From Field to Forest

This project simulates the process of natural evolution of a new stable ecosystem on an abandoned field.
 * Project name: From Field to Forest
 * Class: 4IT495 Simulation of Systems (WS 2013/2014)
 * Author: Alice Peková
 * Model type: Agent-based simulation
 * Software used: NetLogo 5.0.4

Problem definition
When a field is abandoned, it takes aproximately 150 years before it turns into a natural forest. This process called secondary succession was firstly described by F. A. Clements in 1916. In this paper a Tolearance Model of Succession is used as presented in the article Mechanisms of Succession in Natural Communities and Their Role in Community Stability and Organization written by Joseph H. Connell and Ralph O. Slatyer, which is typical for abandoned field behavior. This model of succession suggests that the change in plant species dominance over time is caused by competition for resources. Later species are able to tolerate lower resource levels due to competition and can grow to maturity in the presence of early species, eventually out competing them. The goal of the project is to simulate the process of secondary succession on an abandoned field. The simulation should answer the question how the initial fertility of the field influences the whole process. The sources used are soil fertility and sunshine available. The model is set in central-european climate, which implies that the ecosystem reached with the succession should be a mainly deciduous forest with oaks prevailing.

Method
The succession is a complex process with number of mutual dependancies. Two essential viewpoints can be taken. The focus can be put either on the feedbacks between various plants, soil and the envirnment, or on the individual behavior of each plant in the changing environment. The results of feedback loop analysis of the process of secondary succession was presented for example in the article Feedback dynamics analysis of secondary successional transients in ecosystems written by Luis T. Gutierrey and Willard R. Fey. I have chosen the latter approach, as there are many possibilities of individual plant's reactions. For this approach an agent-based simulation is appropriate where each plant is percieved as an agent with specific requirements and behavior. The NetLogo software was used for its simplicity and easy data analysis.

Model
The species listed below are gathered from a diploma thesis Sekundární sukcese na opuštěných polích v pahorkatině jižní Moravy (Secondary Succession on Abandoned Fields in a Hilly Landscape of South Moravia) by M. Sojneková.

Agents and environment
The model uses 4 breeds af agents (turtles) - grasses, herbs, bushes, and trees. The characteristics of the environment are saved as variables of patches. The initial average fertility of the field is stored as a global variable and can be adjusted by the user before running the simmulation. The patches also store number of seeds in the soil for each type of turtles. The agent characteristics are described in variables as follows: The agent characteristics are typical for each type of turtles, however it is randomized in order to refer more accurately to the reality. The characteristics are set as a value of normal distribution with mean value typical for the type of turtle and standard deviation set as a fraction of the typical value (in case of rfertility, rsunshine and max-age) or as a constant (in case of death-ratio and spread-ratio). The age is naturally set as 0. The characteristics of environment are stored for each patch seperately. One patch represents 1m2 of the soil. The default size of the field is 20m x 20m. The characteristics are as follows: The seeds that don't manage to sprout in the spring stay in the soil and may sprout next year or any year after. The initial number of seeds on a patch is generated randomly on given percentage of patches. This simulates the fact that when a field is abandoned, there is significant number of seeds on the soil.
 * rfertility: sets the minimum soil fertility requirements of the plant. If the actual fertility of the patch is lower, the plant dies.
 * rsunshine: sets the minimum requirement for exposure to the sun. If the actual exposition on the patch is lower, the plant dies. The exposition is influenced by bushes on the same patch and trees on the same patch and in the neighborhood.
 * death-ratio: sets the percentage of plants dying every winter.
 * spread-ratio: sets the number of seeds which the plant produces and which can sprout and grow. The plant can spread on the same patch or on the neighboring patches.
 * age: stores the current age (in years) of the plant. The age change every step of the simulation.
 * max-age: sets the maximum age the plant can reach. When the plant gets older, it dies.
 * local-fertility: percentage of the humus in the soil. The local fertility changes over time according to plants living on the soil and rotting dead plants. The initial value is set as randomized user-selected fertility.
 * sunshine: percentage of sunshine that reaches the ground. The initial values is 100 (no shade is produced), but changes over time according to number of trees on the patch and on the neighboring patches and number of bushes on the patch.
 * seeds-grass: number of grass seeds on the patch. Every spring one half of the seeds manage to sprout.
 * seeds-herbs: number of herbs seeds on the patch. Every spring one half of the seeds manage to sprout.
 * seeds-bushes: number of bushes seeds on the patch. Every spring one quarter of the seeds manage to sprout.
 * seeds-trees: number of trees seeds on the patch. Every spring one fifth of the seeds manage to sprout.

Grasses
The grasses represent annual plants, such as Melica uniflora (wood melica), Poa nemoralis (wood bluegrass) and Brachypodium pinnatum (tor-grass). The initial values of variables:
 * rfertility: 10
 * rsunshine: 50
 * death-ratio: 43.5
 * spread-ratio: 36
 * max-age: 1

Herbs
The herbs represent mostly perrenial herbs, such as Bellis perennis (daisy), Taraxacum officinale (dommon dandelion) and Convallaria majalis (lilly of the valley). The initial values of variables:
 * rfertility: 10
 * rsunshine: 50
 * death-ratio: 44
 * spread-ratio: 27
 * max-age: 8

Bushes
The bushes represent mainly pioneering woods, such as Crataegus monogyna (single-seeded hawthorn), Betula pendula (silver birch) and Picea abies (spruce). The initial values of variables: The bushes produce shade when older than 5 years. This shade consumes 10% of sunshine reaching the ground on the same patch.
 * rfertility: 10
 * rsunshine: 58
 * death-ratio: 20
 * spread-ratio: 27
 * max-age: 50

Trees
The trees represent mainly deciduos trees typical for central europe oak forests, such as Quercus petraea (sessil oak) or Acer campestre (field maple). The initial values of variables: The bushes produce shade when older than 10 years. This shade consumes 15% of sunshine reaching the ground on the same patch. Trees older than 20 years produce shade which consumes 25 % of sunshine reaching the ground on the same patch and 15 % of the sunshine reaching the ground on the neighboring patches.
 * rfertility: 20
 * rsunshine: 40
 * death-ratio: 10
 * spread-ratio: 27
 * max-age: 150

Model dynamics
Before the model run is started the initial values are set: global fertility set by user is randomized for each patch. Seeds are generated for some patches (80% of patches hold grass and herbs seeds and 40% of pathces hold bush and tree seeds). When the model run starts, the first spring come and then the seasons follow for user-set number of years. When growing, the plants decrease the local fertility (more when older). When dying, the plants increase the loceal fertility (also more when older). Anytime during the year a plant dies when the fertility and sunshine requiremnts are not met.
 * Spring: The seeds sprout (i. e. the turtles are initiated). However a lot of sprouts may die. When a sprout manage tu survive, it decreases its likelihood to die a little.
 * Summer: All the plants grow (some of them may die, survivales decrease their death-ratio), grasses and herbs produce new seeds according to their spread-ratio.
 * Autumn: All the plants grow (as in spring and summer), bushes older than 3 years and trees older than 5 years produce seeds.
 * Winter: All the grasses die, 2/3 of herbs die, other plants die according to their death-ratio. The survivals decrease their death-ratio slightly.

User interface
Before running the simulation the user can adjust the initial fertility of the soil and set the number of yoers for which the simulation will be run. When hitting the SETUP button, the values are initiated. When hitting the GO button the simulation starts. The data is recorded and displayed every autumn. However the monitors in the left side of the window (counting each type of turtles) may update continuously. The graphs in the right side of the window show the evolution of numbers of grasses, herbs, bushes and trees, average fertility and sunshine exposure and evolution of nubers of bushes older than 5 years and trees older than 15 years. The latest indicates the maturity of the ecosystem.

Results
During the run of the model it became obvious that with parameters set as described above the balanced ecosystem cannot be estabilished. This can be caused by following isuues:
 * Wrong setting of turtle parameters. The parameters set in the model are artificial (I was not able to find real values of selected parameters for specific plants). Although the parameters provide basic differences in the behavior of the types of plants, the dynamic does not correspond to reality.
 * Wrong selection of parameters. The dynamics of the process may be caused by more or different factors (such as amount of nitrogene in the soil, soil humidity, soil acidity and many others).
 * Too simplified behavior of plants. The behavior of the plants in the simulation is very sipmle. The values of parameters do not change rapidly, the growth is linear and the plants do not inherit their predecessor's values of parameters. The plants also do not react much on other plants around directly(only too big trees die when another big tree is on the same patch) but only through the shade (which is also not influenced by grasses and herbs). The balance of the ecosystem may be reached due to a far more complex behavior of the plants.
 * Insufficient interaction between turtles and (or patches). The process of succession is very complex and there may be hidden interaction which I did not manage to reveal. For example some trees produce chemicals preventing specific kinds of plants to live nearby.
 * Hidden effect of animals in the ecosystem. The influence of animals on the process is usually ommited. Yet the role of animals may be also crucial for developing a balanced ecosystem.

Simulation runs results
Although the model does not reflect reality in the way it was supposed to, it reveal some conclusions. Therefore I provide a brief overview of the data produced. The analysis was performed with following parameters: The data was recorded at the end of each run. The Excel file with the results and brief summary can be downloaded here. From the data gathered we may conclude following: Althoug the model does not reflect real behavior of the succession process, it behaves in an interesting way. The plants organize themselves in clusters with oldest plants tending to be in the center of it and the young plants mainly on the periphery. Also bushes and herbs tend to be organized around trees more likely then standing alone in the field. Grasses grow on the largest area of the field and are able to migrate form one part of the field to another due to their high ability to spread.
 * Varying variables:
 * ["number-of-years" 200]
 * ["Fertility" [20 5 40]]
 * Repetitions: 10
 * Recorded data:
 * count grasses - number of grass individuals (always 0, the grass dies in winter)
 * count herbs - number of herbs
 * count bushes - number of bushes
 * count trees - number of trees
 * tot-grasses - total number of grasses created
 * tot-herbs - total number of herbs created
 * tot-bushes - total number of bushes created
 * tot-trees - total number of trees created
 * year-of-death-grasses - the year when the grasses extinct (have less then 10 individuals)
 * year-of-death-herbs - the year when the herbs extinct (have less then 10 individuals)
 * year-of-death-bushes - the year when the bushes extinct (have less then 10 individuals)
 * year-of-death-trees - the year when the trees extinct (have less then 10 individuals)
 * The ecosystem cannot be estabilished with initial fertility lower then 25%. With higher fertility grasses and trees are able to survive.
 * Meadow can be estabilished with fertility between 20 % and 25 %.
 * Grasses and little trees dying younger than 3 years are crucial for maintaining balanced fertility of the soil.
 * The bushes extinct (probably due to shade provided by the trees).
 * The herbs may survive only if the trees survive as well.

Conclusion
The model does not correspond with the reality, most probably due to wrong setting of initial parameters and too simplified behavior of the plants and their interaction. However with initial fertility over 25 % the "ecosystem" can survive, but without herbs and bushes. The effect of fertility on the process of succession seems to be important, the fertility during the simulation run does not change rapidly and does have a direct impact on survival of the plants.

Code and results files
File:Succession_model.zip File:Succession_results.xlsx