Integrating Multi-Model Simulations to Address Partial Observability in Population Dynamics: A Python-Based Ecological Tool

Abstract

Species richness is a crucial factor in maintaining ecological balance and promoting ecosystem services. However, simulating population dynamics is a complex task that requires a comprehensive understanding of ecological systems. The current tools for wildlife research face three major challenges: insufficient multi-view assessment, a high learning curve, and a lack of seamless secondary development with Python. To address these issues, we developed a novel software tool named WAPET (Wildlife Analysis and Population Ecology Tool) (Python 3.10.12). WAPET integrates Monte Carlo simulation with ecological models, including Logistic Growth, Random Walk, and Cellular Automata, to provide a multi-perspective assessment of ecological systems. Our tool employs a fully parameterized input paradigm, allowing users without coding to easily explore simulations. Additionally, WAPET’s development is entirely Python-based, utilizing PySide6 and Mesa libraries and enabling seamless development in Python environments. Our contributions include the following: (I) integrating multiple ecological models for a comprehensive understanding of ecological processes, (II) developing a no-code mode of human–computer interaction for biodiversity stakeholders and researchers, and (III) implementing a Python-based framework for easy extension and customization. WAPET bridges the gap between comprehensive modeling capabilities and user-friendly interfaces, positioning itself as a versatile tool for both experienced researchers and non-computational stakeholders in biodiversity decision-making processes.

1. Introduction

Accurate simulations help ecologists and conservationists effectively address ecosystem challenges and provide scientific foundations for developing reasonable strategies [1]. Species richness serves as a fundamental element of ecosystems and plays a crucial role in maintaining ecological balance, enhancing ecosystem services, and promoting human well-being [2]. By simulating species interactions, population dynamics, and ecological processes, simulations provide important insights that inform conservation strategies, guide policy decisions, and enhance ecosystem management [3,4]. Additionally, simulations offer a controlled and cost-effective framework for testing hypotheses and exploring potential outcomes of different interventions, providing support for both scientific research and practical applications [5]. As the pressures on ecosystems continue to increase, the demand for precise, user-friendly, and scalable simulation tools becomes more urgent. These tools assist researchers and decision-makers in formulating effective biodiversity conservation and ecosystem sustainability strategies. Furthermore, existing population dynamics simulation tools often rely on a single perspective for analysis, which may lead to partial observability issues during the simulation process. This limitation hinders a comprehensive reflection of the complex processes of species succession. Therefore, the simulator must integrate various biological models to provide multi-perspective analyses of population dynamics.

Based on the research, the current tools for multidisciplinary expert wildlife research still face three major challenges. (I) Insufficient multi-view assessment: specialized simulators [6,7,8], for example, specialized in free-roaming dogs, Ursus arctos, and Maned wolfs, despite their focused and tailored nature, may sometimes provide a narrower perspective on ecological systems, potentially leading to less comprehensive assessments [9]. (II) A high learning curve: General-purpose tools [10,11,12,13,14,15,16,17], including those designed for educational purposes [18,19,20], are applicable to a wide range of biological species and ecological scenarios. But they often require a certain level of programming expertise, which can present challenges for users without a computational background [21]. Programming is necessary for customized development, such as adding new dimensions or influencing factors for population dynamics simulations. Users without programming backgrounds may find it challenging to quickly implement their desired customizations. This technical barrier may unintentionally limit the participation of diverse stakeholders, particularly those from non-computational fields such as ecology or policy-making, potentially impeding the widespread dissemination and educational potential of these valuable tools. (III) The lack of seamless secondary development with Python (3.10.12): With the development of AI and data science, there are higher requirements for simulators, which need to integrate with AI and data science. Since Python is the dominant language in AI and data science, there is a need for simulator tools that integrate seamlessly with Python environments. However, many general-purpose tools tend to be commercial, closed-source, and developed in non-Python languages (e.g., Java), which limits users to specific platforms (e.g., only being able to use a certain company’s products or download specific tools). When users need to integrate with Python development environments, they still face significant difficulties and challenges (e.g., package management and multi-file import issues [22]). Even with open-source options available, they may still be developed in languages other than Python, presenting similar integration challenges. The Python-based simulator tool bridges the gap between comprehensive modeling capabilities and a user-friendly interface. It also facilitates smooth integration with Python to enhance rapid decision-making processes and broaden educational opportunities in the fields of population dynamics and biodiversity conservation [23,24].

To address these issues, we developed a novel software tool named the Wildlife Analysis and Population Ecology Tool (WAPET) (Python 3.10.12, code url: https://github.com/JoshuaYuYide/WAPET.git (accessed on 1 December 2024)) to address these challenges through a three-pronged approach. Firstly, to overcome challenge (I), WAPET incorporates Monte Carlo simulation [25] and integrates ecological models, including Logistic Growth [26], Random Walk [27,28], and Cellular Automata [29]. These models are widely used in ecological research to simulate population dynamics and spatial distribution from particular perspectives.

This approach enables WAPET to simulate population dynamics from multiple perspectives, including competition, reproductive networks, habitat analysis, and population trends, thus providing a more comprehensive assessment of ecological systems. Secondly, to address challenge (II), WAPET employs a fully parameterized input paradigm, allowing users without coding to easily explore simulations. Users can operate WAPET solely by inputting parameters on GUI. The paradigm facilitates the participation of diverse stakeholders, particularly those from non-computational fields such as ecology or policy-making, thereby enhancing the tool’s educational potential and accessibility. Lastly, to tackle challenge (III), WAPET’s development is entirely Python-based, utilizing PySide6 [30] and Mesa [31] libraries. This framework enables seamless development in Python environments, providing robust extensibility while avoiding common issues like package management or multi-file imports. Therefore, WAPET can function as a module for Python projects, imported into AI or data science projects, allowing seamless interaction with packages like Pytorch and scikit-learn, and so on. Additionally, WAPET employs agent-based modeling (ABM) [32] to simulate individual behaviors, offering a more realistic representation of natural systems by accounting for the randomness and variability inherent in individual organisms and their impact on population dynamics. This comprehensive approach positions WAPET as a versatile tool for both experienced researchers and non-computational stakeholders in biodiversity decision-making processes, bridging the gap between comprehensive modeling capabilities and user-friendly interfaces.

Our contributions are as follows: (I) We integrated Logistic Growth, Random Walk, and Cellular Automata via Monte Carlo simulation, offering WAPET multiple perspectives to gain a better understanding of ecological processes. (II) We developed a no-code mode of human-computer interaction in WAPET. The no-code simulation tool for population dynamics offers a low-barrier entry for biodiversity stakeholders and researchers. (III) We implemented a framework combining PySide6 for the GUI and Mesa for agent-based modeling, both Python-based libraries. This integration facilitates easy extension and customization, allowing users to adapt or rewrite components to suit their research needs.

2. Related Work

In the field of population dynamics and ecological simulations, several tools have been developed, each with its distinct advantages and limitations, shown in Table 1. Our work builds upon the foundation laid by various specialized and general-purpose simulators, addressing their gaps while integrating multiple models and a user-friendly interface.

2.1. Specialized Tools for Ecological Systems

Several specialized tools have been developed for simulating specific species or ecosystems. For instance, Shamsaddini et al. [6] focused on dynamic modeling for free-roaming dog population management, particularly in urban settings. Similarly, Wiegand et al. [7] created spatially explicit models for Ursus arctos (brown bear) to link landscape structure with population dynamics. Barbosa et al. [8] explored the consequences of roads on wildlife population dynamics, with a focus on Maned wolves. These specialized tools are invaluable for their focused assessments, but they tend to provide a narrower perspective, which may limit their applicability to broader ecological systems. In contrast, our tool, WAPET, integrates various ecological models to offer a more comprehensive view of population dynamics, addressing the limitations of partial observability in such specialized tools.

2.2. General-Purpose Simulators

General-purpose simulation platforms such as Gephi [10], Vensim [11], AnyLogic [12], and NetLogo [13] are widely used in population modeling. These platforms are versatile and can simulate a broad spectrum of ecological scenarios. However, they often require significant programming knowledge, which can be a barrier for non-computational users like ecologists or policy-makers. For example, platforms like GAMA [14], Repast [15], and Mason [16] provide extensive capabilities for agent-based modeling, but their complexity can deter users unfamiliar with coding. WAPET addresses this challenge by offering a no-code mode, allowing users to run simulations without the need for programming expertise. This feature makes it accessible to a wider audience, including researchers and stakeholders from various disciplines.

2.3. Educational Platforms

Educational tools such as PopBio [18], WiSP [19], and wildlife tracking guides [20] are designed for teaching and learning purposes. These tools are generally more user-friendly but are limited in terms of customization and extensibility. While they offer a good entry point for understanding population dynamics, they lack the flexibility required for in-depth ecological research. WAPET bridges this gap by combining ease of use with powerful modeling capabilities, allowing users to explore complex ecological processes through an intuitive interface.

2.4. Integration with Python and AI

With the rapid advancements in AI and data science, there is a growing need for simulation tools that integrate seamlessly with Python, the dominant language in these fields. Many existing simulators are either commercial, closed-source, or developed in languages other than Python, making it difficult to integrate them into AI-driven research workflows. WAPET, being fully Python-based, addresses this issue by utilizing PySide6 [30] for the GUI and Mesa [31] for agent-based modeling. This integration ensures compatibility with Python environments, allowing users to extend the tool’s functionality with AI and machine learning libraries like PyTorch [33] and scikit-learn [34]. This flexibility is particularly valuable for researchers looking to incorporate population simulations into broader data science projects.

In summary, while specialized tools provide focused assessments and general-purpose simulators offer versatility, both types of tools face challenges in terms of accessibility and integration with modern data science workflows. WAPET addresses these challenges by integrating multiple ecological models, offering a no-code interface for non-programmers, and ensuring seamless integration with Python environments. This makes WAPET a versatile and accessible tool for researchers and stakeholders involved in biodiversity conservation and ecological research.

3. Methodology

The WAPET framework consists of two components: the software framework and ecological models. The software framework provides users with an intuitive interface, allowing them to set parameters for the ecological models and execute simulations through code, subsequently delivering the simulation results back to users. The ecological models conduct more comprehensive simulations by considering the multidimensional influencing factors during the species succession process, thereby reducing the partial observability issues arising from single-factor analysis. Together, these components complement each other, enabling WAPET to effectively support ecological research and the formulation of conservation strategies.

3.1. Architecture

As shown in Figure 1, the architecture is structured into three layers: the Presentation Layer (user interface and visualized results), the Agent Layer (agents for environmental and species objects), and the Interaction Layer (interactions between agents generating simulation results). This structure enhances modularity and maintainability.

The information flow begins with the GUI receiving user input data or parameters, which triggers the initialization of the environment and species agents. These agents then enter the Interaction Engine, where they engage in dynamic interactions, resulting in updates to their states. The Simulation Controller subsequently generates results and directs further cycles according to a predetermined schedule. This process is repeated iteratively until the simulations are complete. Upon completion, the Simulation Controller sends the results to the visualization module, which presents the findings through the GUI.

We can abstract the interaction flow into a Markov decision process (MDP) [35]. An MDP is 5-tuple $\mathbb{M}=⟨\mathcal{S},\mathcal{A},\mathbf{R},\mathbf{T},\gamma ⟩$, where the $\mathcal{S}$ is the state space, the $\mathcal{A}$ is the action space, $\mathbf{R}:\mathcal{S}\times \mathcal{A}\to \mathcal{R}$ is the reward function, $\mathcal{R}$ is the reward set, $\mathbf{T}:\mathcal{S}\times \mathcal{S}\to \mathcal{S}$ is the state transition function, and $\gamma$ is the discount factor.

An agent at the current state, $s_t$, chooses an action, $a_t$, and give this action to the environment. Then, the environment responds to a reward, r, through the reward function $\mathbf{R}(s_t,a_t)$ and the next state $s_{t+1}$ via the state transition function $\mathbf{T}(s_t,a_t)$. This process runs constantly until the end of the episode, $t=T\in \mathcal{T}$.

A state, $s\in \mathcal{S}$, consists of multiple observational dimensions, with each dimension representing an ecological model. However, when the number of dimensions is limited, the state, s, may fail to fully capture the dynamics of the environment. In this case, the agent receives an observation, o, where the dimensionality of o is smaller than that of the true state, s. This leads the agent to face a partially observable Markov decision process (POMDP) [36] problem. Incorporating more ecological models increases the number of dimensions, thereby mitigating the issue of partial observability.

3.2. GUI

Figure 2 and Figure 3 show the human–computer interaction interface. The main interface is divided into three regions: left, center, and right. The central region in Figure 2 includes the Species Data Management Panel for population parameters and the Environment Data Management Panel for environmental parameters. The rightmost section is the Operation Panel for simulation controls.

Figure 3 shows the landscape configuration. The landscape is a grid with adjustable dimensions. Users can generate random maps or manually modify grid types. Population distribution heat maps and environmental carrying capacity maps can be rendered. The left region displays the Species Data Table and Simulation Result Table. The right region features the Plotting Panel for visualizing simulation results.

3.3. Environment Agent

The Environment Agent represents a realistic natural environment, with the primary function of interacting with the Species Agent and providing feedback based on these interactions. In this paper, the Environment Agent is constructed by the Soil Agent and the Climate Agent, which simulate soil conditions and climate factors, respectively. This design allows the Environment Agent to offer more authentic and multidimensional environmental feedback to the Species Agent.

3.3.1. Soil Agent

The Soil Agent simulates the landscape, representing habitat changes and their impacts on population dynamics. It is an array with grid cells containing information on carrying capacity and species occupancy.

The Soil Agent consists of many grids. Mathematically, let a Soil Agent be a function of $\mathsf{\Sigma }\to \mathcal{N}\times \mathcal{M}$, where $\mid \mathcal{N}\mid \le 10$ and $\mid \mathcal{M}\mid \le 10$ are the map’s size. For any grid ${\varsigma }_i=[n,m]\in \mathsf{\Sigma }$, $f_c\left(\mathsf{\Sigma }\right)\to C$ is the carrying ability function, $f_b(\mathsf{\Sigma },K)\to B$ is the function of the Species Agent set K on the grid, and $f_y\left(\mathsf{\Sigma }\right)\to Y$ is the soil type function.

Users initialize the soil type function $f_y\left(\mathsf{\Sigma }\right)$ with the carrying ability mean ${\mu }_c$ and standard deviation ${\sigma }_c$, sampling data from the normal distribution [37]:

(1)$f(x\mid {\mu }_c,{\sigma }_c)={\displaystyle \frac{1}{{\sigma }_c\sqrt{2\pi }}}{e}^{-{\displaystyle \frac{{(x-{\mu }_c)}^2}{2{\sigma }_c^2}}}.$

The Soil Agent updates the carrying ability and Species Agent number via $f_c\left({\varsigma }_i\right),f_b({\varsigma }_i,k)$ at each time-step t.

3.3.2. Climate Agent

Climate impacts carrying capacity, with extreme weather affecting land’s ability to support populations [38,39]. We model extreme weather as events deviating from the normal state [40]. We offer three climate models: tropical rainforest, tropical desert, and the Mediterranean. Each has a discrete probability distribution, sampled each time-step. Prolonged single weather types reduce the carrying capacity, simulating extreme weather impacts.

3.4. Species Agent

Species Agents have survival, locomotion, and reproduction abilities. Survival includes attack and escape abilities, hunger rates, growth rates, and survival rates. Users initialize locomotion speed with the mean ${\mu }_s$ and the standard deviation ${\sigma }_s$. Species Agents sample speed from this distribution at each time-step. Reproductive capacity allows mating within the same grid, with options for monogamy.

New Species Agents arise from initial setup and reproduction. In multi-species scenarios, there are predator, target, and prey agents. This food chain impacts each species’ survival and reproduction, modeled by their interactions.

3.5. Interaction Engine

3.5.1. Logistic Growth Model

The Interaction Engine integrates ecological models through Monte Carlo simulation. We use the Logistic Growth Model for feeding and reproduction success rates. The continuous-time differential equation is as follows:

(2)${\displaystyle \frac{dN}{dt}}=rN\left(1-{\displaystyle \frac{N}{\mathcal{K}}}\right),$

where N is the population size, r is the growth rate, and $\mathcal{K}$ is the carrying capacity. The discrete-time form is as follows:

(3)$N_{t+1}=N_t+r{N}_t(1-{\displaystyle \frac{N_t}{K}}).$

We sample environmental regulation $P\left({\displaystyle \frac{1-N}{K}}\right)$ and individual success rate $P\left(\right(1+D)·r)$, where D is attenuation.

3.5.2. Random Walk Model and Cellular Automata

The random walk model represents the Species Agent movement; the formula of a Species Agent position changing along with the time horizon is as follows:

(4)$x(t+\Delta t)=x\left(t\right)+\eta \Delta t,$

where $\eta$ is movement distance per time-step, and $x\left(t\right)$ is the start location.

Although species move randomly across the landscape, their movement decisions are also influenced by other factors. For instance, they tend to move toward areas where conspecifics are concentrated. The tendency follows the movement of the cellular Automata model, which is as follows:

(5)$G_{\mathrm{new}}(i,j)=\left\{\begin{array}{cc}1\hfill & \mathrm{if}G(i,j)=0\mathrm{and}\sum _{\theta }G(i^{\prime },j^{\prime })=\phi \hfill \\ 1\hfill & \mathrm{if}G(i,j)=1\mathrm{and}\phi -1\le \sum _{\theta }G(i^{\prime },j^{\prime })\le \phi \hfill \\ 0\hfill & \mathrm{otherwise}\hfill \end{array}\right.$

where ${\sum }_{\theta }G(i^{\prime },j^{\prime })$ is the sum of $\phi$ neighbors of cell $[i,j]$. Cells can be as follows:

• Birth: dead cell (0) becomes living (1) if it has three living neighbors.

• Survival: living cell (1) remains alive with two or three living neighbors.

• Death: living cell (1) dies (0) if neighbors are not two or three.

Species Agents tend to move towards survival cells. The tendency probability is $P({\sum }_{\theta }\mid \sum \Theta )$, with the maximum distance $f(x\mid {\mu }_s,{\sigma }_s)·1$.

3.5.3. Other Models

Attack and Escape

In multi-species scenarios, food chain dynamics emerge. Each species has attack, $\circ a$, and escape capabilities, $•a$, as actions, modeled probabilistically. Ås the age $t_k$ grows, the Species Agent’s k attack and escape capabilities fluctuate. For a Species Agent k, the formula of the attack capability changing is

(6)$\circ a_{t+1}^k=\left\{\begin{array}{cc}{(t_k-\circ t)}^{{\displaystyle \frac{1}{2}}}\times t_k\times \beta \times \circ a_t^k\hfill & \mathrm{if}\mathrm{married}\hfill \\ t_k\times \beta \times \circ a_t^k\hfill & \mathrm{if}\mathrm{not}\mathrm{married}\hfill \end{array}\right.$

where $\beta$ is the coefficient of variation of predatory capacity, and $\circ t$ is the marriage age of the Species Agent k. the formula of the escape capability changing is

(7)$•a_{t+1}^k=\left\{\begin{array}{cc}{(t_k-\circ t)}^{{\displaystyle \frac{1}{2}}}\times t_k\times \beta ·•a_t^k\hfill & \mathrm{if}\mathrm{married}\hfill \\ t_k\times \beta \times •a_t^k\hfill & \mathrm{if}\mathrm{not}\mathrm{married}\hfill \end{array}\right.$

Predation success depends on relative strengths, leading to capture or survival. The stronger the predation ability of a predator, the higher the probability of capturing prey. However, if the prey has a stronger escape ability, the probability of evading predation increases. Predation serves as another primary source of food, in addition to the food provided via the environment.

Mating and Giving Birth

In the simulated ecosystem, the reproductive behavior of individual species is a crucial factor for species continuation and population dynamics. To simulate the mating and reproductive processes among individuals, this paper proposes a set of multi-condition mating rules. Specifically, individuals of the same species, $k_1,k_2\in K$, can mate if they meet the following conditions: they must be of opposite sexes, be located in the same grid, reach the appropriate mating age, and be unmarried and in a living state.

In the model, the mating behavior of individuals is controlled through the following methods:

1. Mating condition assessment: First, check whether the individual’s age exceeds the specified minimum mating age, $\beta$, and the individual must be alive. Additionally, the individual must be of the opposite sex to the potential partner. For monogamous individuals, it further checks whether the individual already has a partner. If all these conditions are met, they indicate that the individual can mate; otherwise, they indicate that mating is not possible.

   Mathematically, this rule can be expressed as follows:

   (8)$\mathrm{can}\text{\_}\mathrm{married}(k_1,k_2)=\left\{\begin{array}{cc}\mathrm{True},\hfill & \mathrm{if}{t}_{k1}>\beta \wedge G\left(k_1\right)\ne G\left(k_2\right)\wedge L\left(k_1\right)\wedge \neg D\left(k_1\right)\hfill \\ \mathrm{False},\hfill & \mathrm{otherwise}\hfill \end{array}\right.$

   where $G\left(k\right)$ is the gender of the agent k, $L\left(k\right)$ is the agent k’s survival, and $D\left(K\right)$ is the agent k being married.

2. Mating execution: When the mating conditions are satisfied, the Species Agent $k_1$ attempts to find potential mating partners within the grid where the current individual resides. By traversing the same grid for conspecifics, if both individuals $k_1,k_2$ meet the mating conditions, they will pair up as partners, updating each other’s partner and $D\left(k_1\right)=True$ status to reflect their new married state.

3. Reproductive process: Following successful mating, individuals can reproduce. The reproductive behavior of the population is regulated by population density. Specifically, the population size N and the environmental carrying capacity $\mathcal{K}$ affect the reproductive probability, which is calculated using the logistic growth model:

   (9)$P\left(\mathrm{birth}\right)={\displaystyle \frac{r·N}{\mathcal{K}}}\times \alpha \times (t_k-\beta )$

   where r represents the growth rate, N is the current population size in the grid, $\mathcal{K}$ is the carrying capacity of that grid, and $\alpha$ is the fertility factor for individual fertility as age changes. If the reproductive conditions are met, a new individual is generated, and its attributes (such as location, parent information, etc.) are initialized.

This mating and reproductive rule not only considers biological characteristics such as gender and age but also incorporates mating patterns (e.g., monogamy) and environmental factors (e.g., population density), ensuring the rationality and complexity of the simulation process. Through these rules, the model can dynamically simulate the mating and reproductive behaviors of individual species, further facilitating the analysis of population dynamic changes and their impacts on the ecosystem.

3.6. Simulation Controller

3.6.1. Scheduling Overview

The Simulation Controller manages time, agent, and execution order scheduling. We use the mesa.time.SimultaneousActivation scheduler for simultaneous agent updates. Soil Agents refresh species count and carrying capacity. The Climate Agent selects weather based on its probability distribution at each time-step.

3.6.2. Agent Scheduling

The scheduling order for any Species Agent follows a systematic sequence. Initially, the agent’s vitality status is checked. Deceased agents are removed from the simulation, and new agents are introduced. Next, the agent performs a random movement within the environment. The agent then consumes available food resources. If the agent is not attached, it may participate in mating, forming a pair bond with a compatible agent. Successful mating can lead to reproduction, generating offspring. Finally, the agent undergoes growth, advancing in age and potentially transitioning through different life stages.

3.6.3. Component Interaction

The Simulation Controller coordinates interactions among various components. It initializes the Environment Agent and randomly distributes Species Agents. Based on the user-defined maximum time-step, the Simulation Controller invokes the internal scheduling of each agent class in every iteration. The internal scheduling of agent classes partially derives results from the Interaction Engine. The Simulation Controller then consolidates these results and transmits them to the Visualization Tool. This coordinated workflow enables a comprehensive and synchronized simulation of the ecosystem.

3.7. Visualization Tool

The visualization toolkit offers various graphical display methods designed to help researchers intuitively understand and analyze complex data within simulated ecosystems. First, the gender ratio pie chart clearly illustrates the proportion of male and female individuals in the ecosystem, aiding researchers in understanding the gender distribution. Second, the age structure pie chart categorizes individuals into different age groups, facilitating the observation of the quantity distribution and changing trends of individuals at various age stages.

Additionally, the alive status pie chart displays the ratio of surviving to deceased individuals, reflecting the survival status of individuals in the ecosystem and enabling researchers to assess the health of the ecosystem. The population change curve graph serves as a dynamic tool that illustrates how population numbers fluctuate over time, helping researchers grasp the growth, decline, or volatility of populations from a temporal perspective. Finally, the reproductive network tree provides an intuitive view by showcasing the familial relationships between individuals, allowing researchers to easily trace the breeding history and lineage structure of individuals.

Furthermore, the population distribution map records the density of populations on the landscape after the simulation is complete. This map reflects the geographical distribution of populations following random movement to some extent. The land carrying capacity map displays the environmental carrying capacity of each soil agent (grid) on the landscape, which influences the population density. The real population line chart represents the actual trends in population number changes, allowing for the validation of the results and trends observed in the simulated population line chart.

Figure 4 showcases the diverse visualization capabilities of WAPET, demonstrating its ability to present complex ecological data in an accessible and informative manner. The figure comprises seven simulation-generated visualizations (Figure 4a–g) and one real-world data plot (Figure 4h), illustrating the tool’s comprehensive approach to population dynamics analysis. Specifically, the visualizations include the following: (Figure 4a) the gender ratio, (Figure 4b) the age distribution, (Figure 4c) birth and death rates, (Figure 4d) the simulated population trend, (Figure 4e) the reproductive network, (Figure 4f) the spatial distribution of the population within the habitat, and (Figure 4g) the land carrying capacity distribution. The simulated population trend (Figure 4d) is juxtaposed with the actual species data curve (Figure 4h) from the UK butterfly population study [41] to demonstrate the tool’s predictive capabilities. While the exact replication of real-world data is not the primary goal of this demonstration, the simulated trend closely approximates the general trajectory of population decline observed in the actual data. It is worth noting that the simulated curve (Figure 4d) exhibits non-smooth characteristics and inherent randomness, which are natural consequences of employing ABM and Monte Carlo simulation techniques. Despite these differences in curve smoothness, both the simulated (Figure 4d) and actual (Figure 4h) data converge on a declining population trend, underscoring WAPET’s potential as a valuable tool for predicting broad ecological patterns. These visualizations collectively demonstrate WAPET’s capacity to provide a multi-dimensional understanding of population dynamics, offering valuable insights for biodiversity conservation efforts and policy-making. It is important to note that the visualizations presented in Figure 4 are primarily intended to showcase the functionality of the WAPET tool, rather than to provide definitive ecological analyses. By simulating possible real-world conditions such as butterfly gender, age, survival, population changes, and reproductive networks, WAPET aids in understanding butterfly population dynamics and their influencing factors, even when the direct observation of these factors is challenging.

4. Conclusions

We proposed a framework named WAPET, a user-friendly, programming-free platform designed to simulate wildlife population dynamics and environmental impacts. WAPET addressed the limitations of existing tools by integrating multiple ecological models, including the Logistic Growth Model, Random Walk Model, and Cellular Automata, with Monte Carlo simulation techniques. It was developed entirely in Python using PySide6 and Mesa libraries, enabling seamless secondary development and robust extensibility. Key contributions included providing a low-barrier entry point for biodiversity policy stakeholders and researchers, offering multi-perspective simulations for a high-dimensional understanding of ecological processes, and implementing a novel framework that combined a graphical user interface and agent-based modeling. The Simulation Controller efficiently managed agent interactions, ensuring synchronized and comprehensive ecosystem simulations. The visualization toolkit offered diverse graphical representations, allowing researchers to analyze complex ecological data and identify trends and patterns. WAPET enhanced the prediction of wildlife population dynamics and supported effective conservation strategies through its accessible and comprehensive simulation capabilities.

Author Contributions

Conceptualization, Y.Y.; methodology, Y.Y.; software, Y.Y.; validation, Y.Y.; formal analysis, Y.Y.; investigation, Y.Y.; resources, Y.Y.; data curation, Y.Y.; writing—original draft preparation, Y.Y. and H.L.; writing—review and editing, Y.M. and Y.L.; visualization, Y.Y.; supervision, Y.M. and Y.L.; project administration, Y.Y. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The raw data supporting the conclusions of this article will be made available by the authors on request.

Conflicts of Interest

The authors declare no conflicts of interest.

Footnotes

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Figures and Table

Figure 1. Information flow of the proposed WAPET framework diagram.

Figure 2. Main panel of the proposed WAPET.

Figure 3. Landscape Panel of the proposed WAPET.

Figure 4. Visualization result of butterfly population rehearsal simulation. Subfigures (a,b,c,e) show the sex ratio, age distribution, mortality rate, and reproductive network respectively. For (d,h), the horizontal axis represents the timeline (unit: time steps for (d) and years for (h)), and the vertical axis represents the quantity. For (f,g), red represents the inaccessible land, green represents the population size, and blue represents the land carrying capacity, with darker colors having higher values. The data of (h) indicates smoothing indicators of population change, while (d) depicts non-smoothing indicators.

Table 1

Comparison of WAPET and other classical methods.

Methods	Specialized Tools	General-Purpose Simulators	EducationPlatforms	WAPET
Open-source	✓	−	✓	✓
Multi-view	−	✓	✓	✓
No-code	−	−	−	✓
Python-based	−	−	−	✓

References

1. Coulson, T.; Guinness, F.; Pemberton, J.; Clutton-Brock, T. The demographic consequences of releasing a population of red deer from culling. Ecology; 2004; 85, pp. 411-422. [DOI: https://dx.doi.org/10.1890/03-0009]

2. Mace, G.M.; Norris, K.; Fitter, A.H. Biodiversity and ecosystem services: A multilayered relationship. Trends Ecol. Evol.; 2012; 27, pp. 19-26. [DOI: https://dx.doi.org/10.1016/j.tree.2011.08.006] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/21943703]

3. Clark, J.S.; Carpenter, S.R.; Barber, M.; Collins, S.; Dobson, A.; Foley, J.A.; Lodge, D.M.; Pascual, M.; Pielke, R., Jr.; Pizer, W. et al. Ecological forecasts: An emerging imperative. Science; 2001; 293, pp. 657-660. [DOI: https://dx.doi.org/10.1126/science.293.5530.657] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/11474103]

4. Hanski, I. Habitat loss, the dynamics of biodiversity, and a perspective on conservation. Ambio; 2011; 40, pp. 248-255. [DOI: https://dx.doi.org/10.1007/s13280-011-0147-3]

5. Bankes, S. Exploratory modeling for policy analysis. Oper. Res.; 1993; 41, pp. 435-449. [DOI: https://dx.doi.org/10.1287/opre.41.3.435]

6. Shamsaddini, S.; Ahmadi Gohari, M.; Kamyabi, H.; Nasibi, S.; Derakhshani, A.; Mohammadi, M.A.; Mousavi, S.M.; Baneshi, M.R.; Hiby, E.; Harandi, M.F. Dynamic modeling of female neutering interventions for free-roaming dog population management in an urban setting of southeastern Iran. Sci. Rep.; 2022; 12, 4781. [DOI: https://dx.doi.org/10.1038/s41598-022-08697-w]

7. Wiegand, T.; Moloney, K.A.; Naves, J.; Knauer, F. Finding the missing link between landscape structure and population dynamics: A spatially explicit perspective. Am. Nat.; 1999; 154, pp. 605-627. [DOI: https://dx.doi.org/10.1086/303272]

8. Barbosa, P.; Schumaker, N.H.; Brandon, K.R.; Bager, A.; Grilo, C. Simulating the consequences of roads for wildlife population dynamics. Landsc. Urban Plan.; 2020; 193, 103672. [DOI: https://dx.doi.org/10.1016/j.landurbplan.2019.103672]

9. Fulton, E.A. Approaches to end-to-end ecosystem models. J. Mar. Syst.; 2010; 81, pp. 171-183. [DOI: https://dx.doi.org/10.1016/j.jmarsys.2009.12.012]

10. Bastian, M.; Heymann, S.; Jacomy, M. Gephi: An open source software for exploring and manipulating networks. ICWSM; 2009; 8, pp. 361-362. [DOI: https://dx.doi.org/10.1609/icwsm.v3i1.13937]

11. Eberlein, R.L.; Peterson, D.W. Understanding models with Vensim™. Eur. J. Oper. Res.; 1992; 59, pp. 216-219. [DOI: https://dx.doi.org/10.1016/0377-2217(92)90018-5]

12. Borshchev, A. Multi-method modelling: AnyLogic. Discrete-Event Simulation and System Dynamics for Management Decision Making; John Wiley & Sons: Hoboken, NJ, USA, 2014; pp. 248-279.

13. Tisue, S.; Wilensky, U. NetLogo: A simple environment for modeling complexity. Int. Conf. Complex Syst.; 2004; 21, pp. 16-21.

14. Grignard, A.; Taillandier, P.; Gaudou, B.; Vo, D.A.; Huynh, N.Q.; Drogoul, A. GAMA 1.6: Advancing the art of complex agent-based modeling and simulation. Proceedings of the PRIMA 2013: Principles and Practice of Multi-Agent Systems: 16th International Conference; Dunedin, New Zealand, 1–6 December 2013; Springer: Cham, Switzerland, 2013; pp. 117-131.

15. Collier, N. Repast: An extensible framework for agent simulation. Nat. Resour. Environ. Issues; 2001; 8, 4.

16. Luke, S.; Cioffi-Revilla, C.; Panait, L.; Sullivan, K.; Balan, G. Mason: A multiagent simulation environment. Simulation; 2005; 81, pp. 517-527. [DOI: https://dx.doi.org/10.1177/0037549705058073]

17. Minar, N.; Burkhart, R.; Langton, C.; Askenazi, M. The Swarm Simulation System: A Toolkit for Building Multi-Agent Simulations; SFI Working Paper 1996-06-042 Santa Fe Institute: Santa Fe, NM, USA, 1996.

18. Stubben, C.; Milligan, B. popbio: Construction and analysis of matrix population models. R Package, version 2.8; R Project: Vienna, Austria, 2007.

19. Salguero-Gómez, R.; Plotkin, J.B. WiSP: A web-based interactive simulation program for population dynamics. Environ. Model. Softw.; 2012; 37, pp. 230-232.

20. Conroy, M.J.; Carroll, J.P. A Guide to Wildlife Radio Tracking; Springer Science & Business Media: Berlin/Heidelberg, Germany, 2011.

21. Joppa, L.N.; McInerny, G.; Harper, R.; Salido, L.; Takeda, K.; O’Hara, K.; Gavaghan, D.; Emmott, S. Troubling trends in scientific software use. Science; 2013; 340, pp. 814-815. [DOI: https://dx.doi.org/10.1126/science.1231535]

22. Marru, S.; Gunathilake, L.; Herath, C.; Tangchaisin, P.; Pierce, M.; Mattmann, C.; Singh, R.; Gunarathne, T.; Chinthaka, E.; Gardler, R. et al. Apache airavata: A framework for distributed applications and computational workflows. Proceedings of the 2011 ACM Workshop on Gateway Computing Environments; Seattle, WA, USA, 18 November 2011; pp. 21-28.

23. Decker, D.J.; Riley, S.J.; Siemer, W.F. Human Dimensions of Wildlife Management; JHU Press: Baltimore, MD, USA, 2012.

24. Berkes, F. Evolution of co-management: Role of knowledge generation, bridging organizations and social learning. J. Environ. Manag.; 2009; 90, pp. 1692-1702. [DOI: https://dx.doi.org/10.1016/j.jenvman.2008.12.001]

25. Clark, J.S. Testing disturbance theory with long-term data: Alternative life-history solutions to the distribution of events. Am. Nat.; 1996; 148, pp. 976-996. [DOI: https://dx.doi.org/10.1086/285967]

26. Verhulst, P.F. Notice sur la loi que la population suit dans son accroissement. Corresp. Math. Phys.; 1838; 10, pp. 113-129.

27. Codling, E.A.; Plank, M.J.; Benhamou, S. Random walk models in biology. J. R. Soc. Interface; 2008; 5, pp. 813-834. [DOI: https://dx.doi.org/10.1098/rsif.2008.0014]

28. Kareiva, P.; Shigesada, N. Analyzing insect movement as a correlated random walk. Oecologia; 1983; 56, pp. 234-238. [DOI: https://dx.doi.org/10.1007/BF00379695] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/28310199]

29. Hogeweg, P. Cellular automata as a paradigm for ecological modeling. Appl. Math. Comput.; 1988; 27, pp. 81-100. [DOI: https://dx.doi.org/10.1016/0096-3003(88)90100-2]

30. Fitzpatrick, M. Create GUI Applications with Python & Qt6 (PySide6 Edition); Martin Fitzpatrick: Amersfoort, The Netherlands, 2021.

31. Masad, D.; Kazil, J.L. Mesa: An Agent-Based Modeling Framework. Proceedings of the SciPy; Austin, TX, USA, 6–12 July 2015; pp. 51-58.

32. Grimm, V.; Revilla, E.; Berger, U.; Jeltsch, F.; Mooij, W.M.; Railsback, S.F.; Thulke, H.H.; Weiner, J.; Wiegand, T.; DeAngelis, D.L. Pattern-oriented modeling of agent-based complex systems: Lessons from ecology. Science; 2005; 310, pp. 987-991. [DOI: https://dx.doi.org/10.1126/science.1116681] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/16284171]

33. Paszke, A.; Gross, S.; Massa, F.; Lerer, A.; Bradbury, J.; Chanan, G.; Killeen, T.; Lin, Z.; Gimelshein, N.; Antiga, L. et al. Pytorch: An imperative style, high-performance deep learning library. Adv. Neural Inf. Process. Syst.; 2019; 32, pp. 8026-8037.

34. Kramer, O. Scikit-learn. Machine Learning for Evolution Strategies; Springer: Cham, Switzerland, 2016; pp. 45-53.

35. Bellman, R. Dynamic programming. Science; 1966; 153, pp. 34-37. [DOI: https://dx.doi.org/10.1126/science.153.3731.34]

36. Kaelbling, L.P.; Littman, M.L.; Cassandra, A.R. Planning and acting in partially observable stochastic domains. Artif. Intell.; 1998; 101, pp. 99-134. [DOI: https://dx.doi.org/10.1016/S0004-3702(98)00023-X]

37. Laplace, P.S. Théorie Analytique des Probabilités; Courcier: Paris, France, 1820.

38. Curtis, S.; Fair, A.; Wistow, J.; Val, D.V.; Oven, K. Impact of extreme weather events and climate change for health and social care systems. Environ. Health; 2017; 16, pp. 23-32. [DOI: https://dx.doi.org/10.1186/s12940-017-0324-3]

39. Reichstein, M.; Bahn, M.; Ciais, P.; Frank, D.; Mahecha, M.D.; Seneviratne, S.I.; Zscheischler, J.; Beer, C.; Buchmann, N.; Frank, D.C. et al. Climate extremes and the carbon cycle. Nature; 2013; 500, pp. 287-295. [DOI: https://dx.doi.org/10.1038/nature12350]

40. Naveau, P.; Nogaj, M.; Ammann, C.; Yiou, P.; Cooley, D.; Jomelli, V. Statistical methods for the analysis of climate extremes. Comptes Rendus Géoscience; 2005; 337, pp. 1013-1022. [DOI: https://dx.doi.org/10.1016/j.crte.2005.04.015]

41. UK Office for National StatisticsDepartment for Environment, Food & Rural Affairs. Smoothed [Dataset]. Available online: https://www.gov.uk/government/statistics/butterflies-in-the-wider-countryside-uk (accessed on 1 December 2024).