ODD-Protocolof the Agent-Based Model WELASSIMO Crops

1

ODD-Protocolof the agent-based model “WELASSIMO_crops”

The model description follows the ODD (Overview, Design concepts, Details) protocol for describing individual- and agent-based models (Grimm et al. 2006, 2010).

1.Problem

In lakes and peat bogsof the northwestern pre-alpine forelands,there are proofs of settlements from the late Neolithic until the Bronze Age. Favored by the anoxic conditions of the waterlogged environment,well-preserved organic remains have been studied for decades, and much is known about the subsistence strategies and nutrition habits of the people. However, especially with regard to crop husbandry many questions still remain unanswered, and antithetical hypotheses are discussed.

1.1 Purpose

The aim of WELASSIMO_cropsis to understand the effects and implications of three hypotheses of Late Neolithic crop husbandry methods on the landscape development and the socio-ecological system of the wetland sites.

The main goals of the model are answers to the following questions:

What was the spatial extent and the workload for crop cultivation under the assumption of different crop husbandry methods?

What was the potential cereal share of the overall annual calorie supply of the prehistoric wetland sites assuming certain field sizes and crop husbandry methods?

Which were the socio-ecological implications of the crop husbandry methods?

Can statements regarding the plausibility of the crop husbandry methods be deduced from the simulation results?

1.2 Approach

We use a simulation model of the houses that perform cereal cultivation to explore the land demand, the workload, and the importance of cereals for the annual calorie supply.

2.Background

As lake shores and peat bogs were among the favored settlement sites of Late Neolithic people living in the pre-alpine forelands, organic remains have been preserved in many cases due to the anoxic conditions of the archaeological layers. In a multitude of publications, comprehensive syntheses of the (bio-and geo-) archaeological data are presented that provide detailed insights into the wetland ways of Neolithic life (e.g. Billamboz et al. 2010; Jacomet et al. 2004; Maier et al.2001;Matuschik and Strahm 2010; Menotti 2004; Schibler et al. 1997; Schlichtherle et al. 2010). It is known from archaeological data that social stratification was very weak in these settlements, and that most of the houses were self-subsistent units (Dieckmann et al. 2006).It is well-documented that the people performed crop cultivation and animal husbandry as well as hunting, fishing and gathering to cover their caloric demand. In the spectrum of cultivated plants, various types of cereals can be identified alongside with pulses, flax, opium poppy, and other species (Jacomet 2006, 2007,2009, Jacomet et al. 1989; Maier 2004; Maier et al. 2001

The main goals as described above can be achieved using a simplified representation of the much more complex socio-ecological system. For our purpose, the system can be described as consisting of houses, their environment, and their crop fields. Depending on locational choices of the settlers, the size of the fields and the size of the houses, the extent of the cultivated area can be assessed. The annual variation of the cereal supply of the houses and the according relevance of cereals for the total nutrition can be assessed by simulating the crop harvests with annual variation due to environmental variability and the characteristics of husbandry methods. For these, published hypotheses have been set up using (bio-) archaeological data integrating ongoing research (Bogaard 2004; Brombacher and Jacomet 1997; Ehrmann et al. 2014; 2009; Rösch et al. 2014; Jacomet et al. in press). The following hypotheses are used:

1)Intensive Garden Cultivation (igc). This hypothesis assumes a permanent cultivation onfields that are prepared using hoes.

2)Shifting Cultivation using long fallows of 15 years (scl) or short fallows of 8 years (scs). This hypothesis does not include the use of ploughs as well and is also termed Wald-feldbau, Jhum, and other names. The crop fields are prepared using a burning method by which vegetation is cleared and nutrients are supplied to the future crops.

3)Ard cultivation (ac). The hypothesis assumes the use of simple ploughs that are drawn by draft animals.

The influence of environmental and anthropogenic factors on crop yields is taken from ethnographic literature where available (Kerig 2008; Ehrmann et al. 2009; Jepsen et al. 2006), or comes from crop yield simulations (Nendel et al. 2011),or is assumed if no data were available.

3.Entities, state variables, and scales

The temporal resolution of the model is one year, the spatial resolution is 25 m. A number of standardized model houses as chosen during model setup perform crop husbandry according to defined rules as described below. They are located fixed on one spot (=one settlement). Every house has the dynamic state variables cropyield [kg] (displaying the annual crop yield), fielddistance [m] (for the maximum distances of the houses´ fields), cattle (giving the required minimal number of cattle to obtain manure),and „WL_1“ - „WL_12“ [h], calculating the average daily workload for crop husbandry for every month of the year from the model parameters using ethnographic data.

The houses are located in a simplified model landscape that consists of 200x200 grid cells of 25x25 m each (=1/16 ha); one house cultivates as many cells as defined initially during setup. Cells can either be in natural state, be used as crop field, or left fallowing. They are described by a number of state variables, some of which are static, others are dynamic. The static ones are soiltype, elevation [m], travelcost from the settlement [relative units], slope [degree] andpotential ecosystem (indicating the assumednatural stateon the patch in the Late Neolithic:mixed beech forest, alder-ash-forest, peat bog, or open water).The dynamic ones (which are updated with every time step)are natural soilfertility [0-1], (indicating the forest development phase of the vegetation cover on the cell), field use [y/n] (indicating whether or not the cell is used as field), and fallow [years] (if the cell is currently fallowing).

If cultivated as crop fields, cells have variables denoting the owner (a house), the age of the crop field (influencing the cell´s fertility)and theannual yield on the cell, which is calculated annually as described below. Weather stochasticity and husbandry method are valid on global scale, affecting all crop fields in the same way.

The husbandry method is characterized by its labor demand, the average yield, and the field permanence. If the cells are in natural state, they are covered with their potential ecosystem. If this is beech or alder-ash-forest, forest patch dynamics apply (see below) and the cell has a value for the dynamic state variable standage [years]. If the cell is actually fallowing, it has an age [years] that indicates how many years ago the cultivation on the patch was stopped.The relationship between model concepts is displayed in Fig. 1

Fig. 1: Relationship of the concepts used in WELASSIMO_crops

Fig. 2: Flow Diagram of WELASSIMO_crops

4.Process overview and scheduling

1. The influence of the annual weather on the cropyields is defined
2. The houses check whether a new field has to be prepared
3. The houses open and prepare new fields, if indicated.
4. Crops grow on crop fields, the yield is calculated
5. Houses harvest their crop fields and update their dynamic state variables
6. The cells covered with forest (primary/secondary) age by one year
7. The model output and the plots are updated.

5.Design concepts

Basic principles

The model uses published hypotheses on prehistoric crop husbandry in the Late Neolithic wetland settlements (Rösch et al. 2014; Jacomet et al. in press)and translates these into scenarios. The details of the husbandry methods are taken from ethnographic literature if available, come from crop yield simulations, or are assumptions if no data was available. The results inform about the spatial extent of the crop fields, the workload and the annual cereal share of the total calorie requirements of the houses.

Emergence.Key results of the model are: I) the extent of the area needed for crop husbandry under given conditions. This feature is predetermined, as the field size and the number of houses are defined during setup. It is variable however in different landscapes. II) The workload of the houses for chosen scenarios. This is also predetermined by model settings. Ethnographic data used to calculate the workload is given in Table 4. III) The pattern of the cereal contribution to the total calorie requirements under given conditions. This value is dependent of the annual crop yield, which is calculated from the husbandry method, the patch fertility, the patch elevation, and the weather stochasticity. IV) The map showing the distribution of vegetation patterns under shifting cultivation. This emerges from the initial random distribution of forest stand characteristics.

Adaptation.The agents have no adaptive traits, their behavior is determined by the code.

Objectives. The houses have no individual aim other than cultivating the field size set by the observer. They don’t have any options to choose from.

Sensing. The houses are capable of identifying the most suitable field patches for cultivation.These are defined by travel costs, soiltype, and vegetation cover.

Interaction. Houses interact with cells by clearing their vegetation and using them as crop fields or letting them fall fallow. A house may use one or more cells. A variable of the cells is that they may host one cropfield. The cells´ variable fertility is influenced by the duration of field use on the cell, and influences the crop yield of the patch.

Stochasticity. The crop yield of the fields is calculated integrating the stochastic variability of the annual weather. The spatial pattern of forest development phases is stochastic, too.

Collectives. Houses are not individually dispersed in the landscape but are grouped on one defined cell. However no interaction between houses takes place.

Observation.

The maximal distances of the crop fields, the required walking time and the overall area spanned by the most distant fields are documented.

The cereal share of the houses, the required number of cattle for dung production, and the workload of the houses are documented as well.

Landscape development showing the localization of fields, the opening of the forest areasand the development of fallow areas are given as a map, which is updated annually.

6.Initialization

6.1 Landscape initialization

The initial state of the model at the time t=0 is a hypothetical model environment without previous human influence consisting of 200x200 cells of 25x25 metres. The cells are described by various variablesprovided via input files that are generated using Geographical Information Systems (GIS). In the presented model version, the following data were used, but other spatial data from other landscapes can be applied as well.

The variable soil typemay either be Luvisol, Gleysol, Peat bog or Water. The soil data has been simplified and adapted(using data from LGRB 2013) to better resemble conditions before the onset of wide-spread human influence using simple rules.All sub-varieties of a soil type have been treated as being identical (e.g. Anmoorgley, Quellgley, Auengley have are all treated as Gley); Colluvia have been assigned the value of the dominant neighboring value of Gley or Luvisol, as they were only formed through more intense human influence than in the period in question; the same procedure has been applied to areas with the value no data. The result is a much generalized, but still valid soil map that roughly retrodicts conditions of the mid-Holocene without strong human influence.

The variable travelcost is taken from an externally calculated GIS raster dataset (©

The variable natural soil fertility is dependent on the soil type and has the relative values of 1 for Luvisol, 0.75 for Gleysol and 0.4 for Peat Bog. These values are assumptions.

The natural potential ecosystem of the cell is dependent on the soiltype: Mixed Beech Forest on Luvisol, Alder-Ash-Forest on Gleysol, and Alder Carr on Peat Bog. This very generalized allocation is based upon information given by Rösch et al. (2014, p. 124), Kerig and Lechterbeck (2004, p.24)and Lang (1990).

The cells covered with Beech Forest or Alder-Ash-Forest are grouped to stands consisting of 1-16 cells, equaling 0.06 -1 ha, sharing the same forest development phase (rejuvenation-, initial-, optimal-, terminal, - and decay phase). This roughly reflects conditions for natural forest dynamics as described e.g. in Emborg et al. (2000), Remmert (1991) or in Leibundgut (1959, 1982, 1993). The area covered by the forest development phases is determined by the ratio of 8/8/28/36/20 as described in Mayer and Neumann (1981). With an assumed duration of one complete cycle of forest development from rejuvenation to decay phase lasting 500 years, this translates into a duration of the phases of 40, 40, 140, 180 and 100 years in this simulation. For Alder Carr, no such dynamics is assumed.

6.2 House initialization

The houses are located on patch 103/97, which is chosen because it lies in the middle of the landscape and the location resembles true site preferences, as it is located on a small peninsula facing southward. The houses´ initial state variables are identical, i.e. every house has six inhabitants, 2 full and two half workforces with a total workforce of 24 hours daily, and a standardized calorie requirements equaling 6x2000x365 = 438000 kcal.

6.3 Scenario initialization

Modelparameters have to be chosen before the start of the simulation. The number of houses may be 1-100. Field sizes may range from 0.1 to 2 Ha. The crop husbandry system can have values of “ard cultivation”, “intensive garden cultivation”, “shifting cultivation, long fallow” and “shifting cultivation, short fallow”. The systems are defined following established hypotheses, which are described e.g. in Ehrmann et al (2009, 2014),Jacomet et al. (in press)and Bogaard (2004, 2011). The most important properties of the methods are given in Table 1. In “ard cultivation” and “intensive garden cultivation”, the manuring rate must be chosen (no manure application or 10 T animal manure/ha/a).

Table 1: Description of the crop husbandry methods used in WELASSIMO_crops and allocation to the scenarios.

scen. / husbandry method / description
S_ac / ard cultivation (ac) without manure application / extensive method, field preparation with ard, broadcast seeding, permanent field use, manuring optional; mean crop yield 0.88 t ha–1a–1 without manure; weather – induced standard variation 30 %, crop fields on Luvisol without limiting forest areas in between, soil fertility decrease: y = -0.211ln(x)+ 1.0934
S_igc / intensive garden
cultivation (igc) with manure application / intensive method, field preparation by hoeing, single grain seeding, permanent use of the fields, manuring optional; mean crop yield 1.9 t ha–1a–1with manure; weather – induced standard variation 27 %, crop fields on Luvisol, separate fields of the individual houses, soil fertility decrease: y = -0.151ln(x) + 1.5226
S_scs / shifting cultivation,
short fallow (scs) / labour-intensive and land-demanding method, annual shifting to new areas with vegetation cover suitable for the burning procedure, crop cultivation on areas that are clear-cut and burned, single grain seeding, cycle length 8 a; mean crop yield 2.24 t ha–1a–1; weather – induced standard variation of 22 %, crop fields dispersed on Luvisol or Gleysol
S_scl / shifting cultivation,
long fallow (scl) / same as S_scl, but fallow cycle length is 15 a, mean crop yield is 2.6 t ha
–1a–1

7.Input data

No external datasets such as climate data,which represent variable environmental factors,are used. The variability of annual weather is not simulated in detail, instead it is included in the formula for calculating the annual crop yields.

8.Submodels

8.1 Quantify weather influence on crop yield

The influence of the annual weather on this years´ crop yields is stochastically determined. The modeller has done own research using an agro-ecosystem model (MONICA, Nendel et al. 2011)to be able to estimate the mean and standard deviation of the yield distribution in 100 simulated years. Relevant statistics are given in Table 1. These are integrated using the built-in function “random-normal” in Netlogo, whichpermits the giving of a mean and a standard deviation and will return a normal distribution of the values accordingly.

To-report weather-variance ifelse crop_husbandry_system = „shifting cultivation, long fallow“ or crop_husbandry_system = „shifting cultivation, short fallow“[report random-normal 1 0.22] [ifelse crop_husbandry_system = „intensive garden cultivation“[report random-normal 1 0.25][report random-normal 1 0.3]] end

8.2 Choose and prepare crop field

The houses check whether a new field has to be opened, and eventually do so. The corresponding rules differ for the four methods. With “igc”, new fields will be opened, if the number of houses is less than the number of fields (which is the case only in year 1). Then, the cell with the minimal walking cost with Soiltype = Luvisol not in use by any other houses andhaving a certain distance to other fields is chosen. This cell and the number of cells that are needed accordingly to meetthe fieldsize as chosen by the parameter hectare are then defined as fields of the respective house. For ac, the process is identical, with the single difference that no minimal distance between fields is kept. For scl and scs, the criteria are as follows: minimal walking cost, “Soiltype Luvisol or Gleysol”, forest development phase in the initial-, rejuvenation-, or decay phase, period since last field use > minimal fallow period as described for the methods (8 and 15 years,respectively).

igc/pc:

to choose-and-open-igc(pc) if count patches with [pfieldcenter? = true] < (count HEs)[ask Hes[let owner self let x (5 * sqrt Hektar)let my-field min-one-of patches with [pfield? = false and psoiltype = 1 and count patches in-radius 3 with [psoiltype = 1] > 20 and count Hes in-radius x = 0 ] [ptravelcost]move-to my-field ask patch-here [set pfieldcenter? true set pfield? true set powner owner]let belonging_igcfield min-n-of (Hektar * 16) (patches with [psoiltype = 1 and pfield? = false]) [distance myself]ask n-of (Hektar * 16) belonging_igcfield[set powner owner set pfield? true ]]]end