The Model: Questions and Design

The proximate objective of WELASSIMO is an answer to the question, whether a need for a frequent settlement relocation (as documented by the archaeological record, e.g. Bleicher 2009; Dieckmann et al.

Table 9.1 Primary and derived data used to reconstruct the idealized environment of WELASSIMO.

resolution of 25 X 25 m. These were used as the basis to generate a dynamic model environment with NETLOGO. Using the GIS-Extension of this agent-based programming software, a surface was created where every cell is characterized by a set of certain environmental parameters, some of which are static, such as elevation or soil type, others are dynamic and are updated in annual time steps, such as the state of the vegetation cover, the forest development phases or the quantity of relevant resources.

The modern soil map had to be adjusted to assumed Neolithic conditions, as certain modern soiltypes have only developed with intensified human impact since the Neolithic, and others have changed due to natural or anthropogenic processes (Gerlach 2006; Vogt 2014). Since the exact nature and amount of these changes is nearly impossible to assess, the reconstructions are necessarily tentative to a certain degree. Similarly, the allocation of a certain vegetation type to a grid cell is an idealized simplification. I defined standardized forest compositions, based on the naturally dominant tree species in the 4th Millennium as reconstructed by pollen analysis (Rosch 1990; Lechterbeck 2001) and on the modern natural vegetation in the region (Lang 1990), and assigned these communities to certain soil types. The age of the trees in primary lowland deciduous forests is not homogeneously dis­tributed, but instead, these woods are characterized by a mosaic of stands of dif­ferent development phases, as described by various authors (e.g. Härdtle et al. 2004, p. 85; Bernadzki et al.

The economic processes of interest are simulated as a second step inside of this environment. The basic unit (or “agent”) of the model are the households = houses, which are idealized and represent a standardized group of assumed 6 persons comprising adults, children and elderly. The number of households is selectable in a range from 1 to 20, representing a population of 6-120 people. The main driver of the model is the annual demand in calories and non-finite resources of the households. The annual calorie demand of one household is defined as 365 * 6 * 2000 = 4,380,000 kcal, which is the average demand of all potential inhabitants of the houses. In accordance with the objectives of this research, only non-finite resources that exhibit dynamic properties are considered: arable and fertile soil, livestock pasture, hunting/gathering/fishing areas, and suitable timber (see Table 9.2). Finite resources such as flint for tools or clay for pottery are not considered. The households have a set of subsistence strategies that they apply in order to meet their demand (see Fig. 9.2): crop husbandry, livestock husbandry, hunting/fishing and gathering, and timber extraction. While the basic properties, the timing and the functioning of these strategies is predefined and encoded inside of

Table 9.2 Quantitative data on resources per patch (1/16 ha = 25 X 25 m) as assumed in WELASSIMO.

Fig. 9.2 Schema showing the interplay of the various elements of “Landscape” developing in the vicinity of a hypothetical wetland settlement in the 4th Millennium BC. Landscape is understood as the result of anthropogenic and natural processes shaping the human environment. Rectangles symbolize physical elements, diamonds denote processes, and ovals show system drivers. Arrows either denote influence on or affiliation to a certain element

the model, the exact configuration is defined by the observer, so that the outcome of various assumptions and hypotheses can be studied.

Table 9.3 Choice options for the observer

During a simulation run, the households cover their demand according to the chosen scenario. They always choose the patches with the desired properties which have the minimal traveling cost—a parameter that differs from the mere Euclidean distance in that realistic walking time is calculated by additionally taking into account the decreasing walking speed with increasing slope. For the crop fields, the soil type of the patches must be “Luvisol” and the slope must be below 25°, an arbitrary value. The other soiltypes that supposedly covered relevant areas in the Lake Constance area in the Neolithic are “Gleysol” and “Histosol”, which are less well suited by far and therefore are excluded a priori. The size of the crop fields is calculated at the beginning of the simulation and remains fix for the subsequent years. The underlying assumption is that Neolithic people had acquired or inherited a knowledge about the most appropriate size of their fields, with respect to the available work force and the minimal proportion of crop calories aimed at.

Average yields for the use in WELASSIMO are taken from the Rothamsted Dataset for IGC, IFH and NIC and from Ehrmann et al. (2009) for SC, while loss in

Fig. 9.3 Left Synthetic crop yield series (100 % dry matter) of 100 years simulated with the MONICA-model (Nendel et al. 2011). The scenario is IGC: annual application of manure (10 T/year/ha) and intensive weeding. The declining long-term trend is reflecting loss in soil fertility, while weather variability accounts for the annual variation. Right 100 years yield-series (100 % dry matter) as documented by Rothamsted Research. Yields increase slightly in the beginning due to a very high annual application of 35 T of manure per hectare. The decline in yields after 1900 AD is a consequence of reduced weeding, the rise of the curve is a consequence of renewed engagement and improved husbandry methods (Poulton 2006)

soil fertility is taken from the simulation. The annual demand in patches for live­stock pasture and pollarding is dependent on the configuration set by the observer, and additionally on the vegetation cover of the patches. In the Neolithic wetland settlements, the minimum number of livestock individuals (MNI) per household lies between 0.1 and 4 (Ebersbach 2013). The livestock density in WELASSIMO uses these values as minimum and maximum values. Only cattle are considered, even if pigs, goat and sheep are also documented (e.g. Schibler 2006). If the importance of animal products and the proportion of domestics therein is set to high, or if IGC or IFH are selected and the need for animal dung is large, a higher MNI is used than if SC or NIC is chosen, and no livestock products are consumed at all. As shown by various authors, the area necessary to support one cow is highly dependent on the quality of the pasture or the area used for pollarding. While on grassland meadow (for which there is no proof in the Neolithic), 1.2 ha might be sufficient (Ebersbach 2002, p. 156), in a poor pine forest 20-30 ha are needed (Adams 1975, p. 148). In Welassimo, a minimum of 2 ha per cow on young fallowing areas and a maximum of 12 ha per cow in beech forest of the optimal and terminal phase are assumed. Livestock will choose patches with soil type not Histosol without crop fields with the least walking cost.

Also the area required for gathering is dependent on the model configuration as chosen by the observer and the vegetation cover of the respective patches. I assume that the maximal “gathering value” an area can have is reached when it is covered with secondary forest growth dominated by hazel, a vegetation type typical for the wetland sites as documented by palynology (e.g. Lechterbeck 2001; Rosch et al. 2014). Hazelnut kernels are extremely rich in calories (6000 kcal/kg). Using very conservative numbers given by Holst (2010), on one hectare of this vegetation type about 84 bushes may have grown in the Neolithic. 2000 nuts per bush may be harvested, adding up to an average of 900.000 kcal per year and hectare, divided by 16 for the patches (which have the size of 1/16 ha). These patches get the relative value of 1/16. The minimal gathering value a patch can have is arbitrarily set to 0.005/16. Thus, an area covered with mature hazelnut bushes yields 200 times more calories than the poorest areas (e.g. peat bog or optimal/terminal-phase beech forest). While the maximal value for nuts is backed by other studies, the lower number is speculative, and is open to discussion. The requirement in gathered calories of the households is the relative value of 1 * the proportion set by the observer, equaling a potential importance of minimum 0 and maximum 80 % of gathered calories. So, to meet this number, the households visit successively the patches beginning with the ones exhibiting the lowest travel costs, and add the respective gather value to their annual calories provision until the required amount is met.

The required area for hunting and fishing is not (yet) quantified, due to diffi­culties with finding reliable data. It is hypothetically assumed to have been sus­tainable; the simplifying assumption behind is that neither the populations of wild animals nor fish are severely limited in quantity as long as relatively small human populations exploit them. Furthermore, the degree of uncertainty with the decline in game or fish density due to human exploitation is dependent on too many factors to be simulated with confidence in this project. The only representation of these activities inside the model is the annual filling up of stocks according to the chosen parameters, with slight variations allowed for. The fluctuation of the success of calorie provision by livestock products, hunting, gathering and fishing in WELASSIMO is lower than the fluctuation of crop yields. If the observer chooses an importance of 5 % for one of these foodstuffs, than the actual value will be around 3-7 %. The reason for this is that field size is fix, and bad yields cannot be compensated by simply harvesting more patches. To the contrary, in hunting, fishing and gathering, endless resources are assumed, if the radius of action is extended. This is only possible because worktime limitations are not accounted for in WELASSIMO. The main reason is that they are used for validation; this is discussed below. As can be seen on the center panels of Figs. 9.4, 9.5, 9.6 and 9.7, in many cases the sum of all foodstuffs does not equal 100 % of the aim, but instead, over-and underachieving happens quite frequently. While in a real situa­tion, people would try to mitigate this by increased investment in other calorie sources, this is not accounted for in WELASSIMO in order to highlight this effect.

The demand in timber is determined by the house age. A few logs are needed every year for reparations, while the probability of a complete reconstruction increases every year. In the year of the breakdown of the house, 120 logs of 5­20 cm are needed (combined after Luley 1992; Petrequin 1991). The area necessary to meet these numbers is dependent on the numbers of suitable timber per patch— again, patches with lowest travel cost are considered first, suitable timber extracted as needed and provided, and if needed, the next patch will be used as well. In the model, a beech forest in the optimal phase may have no suitable logs at all, while in the regeneration and in the juvenile phase of a mixed alder-ash forest, 400-600 such

Fig. 9.4 Scenario 1 simulating land use, resource use and economic parameters of a hypothetical Neolithic wetland settlement relying on SC. Grey and blue boxes in the upper left corner are observer choice options, white boxes are monitors displaying current model conditions and results. (MNI = minimum number of individuals = livestock). Blue areas are lakes, different shades of green are forests of different species composition on specific soil, and brownish colors are fallowing patches. Only dark green colors (representing Luvisols) are soils suitable for agriculture. A white house symbol represents the settlement. Black circles mark patches affected by livestock browsing. One raster cell equals 25 * 25 m = 1/16 ha, the largest lake measures 750 m from east to west. Note that the maximum on the y-axis of the lowest panel is higher than in the other three scenarios

trees may be found (calculated after Korpel 1995, S. 127-137). On young secondary forest with ages of 20+ years, an assumed maximum of 700-800 suit­able trees per hectare is simulated.

9.5