landscape was also flexible. Each field could be assigned to one of three rotations, chosen to represent a variety of ‘typical’ systems of arable management practised in lowland Britain (Fig. 12.1). The main model proceeded in weekly time steps (Fig. 12.2), during which the management of each field varied according to the rotation assigned to it. Weed dynamics were modelled each week (Fig. 12.3). The model was not spatially explicit and, hence, the term “field” is loose, and could equally refer to the total area of land under a given form of management, with given densities of crops and weeds. This area might be contiguous, or might be the sum of all equivalent areas within the landscape.
Rot- / Year / Weekation / 1 / 2 / 3 / 4 / 5 / 6 / 7 / 8 / 9 / 10 / 11 / 12 / 13 / 14 / 15 / 16 / 17 / 18 / 19 / 20 / 21 / 22 / 23 / 24 / 25 / 26 / 27 / 28 / 29 / 30 / 31 / 32 / 33 / 34 / 35 / 36 / 37 / 38 / 39 / 40 / 41 / 42 / 43 / 44 / 45 / 46 / 47 / 48 / 49 / 50 / 51 / 52
1 / Winter oilseed rape / Winter wheat
1 / 2 / Winter wheat / Winter wheat
3 / Winter wheat / Winter oilseed rape
1 / Sugar beet
2 / Spring barley
2
3 / Spring peas / Winter wheat
4 / Winter wheat
1 / Spring barley / Winter oilseed rape
3 / 2 / Winter oilseed rape / Winter wheat
3 / Winter wheat
Fig. 12.1. The three rotations modelled included two 3-course and one 4-course management systems. Shaded areas show periods during which crops were in the field. Between these there was no crop in the field and it could either be cultivated in preparation for the next crop, or left as a stubble.
Fig. 12.2. Main model flow chart. Simulations were run for 1,000 years. Weed population dynamics were given 900 years to stabilise before birds were introduced to the model. Weed dynamics (mortality, control, seed set and field cultivation) were modelled in all weeks where appropriate for a given rotation. Bird foraging occurred each week only during winter and, thus, bird populations were augmented by reproduction in the final week of autumn. Where a given population of birds failed to survive through the winter, the population was reduced by 10 and the winter was restarted with all other start conditions identical. This process was iterated until the start population survived the whole winter.
Fig. 12.3. Flow chart for weed dynamics weekly submodel.
a Whether or not a stubble lasted overwinter was determined stochastically. Any stubble that reached its fourth week in autumn or winter, and which was not scheduled to be succeeded by another crop before spring, had a fixed chance of becoming an overwinter stubble. In such cases, the field was left uncultivated until early spring and weeds were permitted to mature and set seed within the stubble.
The dynamics of six arable weed species were modelled. These were fat hen Chenopodium album, annual meadow grass Poa annua, chickweed Stellaria media, black bindweed Fallopia convolvulus, poppy Papaver rhoeas and blackgrass Alopecurus myosuroides. A review of bird diets (see Objectives 3 and 5) indicated that the first four of these were important sources of bird food during winter, whilst the latter two are known to be of commercial importance, due to their influence on the herbicide management of fields. The dynamics of these weeds are described in detail elsewhere (see Objective 4) but parameters used in the model are summarised in Table 12.1. Seed production by arable weeds in stubbles is typically much lower than in crops. Data on seed production in stubbles were available from <???> and were analysed to determine seed production functions (Table 12.2). The timing of seedset in overwinter stubbles is important. The proportions of different weeds maturing and setting seed in stubbles was available from an ongoing study conducted by the BTO (L. Robinson & P. Atkinson, unpublished data) and was modelled as shown in the schematic (Fig. 12.4).
A five-year, multi-site, MAFF-Link project on Integrated Farming Systems (IFS) was run by ADAS during the 1990s, in order to assess the potential benefits of integrated cropping sequences and husbandry techniques. Data on the abundance in spring of different weed species in a variety of different crops managed within different rotations were available from the IFS trials. Weed densities are highly sensitive to current and historical management of individual fields. Nevertheless, IFS data on mean and maximum weed densities in different crops may be used to indicate the approximate orders of magnitude of densities of various weeds in a variety of crops under conventional management. These densities were used to infer mean levels of control of weeds in different crops, in a baseline model of conventional management. Densities of weeds occurring in sugar beet were not available from the IFS trials but <NEED SOME JUSTIFICATION HERE, ROB> indicated that only C. album and F. convolvulus were likely to occur in this crop with any regularity. Levels of control were set such that average densities of C. album in sugar beet were in the order of one plant per m2, whilst F. convolvulus was much less prevalent, occurring at average densities of only one plant per 20m2.
Data from the IFS trials indicated that weed densities in crops are highly variable (maxima may be more than two orders of magnitude greater than means). In the simulation model, variance arose for two reasons. First, the probability of a field being left uncultivated over winter was applied stochastically, such that increases in weed seed banks due to a winter seedset were random. Secondly, the efficacy of control of weeds by herbicides was also varied log-normally about a mean control parameter. To incorporate levels of variance for each weed comparable to field data would have led to a very low degree of repeatability in model output. In order to ensure a realistic range of densities for each weed, therefore, three separate scenarios were developed, with different mean levels of control for each weed. The three scenarios were: (i) a standard scenario, with weed densities of a similar order to the means observed in the IFS trials; (ii) a ‘clean’ scenario, with weed densities between 10 and 25% of the means observed in the IFS trials; and (iii) a ‘weedy’ scenario, with weed densities of a similar order of magnitude to the maximum densities observed in the IFS trials. This permitted the effects of changes in management to be predicted independently for clean, ‘normal’ and weedy farms, or for areas of each of the three scenarios to be combined within a single landscape, to reflect a realistic diversity of weed densities.
Little is known about dispersal dynamics in weeds, although most arable weeds are known to be poor dispersers <REF, ROB?>. However, it is unlikely that weeds extirpated from a field will remain extirpated without continued herbicide management, unless they are extirpated over a substantial region. To prevent complete extinction of weeds in the simulation, any field in which no seedbank remained for a given species was reseeded with a very low number of seeds (5m-2) which were reintegrated into the seedbank.
For all simulations, weed dynamics were permitted to stabilise over a 1,000 year period before recording data from the model or introducing changes in management. Simulations were run
Table 12.1. Parameters used to model arable weed dynamics
Species / Germination coefficient / Germination coefficient / Annual mortality of seed- / Density dependent seed production parametersbin crops / in stubbles / bank due to decay / A / B
Chenopodium album
/ 0.100a / 0.031a / 0.200 / 230,000 / 0.1Poa annua / 0.187 / 0.061 / 0.193 / 100 / 0.00233
Stellaria media / 0.216 / 0.043 / 0.254 / 1,000 / 0.0125
Fallopia convolvulus / 0.140 / 0.045 / 0.195 / 2,898 / 0.05
Papaver rhoeas / 0.131 / 0.026 / 0.261 / 8,000 / 0.008
Alopecurus myosuroides / 0.200 / 0.050 / 0.700 / 300 / 0.004
a Germination in C. album was restricted to spring, early and late summer.
b Density dependent seed production was modelled according to the formula: AN / (1 + BN), where N is the density of mature plants and A and B are the coefficients summarised above. For C. album, competition with sugar beet was also incorporated where appropriate, using the function: AN / (1 + B(N + 11.11)), where 11.11m-2 is the typical density of sugar beet plants (Freckleton & Watkinson, 1998).
Table 12.2. Seed production parameters for arable weeds in winter stubbles
Species / Density dependent seed production parametersaA / B
Chenopodium album
/ 1400 / 0.005Poa annua / 100 / 0.00233
Stellaria media / 100 / 0.000625
Fallopia convolvulus / 2,898 / 0.05
Papaver rhoeas / No seeds produced in winter stubbles
Alopecurus myosuroides / No seeds produced in winter stubbles
a Density dependent seed production was modelled according to the formula: AN / (1 + BN), where N is the density of mature plants and A and B are the coefficients summarised above. The lower density dependence for some weeds in winter than in summer, reflects the different resource uptake of plants in the different seasons.
using a landscape comprising 300 fields. Field size was normally distributed, with a mean of 15ha and a standard deviation of 8ha. The range of field sizes was 5 – 45ha. The average total size of the landscape was 45km2. For ease of comparison with the IFS trial data, weed plants were censused in the spring.
12.2.2bird foraging and population dynamics
The diets of many farmland passerines appear to be quite flexible and dietary divisions between the majority of granivorous farmland birds are indistinct (see Objectives 3 and 5). Furthermore, little is known about interspecific competition between farmland birds. Consequently, a model of an entire community of farmland birds is not yet a possibility (see Objective 2). It is, however, possible to base a model on a generic farmland passerine, varying the preferences of this species towards different types of seed. Thus, it is possible to determine the densities of birds with given diets that the environment can support, and also whether different dietary groups will react to change in their environment in different ways.
The model was flexible to the exact characteristics of the species but birds in the model were generally based loosely on one of two farmland passerines. These were the skylark Alauda arvensis, with a body mass of approximately 38g and a preference (among the weeds modelled) for the seeds of fat hen, and the cirl bunting Emberiza cirlus, with a body mass of 23.5g and an apparent preference for seeds of annual meadow grass, chickweed and bindweed (see Objectives 3 and 5). The model was relatively insensitive to the precise reproductive characteristics of the species modelled (see section 12.3.2.1, below) and, for simplicity, the two species were given equal reproductive parameters. Specifically, all birds surviving winter could form pairs, up to a maximum of 11 pairs km-2. Each pair produced a single clutch of 3.5 eggs. Overall survival of eggs to produce birds entering their first winter was 0.5. Thus, for each pair breeding, 1.75 additional birds entered the following winter. The total number of additional birds was rounded to an integer. In the results section, we refer to the two types of birds as skylarks and cirl buntings. Note, however, that these remain fairly generic birds, only loosely based on those species.
Fig. 12.4. Seed-setting phenology of weed species that produce seeds in winter stubbles, (a) fat hen (b) annual meadow grass, (c) chickweed, (d) bindweed. Source: ???.
Bird foraging was modelled explicitly only during the non-breeding season (Fig. 12.5). Daily energy requirements were calculated using the relationship between body mass (M) and field metabolic rate (FMR) given by Nagy (1987):
log (FMR) = 1.037 + 0.640 log (M)eqn 12.1
Mean energetic contents of seeds have been assessed by Kendeigh & West (1965) and reviewed in detail by the Game Conservancy Trust (J. Crocker, unpublished data). Typical values are in the region of 18-22kJg-1 and for this model, an average value of 20.5kJg-1 was used for all seeds. Average seed masses and energetic contents are summarised in Table 12.3. Assimilation efficiencies may also vary between studies but values of 70-80% are typical (Kendeigh et al., 1977; Kendeigh & West, 1965). For the purposes of this model, a value of 75% was used.
Within a dietary grouping, birds in the model were given no preference for the seeds of any particular species. Rather, seeds were removed in proportion to their availability. All seeds on the surface and to a depth of 1mm beneath the surface were deemed accessible but birds could not forage below a critical energetic density. For this model, the critical density was initially set at 5kJm-2 (see Objective 2 and section 12.3.2.5, below).
12.2.3management options
Basic simulations were run using non-GMHT crops, a landscape divided equally between the three rotations (Fig. 12.1), and a probability of 0.05 that a field would be left uncultivated over winter. However, to generate predictions of how weed and bird populations will respond to changes in agricultural management, three simple aspects of management could be varied within the model. These were the proportions of different rotations, the probability of fields being left uncultivated over winter, and the use of GMHT rape and sugar beet. Control of weeds within GMHT crops is known to be very high and often complete (Watkinson et al., 2000). To model the use of either GMHT rape or sugar beet, control of all weeds in these crops was set to 100%.
12.3Results
12.3.1weed dynamics and responses to agricultural management
12.3.1.1Control parameters and weed densities
Average effectiveness of herbicides for controlling different weeds in different crops is poorly known although, for most weeds, it is typically in the order of 95-100% (ADAS, unpublished data). Most weed populations are extremely sensitive to the precise level of control; a change of 0.1% can often lead to an order of magnitude change in resultant weed densities. For the purposes of this simulation, control parameters (the proportion of emergent weeds killed by typical herbicide regimes) were set to produce weed densities of a similar magnitude to those observed during the IFS trials. Variance was log-normal and was set to maximise variation in weed densities within realistic ranges and within the limits of repeatability. The control parameters used and resultant weed densities are summarised in Table 12.4. Further details of the range and variance of weed densities are given at the end of this section (Table 12.A1).
Weed plant densities averaged over all of the rotations are show for the 50 years following the stabilisation period (see Fig. 12.6). For a standard, clean, or weedy landscape, weed numbers averaged over the whole landscape were fairly constant from year to year (Fig. 12.6a-c). Fluctuations are most marked in the clean landscape, as even very small increases in weeds due, for example, to a winter stubble or re-seeding event in any field, can lead to relatively large changes in the mean densities. For a landscape with a mixture of standard, clean and weedy areas, average weed numbers were sensitive to the rotational stages in the few weedy fields and, thus, showed much greater fluctuations (Fig. 12.6d).
12.3.1.2Response to varying rotations
The response of weed populations to changing the crop rotation in certain proportions of fields is illustrated in Figs. 12.7 to 12.10. These figures illustrate that changing the proportion of the different rotations has a surprisingly limited effect on weed dynamics. In the standard scenario (Fig. 12.7), few patterns are evident, even when increasing the proportion of any of the rotations up to 75% of the total area. Clearly, however, increasing the proportions of either rotation 1 or 2 does lead to a substantial decline in annual meadow grass (in both cases), chickweed (in the latter case) and bindweed (in the former case). Less substantial declines may also be indicated for chickweed (in the former case), bindweed (in the latter case) and poppy (in both cases). Increasing the proportion of rotation 3 has very little effect on the majority of the weeds