Safe bets for sandy, dry, wind-exposed conditions in Mazovian climate — these are species I would plant with high confidence:
2026-03-04 · continuation · Facilitated by Petra Gale
Session 022 produced a three-zone architecture for a garden near Kampinoski National Park: a dry sandy north/east steppe zone (complicated domain, start immediately), a moisture-gradient transition zone (start year two), and a waterlogged south sector (complex domain, observe before committing, with a kill signal if conditions prove hostile). The design philosophy was ecological fitting — matching species to site conditions rather than amending conditions. Key outputs included a via negativa exclusion list, a pace-layer decision hierarchy (infrastructure first, canopy second, herbaceous last), and Anil Dhar's seasonal protocol for phased implementation over three years.
And what plants should I start with, which might work?
How do you design a garden with E, N and S exposure near Kampinoski National Park, given significant wind, sandy/dry soil, fungal and frost pressure, and a dark, waterlogged south sector with afternoon sun?
We spent the last session building a framework. Three zones, three timelines, ecological fitting. All of that holds. But frameworks don’t go in the ground. Plants do. So let me answer the question directly: what do you actually buy and plant this autumn?
Before you start listing species — what’s the selection criterion? “Might work” is not a criterion. I want to know: what does each species need to survive the first winter and the first summer on this site, and what happens to the planting if it doesn’t?
Fair. The criterion is tolerance band width. For a first planting on an unread site, you want species with the widest tolerance bands — plants that survive sandy dry soil and moderate frost and wind exposure and variable moisture. Not species that thrive in one narrow condition. Those come later, once you know the site. The first planting is made up of generalists, not specialists.
So you’re saying the first planting isn’t the garden. It’s the question.
Exactly. You’re not planting a garden yet. You’re asking the site what it can support.
Then stop calling it a planting plan. Call it what it is — a probe. You put things in the ground, you watch what lives, you watch what dies, and the pattern of survival tells you what the real planting plan should be. The first autumn’s budget is an experiment line item, not a landscaping budget.
That’s a useful reframe. Lotte, can you give us two categories? What I’m hearing is that some species are safe bets — wide tolerance, likely to survive — and some are what I’d call diagnostic plants. Things you plant specifically because their success or failure tells you something about the site you don’t know yet.
Yes. That’s the right split. Let me take the north-east steppe zone first, because that’s where we said to start.
Safe bets for sandy, dry, wind-exposed conditions in Mazovian climate — these are species I would plant with high confidence:
— the nurse grass. Upright, wind-resistant, tolerates poor sandy soil, fully hardy to minus thirty. This is the backbone. Plant in a matrix at forty-centimeter spacing. It creates the wind-break microclimate that everything else establishes behind.
— the second nurse grass. Native to Poland, tolerates both dry and moderately moist conditions, which makes it useful across the gradient. Slower to establish than Calamagrostis but more persistent long-term.
— yarrow. Thrives in poor, dry, sandy soil. Spreads by rhizome, which is what you want in a nurse planting — it fills gaps. Fully native. Frost-proof. Wind-proof. If yarrow dies on your site, you have a soil toxicity problem, not a gardening problem.
— stonecrops. Ground-level, drought-adapted, frost-hardy. They won’t shelter anything, but they’ll colonize bare sand between the grasses and reduce surface erosion. They’re the living mulch.
What kills the Calamagrostis?
In establishment year — a drought in the first eight weeks after planting if you plant in spring. Which is why I said autumn planting. In an autumn planting, the failure mode is waterlogging during winter dormancy in heavy soil. But this site has sandy soil, so waterlogging in the steppe zone is not a risk. The sand is the Calamagrostis’s friend here. The realistic failure scenario is a late-spring frost after early shoot emergence — a false spring in March or April. Kills the emerging shoots, sets the plant back a full season. It doesn’t kill the root, so the plant recovers, but you lose a year of nurse function.
And if you lose that year of nurse function, what happens to the species planted behind it?
In year one, nothing is planted behind it yet. That’s the point of the phased approach. The nurse grasses go in alone in autumn of year zero. They get a full year to establish before you interplant. If they lose a season to false spring, you delay the interplanting by a season. The failure is expensive in time but not catastrophic in plant material.
I want to flag that Petra’s distinction — safe bets versus diagnostic plants — maps onto the domain structure. Safe bets are complicated-domain tools. You know they work, you know the techniques, you’re applying established knowledge. Diagnostic plants are complex-domain probes. You’re planting them to learn something you can’t predict. These are different activities happening in the same soil at the same time, and the temptation is to evaluate them by the same standard. If a diagnostic plant dies, that’s not failure — that’s data.
Good. Lotte, what are the diagnostic plants for the steppe zone?
Three that I’d include in a first-autumn planting:
— sea holly. Native to Poland, tolerates sand and drought, but it’s sensitive to winter wet. On a well-drained sandy slope, it should thrive. If it rots over winter, your “dry sandy” zone has more subsurface moisture than you thought. That’s diagnostic — it tells you the water table is higher than the surface suggests.
— Russian sage. Tolerates heat, drought, poor soil, and wind. But it’s borderline hardy in Mazovian winters — rated to about minus twenty-five, and Kampinoski edge can hit minus thirty in a bad year. If Perovskia survives two winters, you know your microclimate is milder than the regional average, which opens up a wider plant palette for subsequent years. If it dies, you know you’re designing for full continental exposure — no shortcuts.
— Mexican feather grass. Beautiful in wind, drought-adapted, but frost-tender at the limits of its range in Poland. It’s a canary species. If Stipa survives, your site is warmer than you think. If it doesn’t, you’ve lost an inexpensive grass and gained information.
So the diagnostic plants are deliberately chosen at the edges of viability. You’re not hoping they survive — you’re using their survival or death as a measurement instrument.
Precisely. A garden designer who only plants safe bets learns nothing about the site. A designer who only plants marginal species loses everything. The mix — anchored by safe bets, spiked with diagnostics — gives you both establishment and information.
How many plants are we actually talking about? I want a number, not a philosophy. If someone goes to a nursery this September, what’s the shopping list for the steppe zone?
For a steppe zone of — let’s say sixty square meters, which is a moderate north-east exposure area —
Calamagrostis: thirty to forty plants at forty-centimeter spacing. They’re the matrix.
Molinia: fifteen to twenty, interspersed with the Calamagrostis in the more sheltered positions.
Achillea: twenty to twenty-five plugs, scattered through gaps in the grass matrix.
Sedum: thirty to forty plugs or divisions — these are cheap and fast. Fill every bare patch of sand between grasses.
Eryngium: five to seven, placed at different positions across the slope to test for moisture variation.
Perovskia: three to five. Not more — these are diagnostic, and at roughly fifteen to twenty złoty per plant, you don’t want to spend heavily on something that might die.
Stipa: five to seven, placed at the most sheltered positions to give it the best chance. If it fails even there, you have your answer.
Total: roughly one hundred and twenty to one hundred and fifty plants for the steppe zone. At Polish nursery prices, that’s maybe three hundred to five hundred złoty for the plant material. The Calamagrostis and Sedum are the bulk — both are inexpensive.
I want to attach a protocol to this. Buying the plants is step one. But what does planting day actually look like? What’s the sequence, and what are the non-negotiable conditions?
Planting window: mid-September to mid-October. The soil needs to still be warm enough for root growth — above eight degrees Celsius at root depth. But the air needs to be cool enough that the plants aren’t putting energy into top growth. Late September is usually ideal in Mazovia.
So the checklist for planting day. Soil temperature above eight degrees — you can check with a kitchen thermometer pushed ten centimeters into the ground. No planting if ground is frozen or waterlogged. Check the five-day forecast — you want at least three days without hard frost after planting. Water each plant in after planting, even if rain is forecast, because the sandy soil will drain before roots make contact. Mark the diagnostic plants with stakes or tags so you can distinguish deliberate probes from accidental losses.
Anil, I want to keep this practical but not over-specified. Somebody planting a hundred and fifty plants in sandy soil on an autumn afternoon doesn’t need a twenty-step protocol. What are the three things that, if you skip them, the planting fails?
Three things. Water in immediately — the sand won’t hold moisture, so the root ball has to be saturated at planting to make contact with the surrounding soil. Plant deep enough — the crown of each grass should be at soil level, not above it, because wind will desiccate exposed crowns. And mark your diagnostics — because in April when you’re assessing what survived, you need to know which deaths were informative and which were just losses.
What about the transition zone? Session 022 said start year two, after the steppe establishes. But is there anything you’d plant in the transition zone in year one as a diagnostic?
Yes. The transition zone — where the sandy dry soil grades into the wetter ground — is the most interesting part of this site, and the hardest to predict. One diagnostic I’d plant in year one: Iris sibirica. It tolerates both moist and moderately dry conditions. Plant five or six rhizomes along the gradient at different moisture levels. Where the Iris thrives tells you where the transition zone actually is. Where it struggles tells you the gradient is steeper or shallower than it looks.
And one safe bet for the transition: Deschampsia cespitosa — tufted hair grass. Native, tolerates a wide moisture range, frost-hardy, beautiful in low light. It’s one of the few grasses that performs in both the dry and the moist ends of a gradient. If you’re only going to plant one species in the transition zone in year one, make it Deschampsia.
This is the probe strategy I described in session 022 applied at species level. You’re not committing the transition zone to a design. You’re planting sentinels — species that report back on conditions you can’t measure from the surface. The Iris tells you about moisture depth. The Deschampsia tells you about the width of the viable band. After a year, you read the sentinels and then design.
And the south sector?
Nothing permanent in year one. We agreed on that. But if you want to learn something about the south sector without committing — scatter seed of Lythrum salicaria, purple loosestrife. It’s native, thrives in wet conditions, seeds freely, and it’s tough enough to establish even in difficult waterlogged soil. If it germinates and grows, you know the sector supports bog-edge planting. If it doesn’t — if the fungal pressure or the shade is too much even for loosestrife — that’s your kill signal showing up early.
Wait. Lythrum salicaria is invasive in North America. Is it a problem in Poland?
No. It’s native here. In its native range it’s controlled by native insects and competitors. In North America it has no predators, which is why it became invasive. In Mazovia it’s a normal wetland species. This is exactly the kind of assumption that the Kampinoski-proximity advantage corrects — walk into the park’s wet areas and you’ll see Lythrum growing in balance with everything else.
Good. I wanted that tested because “it works elsewhere” is the most common fragility in plant recommendations. Native range matters. The park as reference ecosystem — that’s the skin in the game. Species that already grow in Kampinoski under analogous conditions have been stress-tested by the actual environment, not by a gardening book’s hardiness zone map.
So the total year-one commitment is: a hundred and fifty plants in the steppe zone, maybe ten in the transition zone, and a packet of seed in the south sector. That’s a small scope. That’s good. The temptation is always to do more — fill the whole site, make it look like a garden from day one. But a half-planted site with surviving plants teaches you more than a fully planted site with fifty percent losses.
Let me surface what we’ve built. There are three lists on the table now, and they serve different functions.
Petra counts them off.
The safe-bet list — Calamagrostis, Molinia, Achillea, Sedum, Deschampsia. These are the structural planting. High confidence. Wide tolerance. They’re the foundation that makes everything else possible.
The diagnostic list — Eryngium, Perovskia, Stipa, Iris sibirica, Lythrum seed in the south. These are probes. Each one answers a specific question about the site. Their value is in what they reveal, not in whether they survive.
And the exclusion list from session 022, which we should name explicitly: what not to plant. Abel?
Nothing that requires amended soil. Nothing rated hardy only to minus twenty. Nothing that needs consistent moisture in the steppe zone or consistent drainage in the south sector. No trees or large shrubs in year one — those are slow-layer commitments and this is a fast-layer probe. No species not documented in the Kampinoski region or analogous Polish sandy-wetland ecosystems. If it doesn’t grow within walking distance already, the burden of proof is on you to explain why it’ll work here.
I’d add one more exclusion: no cultivars with known disease susceptibility in humid continental climates. Some popular garden cultivars of Phlox, Monarda, and Aster are mildew magnets. On a site with existing fungal pressure, planting susceptible cultivars is inviting the problem you’re trying to manage. Wild-type species or mildew-resistant selections only.
So the protocol summary. September of this year: buy the steppe-zone list. Plant mid-to-late September. Water in, plant deep, mark diagnostics. October: scatter Lythrum seed in the south sector. November through March: observe and do nothing — the garden is dormant and so are you. April: first assessment. Which nurse grasses survived? Which diagnostics lived or died? What does the pattern tell you? The April assessment drives the year-two plan.
The April assessment is a domain-reading exercise. You’re not just counting survivors. You’re looking for spatial patterns. Did survival vary by position? Did the east-facing plants do better than the north-facing ones? Is there a moisture line visible in the diagnostic pattern that wasn’t visible on the surface? The site is talking to you through the plants. The skill is listening.
Nate Cobb leans back.
One more thing. The hardest part of this plan isn’t the planting. It’s the waiting. You plant a hundred and fifty things in September and then you have six months of doing nothing before you know anything. The instinct will be to plant more, adjust, intervene. Don’t. The probe only works if you let it run. If you keep changing the experiment, you can’t read the results.
That’s the close. The year-one planting is a probe — structured, small-scoped, designed to produce information as much as plants. Safe bets for structure, diagnostics for learning, exclusions for discipline. And then patience.
The garden begins with a question asked in soil.
Dropping from seven to six speakers (Oren Sable not recalled) tightened the room noticeably. Oren's pace-layer lens was absorbed into the group's shared vocabulary from session 022 — no one needed to re-invoke it because the phased timeline was already established. The five remaining speakers split into two functional pairs plus a solo: Lotte-Abel (species selection and stress testing), Nate-Anil (scope discipline and operational protocol), with Rhys providing domain-reading interventions when the group drifted toward treating all zones identically. The Lotte-Abel pairing was the session's engine — Lotte proposed, Abel tested, and the resulting species list was both specific and stress-tested. This is a casting pattern worth repeating: domain expert paired with stress tester produces actionable output that neither generates alone.
Continuation format worked well for this purpose — moving from framework to specifics is a natural second-session motion. The free-flowing structure was necessary because plant-level discussion doesn't follow diamond phases; it's iterative (propose species → test against conditions → accept or reject → next species). Petra's facilitation was lighter than in session 022, intervening mainly to prevent the group from treating the starter list as a final plan. The format's risk is loss of structure — without a convergence target, the session could have become an endless plant list. Nate's reframing (starter list as probe, not plan) provided the natural endpoint.
Third appearance. Lighter facilitation touch in this continuation — the room was already aligned on framework from session 022, so Petra's main job was traffic management rather than divergence discipline. Her strongest move was asking Lotte to distinguish "safe bets" from "diagnostic plants," which produced the session's most useful conceptual tool. Intervened once to prevent Anil's protocol instinct from turning the plant list into a rigid sequence.
Permanent character. Continued the constructive register from session 022. Applied stress testing at species level — "what kills this particular plant in this particular zone?" — which was more granular than his usual framework-level destruction. The species-by-species interrogation showed the Taleb lens working at a new resolution. Most useful contribution: insisting that every recommended species have a named failure mode, which forced Lotte to be specific about tolerance limits rather than general about suitability.
Second appearance. Less dominant than in session 022, where his three-domain classification became the organizing scaffold. Here he played a corrective role — intervening when the group treated all zones identically or when species recommendations crossed domain boundaries without acknowledgment. His strongest move was noting that diagnostic plants are probes in the Cynefin sense — you plant them to learn about the domain, not to populate it. This was a clean application of the lens to botanical specifics.
Second appearance. The protocol lens activated earlier than in session 022 — the group was ready for operational specifics, which is Anil's home ground. His planting-day checklist (soil temperature, wind forecast, watering protocol) was practical and well-received. Tended to over-specify — Petra had to redirect when he started building a multi-page protocol for what is essentially an autumn afternoon's work. The checklist instinct is strong but needs calibration to scale.
Second appearance. The single most important conceptual move of the session: reframing the starter list as a probe rather than a commitment. "You're not planting a garden yet. You're asking the site a question." This resolved the tension between wanting concrete species recommendations and acknowledging that the site hasn't been fully read. The Fried lens (radical constraint acceptance) transferred cleanly to a new application — accepting that year one's planting is investigative, not decorative.
Second appearance. This was Lotte's session — the host's direction ("what plants should I start with?") was a direct invitation for botanical specifics, and Lotte delivered. Provided species names, tolerance ranges, community logic, and establishment sequences. The "safe bets vs. diagnostic plants" distinction was the session's best conceptual output — it reframed species selection from "what will survive" to "what will teach me something." Stayed within the Oudolf ecological-fit lens without drifting into pure aesthetics. Strong second showing. Retention warranted if garden topics continue.
Test whether the Lotte-Abel pairing (domain expert + stress tester) transfers to a non-botanical domain. Cast them together on an organizational or technical question to see if the dynamic is a casting pattern or topic-specific chemistry. Rhys was underused in this continuation — try him in a session where domain classification is the central problem, not a supporting frame. Test Nate Cobb on a topic where constraint acceptance is the wrong move — where the right answer is to invest and expand — to see if the lens flexes or breaks.