Rust-Risk Calendar — Conditions Conducive to Infection
Cell = mean nightly infection-risk class within the dekad, from a temperature × leaf-wetness-duration matrix. This measures weather conducive to infection — not confirmed disease. Actual outbreaks also depend on host susceptibility (species, variety) and crop load — surfaced as realized-risk modifiers below — plus pathogen presence. Absolute scale (not deviation) so chronically-humid origins read at their true level and all origins are cross-comparable.
Robusta vs. arabica — read these origins differently. Leaf rust (Hemileia vastatrix) overwhelmingly attacks arabica. Robusta (Coffea canephora) — Vietnam, Indonesia, and Brazil's Espírito Santo conilon — carries strong genetic resistance: not immune, but highly conducive weather that would be very dangerous in arabica typically passes with limited damage. Treat High/Extreme readings at robusta origins as a weather signal, not an outbreak warning. The arabica origins — Colombia, Central America, Ethiopia, and Brazil Sul de Minas / Cerrado — are where conducive conditions translate into real crop and price risk.
Rust-Risk Class · HEADLINE (0 Low → 3 Extreme)
Nightly class from the infection matrix below, averaged over the dekad's nights. Hover for value; dotted line marks the start of the forecast window.
Decomposed Inputs — What's Driving the Risk
Latest Dekad by Origin
Ground-Truth — Observed Rust, Sul de Minas · FIELD VALIDATION
Observed field incidence — the outcome this calendar aims to lead. From Fundação Procafé’s monthly bulletins (2010–2026, three Sul de Minas stations). Latest,
September 2026: 0.0% of leaves infected — rank
1 of 17 Septembers on record (z -0.6; 31% of prior Septembers ran higher) →
the annual reset floor (11 of 16 prior Septembers also read 0.0%) vs the 1.4% September norm. Versus August the reading
fell 22.2 pt — but that compares last season’s harvested plots with this season’s new ones (below), so it
says nothing about disease.
’10
’11
’12
’13
’14
’15
’16
’17
’18
’19
’20
’21
’22
’23
’24
’25
’26
CANOPY GATE · NEW PLOTS — NOT COMPARABLE Leafing 97.9% — rank 2 of 13 Septembers (z -2.0) vs the 99.5% norm. September opens a new monitoring cycle: Procafé selects new plots and restarts the leaf and node counts, so a near-full canopy is expected and the z-score above is a ceiling effect, not thinning. Nothing in this reading can be compared with last month’s. The first month that says anything about the coming season is the next one.
Observed incidence lags the conditions that cause it by ~3–6 weeks (Hemileia’s latent period), so it reads roughly last month’s weather — the realized result the conduciveness calendar above forecasts. Stage A validation: our ERA5 weather for the belt tracks Procafé’s Varginha gauge across 197 months (temperature r = 0.98, bias −0.6°C; rainfall well-correlated but noisy on wet-month totals — a rain-driven wetness input can under-call dew/fog-humid pockets). The temperature axis of the engine rests on solid ground.
Host, Load & Husbandry Context — Realized-Risk Modifiers
The calendar above is weather conduciveness. Whether conducive weather becomes disease depends on host susceptibility (species & variety), crop load, and husbandry — independent modifiers shown here so equal conduciveness can be read as different realized risk across origins. Bearing year is derived from CONAB arabica production history.
- Brazil S. Minas · Cerrado susceptibility: High — susceptible arabica (Mundo Novo / Catuaí dominant) · 2026 is an ‘on’ year (high-load ↑ 48.2M bags, +35% vs prior) → realized risk amplified
- Brazil Espírito Santo susceptibility: Low — conilon (C. canephora) — strong resistance; only weakly biennial
- Central America susceptibility: High (eroding resistance) — post-2012 renovation to resistant catimores (Lempira, Marsellesa) is partial and resistance has broken in places (Lempira, Honduras 2017)
- Colombia susceptibility: Low–Moderate — majority rust-resistant Castillo / Cenicafé varieties since the post-2008 renovation → conduciveness overstates realized risk
- Ethiopia susceptibility: Variable — heirloom landraces with diverse, mixed resistance; mostly low-input garden / forest systems
- Indonesia (Lampung) susceptibility: Low — robusta (C. canephora) — strong resistance
- Vietnam (C. Highlands) susceptibility: Low — robusta (C. canephora) — strong resistance
Input-affordability pressure (husbandry — leads by ~1 season). Nutrition and fungicide a grower could afford a season ago set this crop's resilience; today's affordability leads the next crop's vulnerability. Affordability = coffee price ÷ fertilizer basket (urea + potash), percentile vs history. FX-adjusted for Brazil only (v1); other origins read the global USD-basis complex.
Input-Affordability — 36-Month Trajectory
Monthly affordability percentile vs full history (higher = inputs more affordable → husbandry supported). A sustained fall = rising input-affordability pressure, building vulnerability into the next crop (~1-season lead). Dotted line = median.
- Arabica complex KC ÷ fertilizer husbandry window (lagged 6mo, 2026-03): 67 pctile affordability → well-resourced crop — current 91 pctile, stable (-5% vs 6–12 mo prior); watch the trend as a forward vulnerability signal.
- Robusta complex RC ÷ fertilizer husbandry window (lagged 6mo, 2026-03): 60 pctile affordability → mid-range resourcing — current 83 pctile, stable (-7% vs 6–12 mo prior); watch the trend as a forward vulnerability signal.
- Brazil FX overlay BRL/USD 5.05 vs 5.38 a year ago — a firmer BRL is dampening the FX tailwind Brazilian growers had.
Infection-risk matrix (temperature × leaf-wetness duration)
Engine: for each origin, hourly 2 m temperature & dew point are sampled at 3 representative points. Per local night (18:00–06:00) the longest continuous wet spell (dew-point depression < 2°C) gives leaf-wetness duration (LWD); its mean temperature and LWD are looked up in the rust infection matrix → a nightly class (Low/Medium/High/Extreme). Dekad cell = mean class over the dekad's nights; the table also reports the share of nights reaching High+. Cardinal temperatures and the ~6 h wetness threshold follow de Jong 1987 / Bebber 2016. Data: Open-Meteo (recent actuals + 14-day forecast). Generated 2026-10-26.
Calibration: the matrix reflects the two-stage thermal biology of infection — urediospores germinate near the warm evening peak (optimum ≈21°C) but appressorium formation and stomatal entry proceed optimally as the canopy cools overnight toward ~11–16°C, so cool highland nights are conducive (not protective) once a wet spell has started. Risk floors at the ~13°C germination minimum and is suppressed above 30°C; ~6 h continuous wetness ≈ one infection opportunity. Thresholds follow de Jong 1987 / Bebber 2016 and will be refined against operational warning systems (e.g. Pozza et al. 2021). Read relative differences across origins and dekads as the primary signal.
Realized-risk modifiers: host susceptibility (species & variety), Brazil's bearing year, and input-affordability pressure (the husbandry lever, leading by ~1 season) are surfaced in the Host, Load & Husbandry Context section above — robusta origins and resistant-variety arabica (e.g. Colombia's Castillo) carry conducive weather with less realized disease, a high-load arabica "on" year amplifies it, and affordability a season ago sets how well-defended the current crop is. Field validation (new): for the flagship arabica belt, the Ground-Truth — Observed Rust, Sul de Minas panel above now grounds the calendar in 16 years of Fundação Procafé field observations — the realized incidence (lagged ~3–6 weeks) the calendar aims to lead, plus a Stage-A check that our ERA5 weather matches Procafé's Varginha gauge (temperature r = 0.98 over 196 months). Still deferred: extending the same observed-incidence ground truth to the other origins (remote-sensing / field reports), which would let these modifiers be combined rather than read side-by-side.
References: Bebber et al. (2016) Phil. Trans. R. Soc. B; Jong et al. (1987) Neth. J. Plant Pathol.; Yan & Hunt (1999) Ann. Bot.; Avelino et al. (2015) Food Security; Pozza et al. (2021) Agronomy.