# 안유순박사님 인터뷰 정리 - ROFASSS
###### tags: `Journal Review`

## Interview Title
How can Complex Adaptive Systems Suggest Possible Strategies to Mitigate Soil Degradation in North Korea?

* Interviewer: Dr.Hyesop Shin (Research Associate, University of Glasgow, UK)
* Interviewee: Dr.Yoosoon An (Research Fellow, Institute for Korean Regional Studies, Seoul National University)(Institute~는 균형을 위해 빼도 됨)
* ToC
- Self-Introduction
- Thesis Oveview
- Study Area
- How Agent-based Modelling was applied
- Key findings
- Episodes during research
- Final Comments

---
## Can you introduce yourself?
Hello. I am Dr. Yoosoon An (He/Him), currently serving as a research fellow at the Institute for Korean Regional Studies, Seoul National University. My primary research interests include North Korea, Agent-based Modelling, and other related areas. My focus lies in <span style="color:blue">N.Korea, Agent-based Modelling, and relationship between soil and food security</span>. My journey with complexity science and ABM has been nearly 10 years!

---
## First of all, what made you investigate the food problems in N.Korea?
* Why was N.Korea so attractive so start at the beginning, instead of other areas?
* Any personal motivations? (e.g. research project, )

---
## Can you give us an overview of your PhD thesis?
Of course. My doctoral research, titled "A Study on Land Degradation and Declining Food Production based on the Concept of Complex Adaptive System: Focusing on the North Korean Famine in the 1990s," explores mitigative strategies for land degradation and food shortages in North Korea by delving into the catastrophic famine of the 1990s. I utilized an agent-based model to scrutinize land cover changes from a bottom-up approach and posited scenarios to hypothesize the potential outcomes, given alternate policies during the 1990s.

For context, the North Korean famine of the mid-90s was a calamitous event influenced by a combination of natural hazards, primarily droughts, economic isolation following the Soviet Union's collapse, and agro-economic system failures. Land degradation emerged as an inescapable consequence, putting millions at risk of hunger. Additional information on this event is available on its Wikipage (https://en.wikipedia.org/wiki/North_Korean_famine).

My thesis encompasses three main pillars:

1. Historical Analysis: Before delving into modelling chapters, an examination of past satellite imagery and socio-economic data was conducted to comprehend the spatial and temporal changes <span style="color:red">from 1960s to 1990s. In this chapter, I focused on relationship among them in social-ecological system context, and </span>. (Note: An explanation of the application or relevance of the social-ecological process model could enhance clarity here).

2. Early Warning Signal (EWS) Application: The second pillar involves employing the EWS method to statistically identify critical transitions, with the primary focus on detecting deteriorating soil quality.

3. Multi-Agent System for Land-Use and Cover Change (MAS-LUCC): The final pillar constructs a MAS-LUCC to simulate the agricultural social-ecological system of North Korea. Contrasting the statistical approach, the agent-based model (ABM) study incorporates household ages to manage their crops and runs assorted scenarios to assess whether the famine's impact could have been mitigated to a certain degree.

Through this research, I aim to unravel the intricacies of the relationship between land degradation and food production, providing insights that may pave the way for future policy development and intervention strategies in analogous situations.

<details> Land Degradation
토지생산성과 기근에 대한 연구에서 그동안 다룬 바와 같이 토지황폐화로 인해서 식량생산성이 떨어지면 사람들은 식량에 압박을 느끼고 식량에 대해 집약적(intensive)으로 이용하면서 다시 토지황폐화가 발생하는 악순환을 겪게 됩니다. 개념적으로 Soil Degradation이나 Land Degradation이냐 논쟁은 있지만,<span style="color:blue">Soil Degradation은 토양의 저하 측면에 치중한 개념이고 Land Degradation은 토양에서 생장하는 생물(작물)과 토지이용, 즉 사람에 집중할 뿐, 인간과 자연환경의 피드백으로 발생하는 토양과 인간의 지속가능성 저하라는 점에서 거의 같은 용어입니다. 이 현상에서 토지 또는 토양황폐화 현상 자체는 자연환경에서 발생하는 현상이고, 식량생산 감소와 기근문제는 인간에 의한 현상이라고 구분할 수 있습니다. 그러나, 이들은 사실 따로 떨어트려서 생각할 수 없는 문제입니다. 토지황폐화와 기근문제의 인간과 자연 사이의 긴밀한 상호작용 관계는 사회생태시스템(Social-Ecological System)이라는 개념으로 잘 알려져 있으며, 토지황폐화 관련 문제에도 관련 연구의 도전 요인으로 2000년대 후반부터 지속적으로 제기되어 왔습니다. 토지황폐화와 기근문제가 사회생태시스템이기 때문에 생기는 문제는 Reynolds et al.(2007)이 지적한 바와 같이 이 두 요소간에 상호작용으로 인한 복잡성입니다. 이로 인해 토지황폐화로 인하여 발생하는 사막화, 토양질 저하, 그리고 기근 문제는 그 시점과 범위를 예측하기 어렵다는 문제를 가지고 있습니다. 그래서 최근 10여년간 토지황폐화를 사회생태시스템과 복잡적응계적 관점으로 바라보는 연구가 많이 진행된 바 있습니다. 그러나 지난 30여년 간 토지황폐화와 기근문제 중 가장 심각한 피해를 야기한 사건 중 하나인 1990년대 북한 대기근에 대한 연구에서는 이런 접근이 진행된 바 없습니다. 예를 들어 경제학자들은 경제체제 문제와 공급실패로 인한 기근만 강조하였으며, 생태학자들은 산림파괴로 인한 토양침식과 산사태와 같은 문제만 강조하였습니다. 이 둘을 연결시키려는 시도는 드물었으며 특히 복잡적응계를 적용하려는 시도는 특히 그러했습니다. 이 것이 제가 북한의 토지황폐화와 대기근에 복잡적응계적 방법론과 행위자기반모형을 적용한 이유입니다.</span>
</details>

---
## Study area and the Study Period
**Study Area and Period: North-eastern China and North Korea**
My research focuses on food sustainability and socio-ecological systems in North-eastern Asia, including China and the Korean peninsula. I've noticed similarities between the North Korean famine of the 1990s and previous Korean famines. These factors include land degradation, political isolation, and the effects of climate change. By delving deeper into the North Korean famine of the 1990s, I hope to gain a better understanding of the factors that influence food sustainability in the Korean Peninsula and East Asia.

**Rationale for Simulating the 1960s**
The 1960s stands out for several key structural and pragmatic reasons. Intriguingly, North Korea adopted the "shared ownership system" in 1946, granting individuals private land ownership rights. Yet, in the aftermath of the Korean War (1950-1953), a transition took place wherein all private lands were assimilated into collective farms by 1960. The agricultural paradigm in place today predominantly traces its roots back to the 1960s.

While conducting my research, I noticed a surge in documentation and statistical data collection starting from the 1960s, leading me to surmise that this era must have held significant importance.

From a socio-ecological perspective, I theorize that the severe famine was an outcome of a confluence of various phenomena. Each of these phenomena had distinct thresholds that, once surpassed, contributed to the crisis. Thus, I deemed it insightful to begin my simulation from the 1960s, examining land cover changes up until 2020. This timeframe should allow for a comprehensive analysis of the critical tipping points of each parameter, enabling the formulation of effective mitigation scenarios.

---
## Agent-based Modelling
### You used a Proof-Of-Concept (POC) model for your simulation. Can you tell us the pros and cons of using conceptual models?
The primary advantage is that it can function effectively in crucial areas where data is absent. Even though my model was somewhat abstract due to the scarcity of micro-data from N.Korea, I was able to simulate the spatial and temporal progression of N.Korea's land degradation using the remarkably consistent information in their reports.

Furthermore, Proof of Concept (POC) models generally offer a more straightforward interface, making it easier for users to comprehend the intended actions. This simplicity allows users to adjust parameters and apply scenarios with minimal challenges. Overall, debugging the model is straightforward. In contrast, with models grounded in real-world data, it can be challenging to determine if results stem from the spatial settings or the variables themselves, given their fixed positions. This complexity is why [Parker et al. (2003)](https://onlinelibrary.wiley.com/doi/abs/10.1111/1467-8306.9302004) referred to POC models as social laboratories.

On the downside, the outcomes from these models can be somewhat ambiguous, meaning they might not always align closely with real-world applications. While a virtual model can support a claim or hypothesis, it may not be the best tool for assessing policy impacts. dditionally, every time the additional data or an algorithm is added to the model, a “process of abstraction” is required e.g. one grid pixel represents 1 household. The small assumptions can add up to a greater crudeness. I should also have to mention the less intuition from the interface.

### You set the unit of tick to 1 year. What factors did you consider while using Discrete Event Simulation?
The model I created represents both agricultural landscapes and social-ecological systems. In temperate to cold agricultural regions like North Korea, farming and harvest schedules are typically based on an annual cycle. Hence, the model was designed with a one-year time frame in mind. However, this annual design wouldn't be suitable for places like Thailand or Myanmar where double-cropping within a year is common.

While I currently designed the model on an annual basis, I am considering modifying it to a monthly scale to incorporate satellite imagery data (which often provides monthly insights) and to better reflect crop selection scenarios.

To be very honest with you, the decision to use a one-year tick wasn't just due to the typical farming schedule in regions like North Korea. It was also because I lacked datasets with finer temporal resolution. However, in comparison to countries where double-cropping is prevalent, many countries operate their agriculture on an annual schedule. Therefore, it would be reasonable to align all the environmental and socioeconomic indicators based on a one-year cycle.

Given more time and access to data with higher temporal resolution, such as monthly datasets, I would also like to explore the seasonal intricacies of land cover changes. In North Korea, farming practices are influenced by factors like annual variations in precipitation and sunlight hours. The country's limited flat terrains have also led to farming in mountainous areas, primarily for corn cultivation, which has exacerbated soil erosion at an accelerated pace.

### Why did you choose the 'Cooperative household' as an agent?
I believed that creating a model that simulates a North Korean cooperative farm would be the right scale, especially since the "cooperative farm" is the fundamental agricultural unit in North Korea. During the 1960s, these cooperative farms were the primary unit of the agricultural landscape. Even current administrative districts indicate that the boundaries of these districts align well with those of the cooperative farms.

While there is a system dynamics method that represents energy and food flow in the model simulating all of North Korea, its scale is too broad to observe detailed dynamics. In the long run, I think it would be beneficial to establish system dynamics for the entire North Korean system. This could be integrated into a hybrid model where agent-based modeling (abm) related to the specific properties of each region functions as a "localised" model.

### It must have been difficult to verify the research because it was North Korea. How did you verify the modelled outcome?
The validation procedure for environmental variables was based on satellite images and data from previous studies. <span style="color:red">For example... </span>

As for the human (agent), extensive literature and data containing information about North Korean cooperative farms were utilised. For human behaviors, only minimal rules were applied, such as changing behavioral patterns when hungry, maximising land potential first, and resorting to inter-mountain cultivation when there's still a shortage. Regarding labor hours and land potential, due to the lack of data, South Korean data was used as a starting point. This was then scaled up or down based on iterative simulations.

---
## Can you tell your key findings from the simulation?

The results of the baseline scenario of **North Korean cooperative farm** model illustrated the negative cycle of land degradation and declining food production in North Korea, culminating in famine. The simulation, which used AD 1960 as the starting year, suggest that a famine could occur around 35 years later, roughly corresponding to 1995 in reality.

Implementing the **additional food supply** scenario effectively postponed the famine in North Korea. This underscores the importance of addressing North Korea's closed-off characteristics in mitigating famine threats. However, a holistic solution requires more than just making the food system more open; it necessitates interventions in other areas as well.

## What does this figure tell us?


In many cases, it manifests as a vicious cycle initiated by human land use: reduced land productivity due to degradation leads to an increased strain on food resources, prompting more intensive land use, and resulting in further land degradation—a classic positive feedback loop. Such positive feedback can destabilize systems, nudging them to the brink of chaos and triggering significant disruptive changes.

The model was fundamentally designed around this feedback loop: "Reduced land productivity (due to degradation) leads to increased food pressure (from reduced food production), resulting in intensified land use, and subsequently further reduced land productivity."

In our simulations, internal metrics showed a linear decline in soil quality (or land productivity), whereas the corresponding agricultural productivity exhibited an exponential decline. This trend underscores that their relationship isn't simply linear but emerges from a feedback mechanism.

---

## Do you have any other stories to share regarding your research?
My supervisor and I had tentatively agreed that my Ph.D. thesis would be an "extension of my Master's thesis". I used this approach to improve and refine my Master's thesis on the North Korean land degradation-famine ABM model (also known as the "Pyongando Model"). However, due to a lack of data, this proved to be a difficult task, and I was quite concerned. During this time, a visiting professor specialising in ecological modelling suggested, "Why not try simplifying or abstracting the model?" I sketched out the concepts of "creating a virtual North Korea" and "establishing a virtual collective farm" at the time. This occurred between June and July of 2018.

By July 2018, I had developed a preliminary model and sought feedback, so I applied for an oral presentation at Computational Social Science (CSS) 2018. I was not aware that I had to submit a full paper other than the abstract. The Computational Social Science Society of the Americas (CSSSA) Board pointed out this oversight, and although I didn't meet the submission requirements, they commented, "The topic is intriguing, so we'll give you an additional two weeks to write it." Within those two weeks, I completed the model, and wrote the paper. The review results were split, with two accepting and two rejecting, but it was ultimately accepted. I believe this was two-thirds of my thesis chapter was completed during these two weeks!

One memorable comment I received during the conference presentation was from a grad student from India: "The model's outcome, which suggests that North Korea's closed and inefficient collective farm system collapses around 35 years in, isn't only pertinent to North Korea. Similar patterns are seen in other countries. For example, countries colonized by Britain, for example India and Ghana, whenever the British interfered with their agricultural system (with authoritative and peculiar regulations), often experienced famine 30-40 years later. The famine in Ethiopia also occurred about this long after transitioning to a socialist regime. Why not evolve this model into a general famine model? Dr. Amartya Sen's theory, which holds that freedom leads to the resolution of inequalities and the promotion of development, could serve as a good theoretical foundation." Although the feedback was insightful, I considered it a long-term project.

Between January and February 2021, I conducted interviews with experts and North Korean defectors based on my model's results. While some feedback was hard to accept or exceeded the scope of my thesis, such as including North Korea's reclamation projects or the introduction of manure fertilizers for soil improvement, I was encouraged by a comment from a North Korean defector who is an agriculture specialist. He noted, "The results reflect the reality of North Korea very well. Although there are criticisms that the pre-collapse results of the model are unrealistic (how can land be abandoned in North Korea?), the agricultural lands in North Korea in the early 1990s were essentially lost in potential, which was strikingly similar to the situation within the model."