March 2022. My advisor assigned me a modeling competition.
The labmate who had helped test NetLogo models during summer school sent me a message.
He said he wanted to improve.
I stared at the screen for a long time. The last time I heard something like that was during undergrad thesis defenses. A classmate had said he chose his topic because he "wanted to challenge himself." The panelist shot back: so, how's that going?
My labmate, this time, meant it.
Our advisor's recent projects were all about carbon neutrality. System dynamics, Anylogic, plant population life history, carbon sequestration capacity estimation. He could rattle off these terms fluently. When it came to actual modeling and simulation, though β that was a different story.
I get it. I really get it. He only knows what he knows, and beyond that, he plays deaf and dumb.
The project was a plant carbon sequestration potential assessment model based on population life-history processes. System Dynamics is essentially a modeling method that describes feedback loops through differential equations. It treats a plant population as a series of state variables β seed bank, seedling bank, juvenile tree bank and mature tree bank β evolving over time. An elm tree goes through total 5 stages from seed to over mature tree. Carbon fixation efficiency varies dramatically across these stages. The model's job was to track carbon flow along this chain.
We used Anylogic for simulation, examining how different precipitation scenarios affect carbon sequestration capacity.
The data was ready-made. Native elm woodlands in the Hunshandake and Horqin sandy lands, with regional precipitation data in Shapefile format. Elm woodlands are the climax community of sandy land ecosystems β deep roots, long lifespans, high carbon storage. But elm seedling-to-mature-tree survival rates vary enormously. Seed germination requires ample soil moisture, seedlings have weak drought resistance, and only mature trees with deep roots can tap into groundwater. This life-history strategy means carbon contribution differs drastically across stages. The model was also ready-made β our advisor had worked on similar projects before, so we had a foundation.
The problem was the code.
Anylogic is simulation software with its own modeling language and environment. Honestly, it's a lot more complicated than NetLogo. NetLogo at least has readable documentation. When you hit a problem in Anylogic, you're basically relying on arcane Google-fu.
My labmate said he'd write the code. I thought he was joking. But he did. Almost all of it.
I only handled the technical problems he couldn't solve. Where he got stuck, I supplemented with my English search skills β trawling through Stack Overflow and the Anylogic Community Forum, looking for anyone who had faced something similar.
Sometimes we found something, and the solution worked with minor tweaks. Sometimes we didn't, and we had to figure it out ourselves.
The most memorable issue was parameter calibration for the plant growth curve. The model required carbon storage conversion efficiency values for different growth stages. The literature data was either in the wrong units or came from tiny sample sizes. Plug it in, and the curve would come out absurdly far from reality.
My labmate tried for three days. Couldn't tune it.
I went searching too, and also spent half a day on it. We sat together, reviewed the code, debugged together.
Eventually, it ran.
The competition results were unexpected. We went in with low expectations, but our model was shortlisted. I was forced to wake up early one summer morning β a morning I could have otherwise slept in β for an online presentation.
My labmate did the presenting. He did better than I expected. No stage fright, clear logic, emphasizing what mattered and glossing over what didn't.
The judges asked a few questions. One about the model's spatial scale effects β the judge pointed out we hadn't accounted for competition between individual elm trees.
I wanted to push back, but I couldn't find the right argument on the spot.
My labmate jumped in. His answer didn't quite land, but it became his research direction.
I watched from the side, thinking: when did this guy get so articulate?
He really did want to improve. And he was improving fast.
The judges didn't press further. By the time the presentation ended, the online meeting had dwindled to about a dozen people.
Afterward, I asked him who came up with that spatial scale response.
He said he did.
Nice.
That's life sometimes. You think some people will always be a certain way, and then suddenly they change. The change catches you off guard.
Maybe it was the right opportunity. Maybe someone lit a fire under him. Maybe he just hit the age where you're supposed to step up.
Either way, this carbon neutrality project unexpectedly became a turning point for him.
And for me.
I'll probably remember for a long time that our advisor was a hack. But working under a hack isn't entirely without benefit β at least the freedom is ample. Ample enough for my labmate and me to muddle through things he didn't understand at all, then stand before him and pretend it was all based on his "wise guidance."
The adult world is that cynical and practical.
My labmate reached out a few more times after that, asking technical questions. Some I could answer. Some I couldn't. For the ones I couldn't, I'd say: let me look into it.
And then I'd go search.
In English.
On countless late nights. Quietly grinding.