Saketh's Blog

If you’ve spent any time in West Oakland, you’ve probably walked past a community garden, a backyard vegetable patch, or a vacant lot someone’s eyeing for a future one. What you probably haven’t thought about is what’s underneath: decades of lead from car exhaust, deteriorating paint, and industrial activity, quietly sitting in the topsoil, waiting to be disturbed.

Lead doesn’t break down. It doesn’t wash away. Once it’s in the soil, it stays there until something physically removes it, and for kids especially, that’s a real problem. Children absorb roughly ten times the share of ingested lead that adults do, and even low-level exposure is linked to reduced IQ and shortened attention spans. So when I found out that some plants can actually pull lead out of the ground and concentrate it in their leaves and stems, I wanted to know: could that actually work as a cleanup strategy, and if so, how well, and how fast?

The problem with “phytoremediation is promising”

Search around and you’ll find plenty of studies on lead-absorbing plants, usually framed with some version of “promising results.” Grow Brassica juncea (Indian mustard) in a pot for three weeks, and you might see an 80% reduction in soil lead. Sounds great. But almost none of that research asks the next question: what happens if you plant this in a real yard, not a lab pot, for years instead of weeks?

That gap is what I’m building PhytoRemedy to close. It’s a tool that takes a real site — a specific location, a specific soil lead level — and tells you which of a handful of candidate plants would actually do the best job, and realistically how many years that would take. Right now it’s comparing four: amaranth, Indian mustard, vetiver grass, and sunflower.

Almost everything the tool knows about lead specifically traces back to one 2013 experiment that grew three of those four plants side by side in contaminated soil for 56 days and measured exactly how much lead ended up in their shoots versus their roots. From that, the tool calculates a bioconcentration factor for each plant (basically, “how much does this plant concentrate lead relative to what’s in the soil around it”) and uses that to model years of repeated planting and harvesting using an exponential decay curve, the same math you’d use for radioactive half-life, just applied to lead in soil instead.

What It’s Showing

I’ve been running the tool on a real West Oakland scenario using soil-lead data from a 2012 academic survey of the neighborhood. The early numbers are sobering. At the neighborhood median lead level, an unamended planting of sunflower comes out at roughly 113 years to bring the soil down to California’s residential safety standard. Even in the best case the tool will model, fertilized, irrigated, harvested as often as possible, that only drops to about 48 years.

The result I didn’t expect is how much harvest frequency matters relative to how much lead a plant actually holds. A plant with a lower concentration in its tissue can close most of the gap on a stronger one simply by completing more cycles in a year. That’s still moving as I refine the growth assumptions, and it’s the part of the model I’m testing hardest right now, because it’s the difference between ranking plants on one number and ranking them on how they actually behave in a season.

And at the very worst yards in the city, over 3,000 mg/kg of lead, more than 80 times the concentration the science was even tested at, the model refuses to give a confident number. That’s on purpose. A number with no real backing is more dangerous than an honest “we don’t know.”

Where it stands

Phytoremediation isn’t a quick fix. At the concentrations found in parts of West Oakland, it’s closer to a decades-long land management strategy than a cleanup you’d notice in your lifetime. But for the much larger number of moderately contaminated yards and lots across the city, a slow, low-cost, plant-based approach running quietly in the background might genuinely be the most realistic option available, especially compared to the alternative of expensive excavation that most homeowners and community gardens simply can’t afford.

The tool isn’t finished. The species list is still shifting, the cost side is rough, and some of the growth numbers are waiting on sources I haven’t run down yet. But the shape of the answer is already clear enough to be worth saying out loud: this doesn’t hand you a magic answer, it hands you an honest one, with every assumption labeled and every gap left visible instead of papered over. In a field where most of the research is 56 days long and most of the real-world questions take decades, that’s the part I most want to get right.