Skip to content

Health-zone estimates ​

This page splits each patch of the joint fit across its health zones, giving each zone's reproduction number, share of the patch and one-week forecast. The health-zone model on the methods page gives the maths, and the interactive map is on the dashboard. The one-week zone forecast is on the health-zone forecasts page and its scores in the health-zone forecast evaluation.

Load packages, data and fitted chains
julia
# Shared setup: packages, observations and the fit registry. See
# `docs/pages/_setup.jl`.
using BVDOutbreakSize
include(joinpath(pkgdir(BVDOutbreakSize), "docs", "pages", "_setup.jl"))
validation_forecast_from (generic function with 1 method)
julia
# The fits this page reads, loaded from the cache here: the headline joint,
# the zone fit melded from it, and the frozen zone fit the sampler
# diagnostics read.
chn_joint = load_fit("joint");
chn_local = load_fit("local");
frozen_local = load_fit("local_frozen_validation");

# The zone stage's fixed inputs with the joint's forecast, from the shared
# setup, so the health-zone forecast page draws the same forecast from the
# same inputs.
zone_inputs = zone_stage_inputs(; forecast = true);
zone_patch = zone_inputs.patch_of_zone;

Last updated: 6 November 2026.

Data as of: 1 November 2026.

Estimates by zone ​

The interactive map on the dashboard adds the probability of at least cases at a chosen and a filter by recent cases.

Health-zone post-processing
julia
# The geojson keys a zone without the province prefix the manifest carries.
zone_map_keys = [
    String(last(split(k, "."; limit = 2)))
        for k in zone_inputs.zone_keys
];
# Daily zone reproduction numbers over the zone grid, each zone draw on
# its own draw of the patch trajectory so the patch uncertainty is
# carried, and the cut-off values the map and table read. A zone's
# reproduction number is reported from the day its cumulative infections
# reach the floor in the median draw.
zone_rt_traj = reconstruct_zone_rt(chn_local, zone_inputs);
zone_RT_finite = [filter(isfinite, m[:, obs.n]) for m in zone_rt_traj];
zone_RT_reported = findall(!isempty, zone_RT_finite);
zone_share_T = let vs = vec(collect(chn_local[:share_T_zone]))
    [Float64[v[z] for v in vs] for z in eachindex(zone_map_keys)]
end;
_zq(v, p) = quantile(v, p)
# The one-week zone forecast drawn from the zone model, and the
# probability of at least K cases per zone from the same draws.
zone_fc = zone_forecast(chn_local, zone_inputs);
zone_fc_draws = zone_forecast_draws(zone_fc, zone_inputs);
ZONE_THRESHOLDS = (1, 5, 10, 20)
zone_fc_probs = zone_forecast_probabilities(
    zone_fc, zone_inputs; thresholds = ZONE_THRESHOLDS, draws = zone_fc_draws
);
zone_overview = zone_overview_table(chn_local, zone_inputs);
# Cases allocated to each zone over the past one, two and four weeks,
# and the last vintage on which each zone's count rose.
zone_recent = Dict(
    w => zone_recent_cases(zone_inputs; window = w) for w in (7, 14, 28)
);
zone_last_case = zone_last_case_dates(zone_inputs);

zone_map_fig = plot_zone_map_panels(
    [
        (;
            values = [median(zone_RT_finite[z]) for z in zone_RT_reported],
            zones = zone_map_keys[zone_RT_reported],
            lower = [_zq(zone_RT_finite[z], 0.05) for z in zone_RT_reported],
            upper = [_zq(zone_RT_finite[z], 0.95) for z in zone_RT_reported],
            diverging_at = 1.0, scale = log10,
            title = "Reproduction number at the cut-off",
            colorbar_label = "R",
        ),
        (;
            values = [_zq(v, 0.05) for v in zone_fc_draws.zones],
            zones = zone_map_keys, scale = CairoMakie.Makie.pseudolog10,
            title = "Confirmed cases over the coming week (lower 90%)",
            colorbar_label = "cases",
        ),
        (;
            values = [_zq(v, 0.95) for v in zone_fc_draws.zones],
            zones = zone_map_keys, scale = CairoMakie.Makie.pseudolog10,
            title = "Confirmed cases over the coming week (upper 90%)",
            colorbar_label = "cases",
        ),
        (;
            values = Float64.(zone_inputs.cumulative), zones = zone_map_keys,
            scale = CairoMakie.Makie.pseudolog10,
            title = "Confirmed cases to date", colorbar_label = "cases",
        ),
    ];
    ncols = 2, title = "Health zones at the cut-off"
);

The reproduction number at the cut-off, the bounds of the 90% interval on next week's confirmed cases, and the confirmed cases to date, by health zone. A zone whose reproduction number interval straddles one is washed towards white, and zones with no confirmed case or too few infections for a reproduction number are grey.
Compute the weekly cases and reproduction number by zone
julia
zone_cases_rt = cases_rt_table(
    zone_rt_traj, zone_inputs.days, zone_inputs.counts;
    cutoff = obs.cutoff, n = obs.n, areas = zone_inputs.zone_labels,
    patch = zone_patch
);
zone_cases_rt_fig = plot_cases_rt(
    zone_cases_rt; patch_labels = zone_inputs.patch_labels, top = 12,
    unit = "health zone"
);

The twelve zones with the most confirmed cases over the past two weeks, each against its reproduction number, with the week before. Faint bars span the 90% credible interval of the reproduction number.
Zone overview table
julia
# The overview as displayed: the interval strings, without the numeric
# columns.
zone_overview_display = let d = zone_overview[
        :,
        [:zone, :patch, :cases, :share, :R_T, :p_R_above_1, :delta_T],
    ]
    d[!, "R modelled separately"] = [
        w ? "yes" : "no" for w in zone_overview.walking
    ]
    d
end;
zonepatchcasesshareR_Tp_R_above_1delta_TR modelled separately
MutwangaNord-Kivu667.2 (2.5–20.0)1.26 (0.7–2.06)0.760.37 (-0.16–0.92)yes
BeniNord-Kivu72050.8 (33.2–66.8)1.08 (0.79–1.38)0.670.17 (-0.33–0.7)yes
OichaNord-Kivu938.3 (3.6–19.6)1.12 (0.68–1.76)0.640.21 (-0.43–0.82)yes
IsiroHaut-Uele11517.1 (6.2–44.0)1.07 (0.54–1.81)0.570.14 (-0.4–0.75)yes
KalungutaNord-Kivu994.6 (2.3–8.8)1.04 (0.68–1.51)0.56-0.04 (-0.5–0.38)yes
KyondoNord-Kivu460.8 (0.3–2.1)1.05 (0.64–1.69)0.56-0.12 (-0.65–0.39)yes
MambasaIturi10516.1 (3.5–53.3)1.08 (0.39–2.09)0.550.57 (-0.55–1.65)yes
Boma MangbetuHaut-Uele466.9 (2.9–13.8)1.01 (0.63–1.47)0.51-0.12 (-0.59–0.25)yes
ButemboNord-Kivu3444.9 (2.2–10.2)1.0 (0.59–1.53)0.49-0.11 (-0.52–0.31)yes
KatwaNord-Kivu6787.4 (3.3–14.8)0.98 (0.58–1.54)0.48-0.15 (-0.54–0.25)yes
WambaHaut-Uele17742.3 (19.7–66.3)0.98 (0.58–1.41)0.470.03 (-0.43–0.48)yes
MusieneneNord-Kivu1621.8 (0.7–5.2)0.95 (0.51–1.69)0.45-0.06 (-0.62–0.47)yes
MaserekaNord-Kivu430.6 (0.2–2.0)0.95 (0.54–1.63)0.44-0.11 (-0.64–0.41)yes
VuhoviNord-Kivu561.1 (0.5–2.6)0.96 (0.59–1.47)0.44-0.11 (-0.55–0.29)yes
TchomiaIturi860.6 (0.2–2.2)0.89 (0.47–1.68)0.390.05 (-0.86–0.95)yes
LuberoNord-Kivu430.9 (0.2–3.4)0.87 (0.41–1.8)0.39-0.01 (-0.62–0.61)yes
MabalakoNord-Kivu1164.2 (1.7–10.6)0.9 (0.5–1.54)0.39-0.01 (-0.57–0.55)yes
PawaHaut-Uele11626.5 (12.6–48.1)0.94 (0.55–1.4)0.39-0.03 (-0.44–0.38)yes
KiloIturi400.3 (0.1–1.2)0.91 (0.51–1.6)0.370.02 (-0.83–0.84)yes
LitaIturi3314.5 (1.4–13.4)0.88 (0.38–1.62)0.370.25 (-0.54–1.06)yes
LolwaIturi370.4 (0.1–1.5)0.87 (0.39–1.63)0.35-0.15 (-1.22–0.85)yes
DamasIturi410.2 (0.0–0.7)0.85 (0.43–1.66)0.34-0.18 (-1.2–0.76)yes
BuniaIturi201225.2 (9.2–47.2)0.87 (0.41–1.34)0.330.28 (-0.42–1.02)yes
NyankundeIturi1391.0 (0.4–3.1)0.85 (0.44–1.54)0.32-0.08 (-1.06–0.81)yes
FatakiIturi840.2 (0.0–0.7)0.8 (0.4–1.72)0.3-0.12 (-1.08–0.81)yes
NiziIturi8554.9 (1.6–13.5)0.82 (0.37–1.53)0.30.21 (-0.5–0.91)yes
BambuIturi2201.0 (0.3–2.8)0.84 (0.44–1.5)0.29-0.06 (-0.79–0.67)yes
RwamparaIturi12048.5 (3.1–18.3)0.84 (0.39–1.4)0.290.05 (-0.74–0.78)yes
MangalaIturi3982.2 (0.8–7.5)0.77 (0.37–1.59)0.270.03 (-0.75–0.83)yes
MongbwaluIturi7220.9 (0.2–3.8)0.65 (0.27–1.55)0.19-0.19 (-1.27–0.78)yes
Nia-NiaIturi3431.6 (0.4–8.1)0.6 (0.19–1.68)0.19-0.14 (-1.44–1.06)yes
KomandaIturi1931.0 (0.3–3.2)0.63 (0.26–1.28)0.15-0.44 (-1.49–0.57)yes
MandimaIturi49114.3 (3.8–41.9)0.54 (0.18–1.19)0.12-0.02 (-1.15–1.06)yes
MangoboTshopo1017.3 (11.3–25.6)1.07 (0.84–1.34)0.690.0 (-0.0–0.0)no
Makiso-KisanganiTshopo3744.7 (35.7–54.1)1.06 (0.84–1.3)0.660.0 (-0.0–0.0)no
KabondoTshopo2016.9 (10.9–24.9)1.05 (0.84–1.3)0.65-0.0 (-0.0–0.0)no
BafwasendeTshopo1811.8 (6.9–18.3)1.03 (0.83–1.25)0.58-0.0 (-0.0–0.0)no
BienaNord-Kivu260.7 (0.4–1.2)1.0 (0.7–1.33)0.5-0.0 (-0.0–0.0)no
AlimbongoNord-Kivu60.2 (0.1–0.4)0.99 (0.68–1.34)0.48-0.0 (-0.0–0.0)no
ManguredjipaNord-Kivu160.4 (0.2–0.8)0.98 (0.69–1.3)0.46-0.0 (-0.0–0.0)no
RunguHaut-Uele91.8 (0.8–3.3)0.98 (0.69–1.29)0.46-0.0 (-0.01–0.0)no
AriwaraIturi80.1 (0.0–0.2)0.85 (0.57–1.17)0.21-0.0 (-0.0–0.0)no
AruIturi250.4 (0.2–0.8)0.86 (0.55–1.15)0.21-0.0 (-0.01–0.0)no
AungbaIturi120.1 (0.0–0.2)0.84 (0.54–1.15)0.2-0.0 (-0.0–0.0)no
LogoIturi140.2 (0.1–0.5)0.84 (0.55–1.14)0.19-0.0 (-0.0–0.0)no
MahagiIturi50.1 (0.0–0.2)0.85 (0.56–1.14)0.19-0.0 (-0.0–0.0)no
RimbaIturi250.3 (0.1–0.5)0.85 (0.56–1.14)0.18-0.0 (-0.0–0.0)no
DrodroIturi150.2 (0.1–0.3)0.82 (0.55–1.12)0.160.0 (-0.0–0.0)no
GetyIturi110.2 (0.1–0.5)0.83 (0.56–1.1)0.160.0 (-0.0–0.0)no
KambalaIturi20.0 (0.0–0.1)0.82 (0.53–1.11)0.16-0.0 (-0.0–0.0)no
BogaIturi20.1 (0.0–0.3)0.81 (0.53–1.08)0.130.0 (-0.0–0.0)no
AdjaIturi110.1 (0.0–0.2)NaN-0.0 (-0.0–0.0)no
GomaNord-Kivu10.0 (0.0–0.1)NaN-0.0 (-0.0–0.0)no
KaynaNord-Kivu30.1 (0.0–0.2)NaN-0.0 (-0.0–0.0)no
DunguHaut-Uele10.2 (0.1–0.6)NaN-0.0 (-0.01–0.0)no
GombariHaut-Uele30.3 (0.1–0.8)NaN-0.0 (-0.0–0.0)no
Miti-MurhesaSud-Kivu3100.0 (100.0–100.0)NaN0.0 (0.0–0.0)no
LubungaTshopo11.9 (0.6–4.9)NaN0.0 (-0.0–0.0)no
TshopoTshopo53.7 (1.6–7.3)NaN-0.0 (-0.0–0.0)no
Wanie-RukulaTshopo11.6 (0.5–4.2)NaN-0.0 (-0.0–0.0)no
ButaBas-Uele116.3 (5.6–32.8)NaN0.0 (-0.0–0.0)no
GangaBas-Uele542.2 (26.3–58.2)NaN0.0 (-0.0–0.0)no
ViadanaBas-Uele640.2 (22.9–59.0)NaN0.0 (-0.0–0.0)no
BuluSud-Ubangi2100.0 (100.0–100.0)NaN0.0 (0.0–0.0)no
WapindaNord-Ubangi352.6 (31.9–75.3)NaN0.0 (-0.0–0.0)no
YakomaNord-Ubangi347.4 (24.7–68.1)NaN-0.0 (-0.0–0.0)no
Every health zone, the zones whose reproduction number is modelled separately first, each group by decreasing probability that it exceeds one. The share is of the patch's infections at the cut-off, in percent, and the share, reproduction number and deviation are medians with 90% intervals. A zone whose reproduction number is not modelled separately takes it from its province.

Point prevalence ​

The zone estimates split the infections in the community at the cut-off in the same way as the province estimates. The point prevalence section of the Methods gives the calculation. The figure shows the twenty zones with the highest rate in the community per 100,000 residents. The table lists every zone in the same order.

Compute point prevalence by zone
julia
zone_population = Dict(
    string(z.province, ".", z.zone) => z.population for z in load_health_zones()
);
zone_prevalence_overview = prevalence_overview(
    zone_prevalence(chn_local, zone_inputs, chn_joint),
    [zone_population[k] for k in zone_inputs.zone_keys];
    labels = zone_inputs.zone_labels, patch = zone_patch
);
zone_prevalence_fig = plot_prevalence_ranking(
    zone_prevalence_overview;
    patch_labels = zone_inputs.patch_labels, max_rows = 20,
    title = "Point prevalence by health zone"
);

ZonePopulationInfected but not yet symptomatic per 100,000Symptomatic, undetected per 100,000In the community per 100,000Number in the community
Beni42885448.4 (23.8–98.4)52.9 (26.8–101.7)101.0 (54.4–194.3)433 (233–833)
Mambasa15080730.9 (7.4–104.5)32.1 (12.2–76.8)64.6 (21.0–175.2)97 (32–264)
Mandima28663715.9 (4.6–45.6)28.5 (12.9–57.7)45.0 (19.5–100.0)129 (56–287)
Bunia50906014.4 (5.7–32.2)18.2 (9.1–34.9)32.6 (16.0–65.0)166 (81–331)
Lita11336411.6 (4.0–36.3)14.9 (6.5–34.0)26.9 (11.5–70.0)30 (13–79)
Nizi12969711.2 (3.7–33.6)14.4 (6.1–33.5)26.4 (10.5–64.9)34 (14–84)
Oicha27499812.7 (5.0–32.9)13.1 (5.9–28.5)25.9 (11.5–59.7)71 (32–164)
Mutwanga25527811.6 (3.5–34.8)10.7 (4.1–24.9)22.5 (7.9–58.0)57 (20–148)
Rwampara25232410.0 (3.6–24.2)12.3 (5.9–25.4)22.3 (10.3–47.7)56 (26–120)
Kalunguta19173210.0 (4.2–23.8)11.0 (4.9–23.3)21.2 (9.7–46.0)41 (19–88)
Mabalako1971369.1 (3.4–25.2)11.1 (4.7–25.2)20.3 (8.5–48.6)40 (17–96)
Mangala897137.7 (2.5–25.6)10.6 (4.4–26.6)18.4 (7.4–50.7)17 (7–45)
Wamba2247578.3 (3.2–20.2)9.2 (4.2–19.4)17.6 (8.0–38.7)39 (18–87)
Butembo2893787.0 (2.9–17.2)8.2 (3.8–17.3)15.3 (7.0–33.5)44 (20–97)
Nia-Nia917565.4 (1.4–24.9)9.6 (3.8–25.3)15.2 (5.8–49.4)14 (5–45)
Katwa6122424.9 (2.1–12.2)5.8 (2.8–12.0)10.8 (5.0–24.0)66 (31–147)
Pawa3017684.1 (1.5–10.7)4.8 (2.1–11.0)8.9 (3.8–21.0)27 (11–64)
Nyankunde940043.1 (1.0–9.9)4.0 (1.6–10.3)7.3 (2.8–20.3)7 (3–19)
Lolwa413142.9 (0.8–11.5)3.8 (1.2–11.1)7.0 (2.2–22.4)3 (1–9)
Vuhovi1519533.1 (1.1–8.5)3.6 (1.6–8.8)6.7 (2.8–16.8)10 (4–26)
Manguredjipa680142.5 (1.1–6.2)2.9 (1.2–6.5)5.5 (2.3–12.5)4 (2–9)
Kilo431392.2 (0.7–8.9)2.7 (1.0–8.1)5.0 (1.7–16.6)2 (1–7)
Tchomia839172.2 (0.7–7.5)2.7 (0.9–7.6)4.9 (1.8–14.9)4 (1–13)
Musienene3483922.2 (0.8–6.7)2.6 (1.1–6.2)4.9 (2.0–12.5)17 (7–44)
Bambu1361512.0 (0.7–6.7)2.7 (1.0–6.9)4.9 (1.8–13.3)7 (2–18)
Mongbwalu1577141.7 (0.5–7.1)2.9 (1.0–8.0)4.7 (1.5–14.8)7 (2–23)
Biena1361592.0 (0.9–4.6)2.3 (1.0–4.9)4.3 (2.0–9.3)6 (3–13)
Komanda2199481.4 (0.4–4.5)2.3 (0.9–5.7)3.7 (1.4–9.9)8 (3–22)
Isiro5124011.5 (0.5–5.3)1.6 (0.6–4.2)3.1 (1.2–9.1)16 (6–47)
Boma Mangbetu2301041.3 (0.5–3.7)1.5 (0.6–3.6)2.9 (1.1–7.3)7 (3–17)
Kyondo2703441.2 (0.4–3.9)1.3 (0.5–3.4)2.6 (1.0–7.2)7 (3–20)
Mangobo1886521.2 (0.5–2.9)1.3 (0.5–2.9)2.5 (1.1–5.8)5 (2–11)
Bafwasende1322641.2 (0.5–2.7)1.2 (0.5–2.9)2.4 (1.1–5.5)3 (1–7)
Lubero3775061.0 (0.3–4.0)1.2 (0.4–3.5)2.3 (0.7–7.4)9 (3–28)
Makiso-Kisangani5357981.1 (0.5–2.5)1.2 (0.5–2.6)2.3 (1.0–4.9)12 (6–26)
Wapinda977121.2 (0.3–5.1)1.1 (0.3–3.8)2.3 (0.6–8.9)2 (1–9)
Kabondo2046291.1 (0.4–2.6)1.1 (0.5–2.7)2.2 (1.0–5.2)5 (2–11)
Masereka2986360.9 (0.3–3.0)1.1 (0.4–2.8)2.0 (0.8–5.7)6 (2–17)
Yakoma1670620.6 (0.1–2.7)0.5 (0.1–2.0)1.2 (0.3–4.6)2 (0–8)
Gety1434420.5 (0.2–1.3)0.6 (0.2–1.6)1.1 (0.5–2.9)2 (1–4)
Damas948200.5 (0.1–2.1)0.6 (0.2–2.1)1.1 (0.3–4.0)1 (0–4)
Boga720280.4 (0.1–1.3)0.5 (0.1–1.7)0.9 (0.2–3.0)1 (0–2)
Rungu1948040.4 (0.1–1.1)0.5 (0.2–1.1)0.9 (0.3–2.2)2 (1–4)
Rimba2222820.3 (0.1–0.9)0.4 (0.2–1.1)0.8 (0.3–1.9)2 (1–4)
Aru3296670.3 (0.1–0.9)0.4 (0.2–1.1)0.7 (0.3–1.9)2 (1–6)
Logo2169440.3 (0.1–0.8)0.4 (0.1–1.0)0.7 (0.3–1.8)2 (1–4)
Fataki1580320.3 (0.1–1.4)0.4 (0.1–1.3)0.7 (0.2–2.7)1 (0–4)
Tshopo1728850.3 (0.1–0.8)0.3 (0.1–0.9)0.6 (0.2–1.7)1 (0–3)
Drodro1999960.2 (0.1–0.6)0.3 (0.1–0.8)0.5 (0.2–1.3)1 (0–3)
Alimbongo2968210.2 (0.1–0.7)0.3 (0.1–0.7)0.5 (0.2–1.4)2 (1–4)
Viadana1878730.2 (0.1–0.7)0.2 (0.1–0.7)0.4 (0.1–1.4)1 (0–3)
Ganga2354010.2 (0.0–0.6)0.2 (0.1–0.6)0.4 (0.1–1.2)1 (0–3)
Bulu1674810.2 (0.2–0.2)0.2 (0.1–0.3)0.4 (0.3–0.4)1 (0–1)
Lubunga1547120.2 (0.0–0.6)0.2 (0.0–0.6)0.3 (0.1–1.2)1 (0–2)
Wanie-Rukula1552970.1 (0.0–0.5)0.2 (0.0–0.5)0.3 (0.1–0.9)0 (0–1)
Adja1269330.1 (0.0–0.4)0.2 (0.0–0.5)0.3 (0.1–0.9)0 (0–1)
Gombari1229010.1 (0.0–0.4)0.1 (0.0–0.4)0.3 (0.1–0.8)0 (0–1)
Aungba2167870.1 (0.0–0.3)0.1 (0.0–0.4)0.2 (0.1–0.7)1 (0–2)
Kayna3226280.1 (0.0–0.4)0.1 (0.0–0.4)0.2 (0.1–0.7)1 (0–2)
Mahagi2478930.1 (0.0–0.3)0.1 (0.0–0.3)0.2 (0.1–0.6)1 (0–2)
Miti-Murhesa3039830.1 (0.1–0.1)0.1 (0.1–0.1)0.2 (0.2–0.2)1 (0–1)
Kambala1409560.1 (0.0–0.3)0.1 (0.0–0.4)0.2 (0.0–0.6)0 (0–1)
Ariwara3147300.1 (0.0–0.2)0.1 (0.0–0.3)0.2 (0.0–0.5)1 (0–2)
Buta2749640.1 (0.0–0.2)0.1 (0.0–0.3)0.1 (0.0–0.5)0 (0–1)
Goma2828380.0 (0.0–0.1)0.0 (0.0–0.2)0.1 (0.0–0.3)0 (0–1)
Dungu3077630.0 (0.0–0.1)0.0 (0.0–0.1)0.1 (0.0–0.2)0 (0–1)

Composition checks ​

Whether the model reproduces each zone's observed share of its patch's confirmed cases and deaths is on the in-sample checks page.

Data currency ​

The zone blocks are the per-zone confirmed-case and confirmed-death tables of the situation reports. A block that stops being picked up shows its last date before the cut-off.

Zone block currency
julia
# Every zone block whose last vintage falls before the cut-off. The grace
# the national table allows is not applied here: one vintage behind is
# already worth reading.
zone_currency = let
    status = stream_report_status(obs; stratum = :zone)
    behind = status[[ismissing(d) || d > 0 for d in status.days_since], :]
    isempty(behind) ?
        Markdown.parse(
            "Every zone block reports to the cut-off, $(obs.cutoff)."
        ) :
        MarkdownTable(
            DataFrame(
                "Block" => behind.label,
                "Last reported" => [
                    ismissing(d) ? "never" : string(d)
                    for d in behind.last_date
                ],
                "Days before cut-off" => [
                    ismissing(d) ? "-" : string(d)
                    for d in behind.days_since
                ]
            )
        )
end;

Every zone block reports to the cut-off, 2026-11-01.

Health-zone fit diagnostics ​

Zone fit diagnostics
julia
_zone_per_chain(v) = join(string.(round.(v; sigdigits = 3)), " / ")
function _zone_sampler_row(label, chn)
    d = zone_sampler_diagnostics(chn)
    return (
        fit = label, max_rhat = round(d.max_rhat; digits = 3),
        min_ess_bulk = round(d.min_ess_bulk; digits = 0),
        min_ess_tail = round(d.min_ess_tail; digits = 0),
        divergences = d.n_divergent,
        depth_cap = _zone_per_chain(d.depth_cap_fraction),
        ebfmi = _zone_per_chain(d.ebfmi),
        step_size = _zone_per_chain(d.step_size),
    )
end
zone_sampler_table = DataFrame(
    [
        _zone_sampler_row("health zones", chn_local),
        _zone_sampler_row("health zones (frozen)", frozen_local.chn),
    ]
);
zone_diagnostics = let d = zone_diagnostics_table(chn_local, zone_inputs)
    for c in names(d)[3:end]
        d[!, c] = round.(d[!, c]; digits = startswith(c, "rhat") ? 3 : 0)
    end
    DataFrame(
        [
            n == "walking" ?
                "R modelled separately" => [w ? "yes" : "no" for w in d[!, n]] :
                n => d[!, n]
                for n in names(d)
        ]
    )
end;
fitmax_rhatmin_ess_bulkmin_ess_taildivergencesdepth_capebfmistep_size
health zones1.01324545000.001 / 0.00.897 / 1.010.00871 / 0.00671
health zones (frozen)1.01914926100.0 / 0.00.839 / 0.8720.00667 / 0.00808
Sampler diagnostics of the two zone fits. The worst R-hat and smallest effective sample sizes cover every stored quantity except the zone reproduction number. Per chain, the table also gives the fraction of iterations at the tree-depth cap, the energy fraction of missing information and the adapted step size.
Per-zone R-hat and effective sample sizes
zoneR modelled separatelyrhat_R_Tess_bulk_R_Tess_tail_R_Trhat_share_Tess_bulk_share_Tess_tail_share_Trhat_delta_Tess_bulk_delta_Tess_tail_delta_T
AdjanoNaNNaNNaN0.999224213441.00110631306
AriwaranoNaNNaNNaN1.00115041352112311282
Aruno1191214740.999100314671.0034451370
AungbanoNaNNaNNaN1174015171.0027531157
Bambuyes1.001222414390.999187514841.00226641443
BoganoNaNNaNNaN1.003207015861.00212101245
Buniayes1.001231716141.00125121718120671436
Damasyes1.001184713181.00115861278129141240
DrodronoNaNNaNNaN1135713961.00113521133
Fatakiyes1250712481270416571.00230941218
GetynoNaNNaNNaN0.999160815921.00319561468
KambalanoNaNNaNNaN0.999164313451.0018531355
Kiloyes1.001208115261.001224316161.00426241311
Komandayes1.00119121359121431605127051346
Litayes1219915031.004246815521.00322811439
Logono1.001185914931.001189715531.0017081298
Lolwayes1.003227313061200815710.99931011611
MahaginoNaNNaNNaN1.001181814371.0015681164
Mambasayes1.00122371382121831128123391198
Mandimayes1267014781.001257916571.00430111459
Mangalayes1221314181.00121941625129211668
Mongbwaluyes1238218421253117901.00129611156
Nia-Niayes1.001260614941253416741.00132181427
Niziyes1.001228816211.00122751691124531552
Nyankundeyes1217216360.99921611588129211587
RimbanoNaNNaNNaN1129714361.0015291341
Rwamparayes0.99925081690122781633125331556
Tchomiayes1235514411216516390.99925661459
AlimbongonoNaNNaNNaN12300189516281068
Beniyes0.999267115580.99922781586127191503
Bienano1247616311202918531.00124271187
Butemboyes1.004238112941.003238214851.00232381478
GomanoNaNNaNNaN120271517111381278
Kalungutayes1.001243015121.001219713951.00130141385
Katwayes1.002236915111194715011.00130651624
KaynanoNaNNaNNaN1207818331.0028321294
Kyondoyes1.001264416621.002221817941.00125361510
Luberoyes1.005309216331.00229711741129491505
Mabalakoyes1.004252812591.001218012951.00427741175
ManguredjipanoNaNNaNNaN1150316541.00213531127
Maserekayes1.001296814811.00126581581135011321
Musieneneyes1.00130921666125991681139591416
Mutwangayes1.001303915601.001245617921.00331901479
Oichayes1.00232321413132171537137001584
Vuhoviyes1233115101.00118791943125421374
Boma Mangbetuyes1.001242414071.001204617601.00224721475
DungunoNaNNaNNaN1.0021870175417411493
GombarinoNaNNaNNaN1.0021769193918811276
Isiroyes1.001278913491269019631.00228701447
Pawayes1.001210114741239114221.00222301219
RungunoNaNNaNNaN1155818281.0037021459
Wambayes1.001307316951.001297918531.00129581630
Miti-MurhesanoNaNNaNNaN0.9992024NaNNaNNaNNaN
BafwasendenoNaNNaNNaN1.00315011768117791336
KabondonoNaNNaNNaN1164315661.00121581214
LubunganoNaNNaNNaN1.001288718821.00217001203
Makiso-Kisanganino0.999260613731.00119091698111541304
MangobonoNaNNaNNaN1.004181018131.00112241167
TshoponoNaNNaNNaN1277520081.00119951388
Wanie-RukulanoNaNNaNNaN1.00125631757120951441
ButanoNaNNaNNaN1.001326117411.00125171301
GanganoNaNNaNNaN0.999244817861.00125531258
ViadananoNaNNaNNaN126571826122151539
BulunoNaNNaNNaN11682NaNNaNNaNNaN
WapindanoNaNNaNNaN1327914101.00328001448
YakomanoNaNNaNNaN1327914101.00328001448
Per-zone R-hat and effective sample sizes of the cut-off reproduction number, share and deviation. The rows whose reproduction number is modelled separately are the ones to read.
Reproduction number from the zone stage and the joint fit
julia
# Each patch's implied reproduction number from the joint draws with the
# joint posterior mean generation interval, so the patch line is close to
# the quantity the zone values average to.
zone_grid = zone_inputs.t0:obs.n
_patch_infection_draws = vec(collect(chn_joint[:infections_patch]));
patch_implied_rt = [
    let m = Matrix{Float64}(
            undef,
            length(_patch_infection_draws), obs.n
        )
        for (i, v) in enumerate(_patch_infection_draws)
            I = reshape(Float64.(v), N_PATCHES, obs.n)
            m[i, :] .= implied_national_Rt(I[p, :], zone_inputs.g)
        end
        m
    end
        for p in 1:N_PATCHES
];
# The implied reproduction number of each row of `draws` (draws × days).
function _implied_rt_matrix(draws::AbstractMatrix)
    m = similar(draws, Float64)
    for i in axes(draws, 1)
        m[i, :] .= implied_national_Rt(view(draws, i, :), zone_inputs.g)
    end
    return m
end
# The zone stage's deformed patch trajectories on one side and the joint's
# draws on the other, nationally then per patch.
zone_stage_rt = let I = zone_patch_infections(chn_local, zone_inputs)
    vcat([_implied_rt_matrix(sum(I))], _implied_rt_matrix.(I))
end;
joint_stage_rt = let draws = _patch_infection_draws
    national = Float64[
        sum(reshape(Float64.(v), N_PATCHES, obs.n)[:, t])
            for v in draws, t in 1:obs.n
    ]
    vcat([_implied_rt_matrix(national)], patch_implied_rt)
end;
zone_meld_rt_fig = plot_rt_zones(
    [m[:, zone_grid] for m in zone_stage_rt],
    vcat(["National"], zone_inputs.patch_labels),
    vcat([N_PATCHES + 1], 1:N_PATCHES);
    patch_labels = vcat(zone_inputs.patch_labels, ["National"]),
    patch_colours = vcat(BVDOutbreakSize._ZONE_PATCH_COLOURS, [:black]),
    dates = grid_date.(zone_grid), as_of_date = obs.cutoff,
    top = N_PATCHES + 1, ncols = 3, unit = "patch",
    reference_rt = [m[:, zone_grid] for m in joint_stage_rt],
    reference_label = "Headline joint fit",
    title = "Reproduction number from the zone stage and the joint fit"
);

The reproduction number implied by the zone stage's patch trajectories against that implied by the joint fit, nationally and for each patch, both using the joint posterior mean generation interval. Agreement means the melding has kept the joint's patch trajectories.

Health-zone parameters against their priors ​

Draw from the zone model's prior
julia
prior_chn_zone = zone_prior_draws(zone_inputs);
Compute the zone prior and posterior table
julia
# Each scalar zone hyperparameter with its table label and pair-plot axis
# label, kept where both chains carry it: the mixing and correlation
# blocks are sampled only when their inputs are on.
zone_hyper = [
    h for h in (
            (:region_sd_zone, "Level spread σ_level", "σ_level"),
            (:region_halflife_zone, "Deviation half-life (days)", "half-life (days)"),
            (:correlation_reference_zone, "Correlation ρ_corr", "ρ_corr"),
            (
                :zone_ascertainment_sd, "Ascertainment spread σ_ascertainment",
                "σ_ascertainment",
            ),
            (
                :zone_severity_sd, "Severity spread σ_severity (logit)",
                "σ_severity",
            ),
            (:mixing_within_zone, "Within-patch mixing ε_within", "ε_within"),
            (:mixing_departure_zone, "Mixing departure τ_mix", "τ_mix"),
            (:composition_rho_zone, "Case composition ρ", "ρ"),
            (:composition_rho_death_zone, "Death composition ρ_death", "ρ_death"),
        )
        if BVDOutbreakSize._has_key(chn_local, h[1]) &&
        BVDOutbreakSize._has_key(prior_chn_zone, h[1])
]
zone_hyper_keys = first.(zone_hyper)
_hyper_draws(chn, k) = Float64.(vec(collect(chn[k])))
_drift_draws(chn, p) = Float64[
    v[p] for v in vec(collect(chn[:region_drift_sd_zone]))
]
function _prior_posterior_row(label, post, prior)
    f(x) = string(round(x; sigdigits = 3))
    ci(v) = string(
        f(median(v)), " (", f(quantile(v, 0.05)), "–",
        f(quantile(v, 0.95)), ")"
    )
    return (
        parameter = label, posterior = ci(post), prior = ci(prior),
        sd_ratio = round(std(post) / std(prior); digits = 2),
    )
end
zone_prior_table = DataFrame(
    vcat(
        [
            _prior_posterior_row(
                label, _hyper_draws(chn_local, k),
                _hyper_draws(prior_chn_zone, k)
            )
                for (k, label, _) in zone_hyper
        ],
        [
            _prior_posterior_row(
                "Drift scale σ_δ ($(zone_inputs.patch_labels[p]))",
                _drift_draws(chn_local, p), _drift_draws(prior_chn_zone, p)
            )
                for p in eachindex(zone_inputs.patch_labels)
        ]
    )
);
# The pair plot's axes take the symbols alone.
zone_hyper_pair_fig = plot_pair(
    chn_local, zone_hyper_keys;
    prior = prior_chn_zone,
    labels = Dict(k => sym for (k, _, sym) in zone_hyper)
);
parameterposteriorpriorsd_ratio
Level spread σ_level0.485 (0.0662–0.87)0.224 (0.02–0.628)1.23
Deviation half-life (days)19.2 (11.7–30.1)42.1 (15.0–103.0)0.19
Correlation ρ_corr1.23e-5 (9.83e-8–0.000371)0.394 (0.0453–0.855)0
Ascertainment spread σ_ascertainment0.185 (0.0905–0.274)0.122 (0.0444–0.402)0.39
Severity spread σ_severity (logit)0.169 (0.0175–0.4)0.104 (0.0078–0.307)1.27
Within-patch mixing ε_within0.0875 (0.0462–0.157)0.0332 (0.00337–0.141)0.74
Mixing departure τ_mix0.6 (0.102–1.09)0.308 (0.0272–1.02)0.96
Case composition ρ0.0282 (0.0255–0.0311)0.0343 (0.0258–0.0438)0.32
Death composition ρ_death0.0382 (0.0331–0.0437)0.0497 (0.0372–0.0639)0.4
Drift scale σ_δ (Ituri)0.477 (0.361–0.627)0.107 (0.0556–0.206)1.71
Drift scale σ_δ (Nord-Kivu)0.295 (0.211–0.432)0.104 (0.0594–0.183)1.67
Drift scale σ_δ (Haut-Uele)0.298 (0.186–0.477)0.11 (0.0574–0.196)2.11
Drift scale σ_δ (Sud-Kivu)0.107 (0.0563–0.208)0.102 (0.0594–0.194)1.04
Drift scale σ_δ (Tshopo)0.107 (0.0578–0.197)0.109 (0.0571–0.198)0.98
Drift scale σ_δ (Bas-Uele)0.106 (0.0555–0.2)0.106 (0.058–0.194)1.08
Drift scale σ_δ (Sud-Ubangi)0.107 (0.0581–0.192)0.106 (0.0573–0.189)0.99
Drift scale σ_δ (Nord-Ubangi)0.107 (0.0581–0.195)0.106 (0.0602–0.191)0.95
Posterior against prior for each zone hyperparameter, with the ratio of their standard deviations, and the drift scale given per patch. A ratio near one means the zone data add little to the prior, and a ratio above one means the data move the parameter into the prior's wider tail.
Zone hyperparameter pair plot (prior overlaid)

Joint posterior of the scalar zone hyperparameters, with the prior overlaid.

Change over the past week ​

How each zone's reproduction number has moved between the fit a week earlier and the current one is on the in-sample checks page.

Saving zone assets ​

The dashboard map reads the per-zone estimates and time series written here, and the summary dashboard's health-zone table is built from them. The zone forecast assets are written by the health-zone forecasts and health-zone forecast evaluation pages.

Write the zone dashboard and release assets
julia
dashboard_dir = joinpath(
    pkgdir(BVDOutbreakSize), "docs", "src", "summary_assets"
)
mkpath(dashboard_dir)
# The health-zone maps, the weekly cases against R and the one-week zone forecast for the summary
# dashboard, and the per-zone estimates the interactive map reads: one row per zone keyed as the
# geojson keys it, with the cases and deaths to date, the reproduction
# number and the chance it exceeds one, the one-week forecast and the
# share of the patch's infections, whether the zone's reproduction number
# is modelled separately (`walking`) and the data cut-off. A zone below the
# reporting floor carries no reproduction number.
CairoMakie.save(joinpath(dashboard_dir, "zone_rt_map.png"), zone_map_fig)
CairoMakie.save(
    joinpath(dashboard_dir, "cases_rt_zones.png"),
    plot_cases_rt(
        zone_cases_rt; patch_labels = zone_inputs.patch_labels, top = 12,
        unit = "health zone", size = (680, 560), fontsize = 12
    )
)
_zone_deaths = [
    let h = obs.zone_death_history
        haskey(h, prov) && haskey(h[prov], z) &&
            !isempty(h[prov][z].counts) ? Int(h[prov][z].counts[end]) :
            0
    end
        for (prov, z) in zip(
            zone_inputs.zone_province,
            zone_inputs.zone_names
        )
]
_zone_rt_stat(f) = [isempty(r) ? missing : f(r) for r in zone_RT_finite]
zone_estimates = DataFrame(
    zone = zone_map_keys,
    label = zone_inputs.zone_labels, province = zone_inputs.zone_province,
    patch = zone_inputs.patch_labels[zone_patch],
    cases = zone_inputs.cumulative, deaths = _zone_deaths,
    R_T_median = _zone_rt_stat(median),
    p_rt_above_one = [
        isempty(r) ? missing : mean(r .> 1) for r in zone_RT_finite
    ],
    R_T_lower = _zone_rt_stat(r -> quantile(r, 0.05)),
    R_T_upper = _zone_rt_stat(r -> quantile(r, 0.95)),
    forecast_median = [median(v) for v in zone_fc_draws.zones],
    forecast_lower = [quantile(v, 0.05) for v in zone_fc_draws.zones],
    forecast_upper = [quantile(v, 0.95) for v in zone_fc_draws.zones],
    share_median = [median(v) for v in zone_share_T],
    share_lower = [quantile(v, 0.05) for v in zone_share_T],
    share_upper = [quantile(v, 0.95) for v in zone_share_T],
    p_ge_1 = zone_fc_probs[:, 1], p_ge_5 = zone_fc_probs[:, 2],
    p_ge_10 = zone_fc_probs[:, 3], p_ge_20 = zone_fc_probs[:, 4],
    cases_last_7 = zone_recent[7], cases_last_14 = zone_recent[14],
    cases_last_28 = zone_recent[28],
    last_case_date = [ismissing(d) ? "" : string(d) for d in zone_last_case],
    walking = Int.(zone_inputs.walking),
    as_of = fill(string(obs.cutoff), length(zone_map_keys))
)
CSV.write(joinpath(dashboard_dir, "zone_estimates.csv"), zone_estimates)
# Each zone's daily reproduction number, weekly allocated confirmed cases
# with their posterior predictive, and weekly confirmed deaths, for the
# dashboard map's detail column. A zone's deaths are its own cumulative
# series, so each zone is binned on its own vintages.
#
# The predictive is the zone composition's, which splits each patch's
# observed count across its zones, so it checks the allocation rather than
# the patch total. The deaths have none: the model zeroes the reattributed
# death vintages that the observed series keeps, so the two would not bin
# alike.
_zone_case_ppc = zone_composition_draws(chn_local, zone_inputs);
_with_series(t, s) = (t[!, :series] .= s; t)
_zone_weekly_deaths = let h = obs.zone_death_history
    out = DataFrame(area = String[], date = Date[], count = Int[])
    for (prov, z, key) in zip(
            zone_inputs.zone_province, zone_inputs.zone_names, zone_map_keys
        )
        haskey(h, prov) && haskey(h[prov], z) &&
            !isempty(h[prov][z].counts) || continue
        c = h[prov][z].counts
        append!(
            out,
            weekly_count_table(
                h[prov][z].days, reshape([c[1]; max.(diff(c), 0)], 1, :),
                [key]; cutoff = obs.cutoff, n = obs.n
            )
        )
    end
    out
end
CSV.write(
    joinpath(dashboard_dir, "zone_timeseries.csv"),
    vcat(
        _with_series(
            rt_quantile_table(
                zone_rt_traj, zone_map_keys; cutoff = obs.cutoff,
                n = obs.n, from = zone_inputs.t0
            ), "rt"
        ),
        _with_series(
            weekly_count_table(
                zone_inputs.days, zone_inputs.counts, zone_map_keys;
                cutoff = obs.cutoff, n = zone_inputs.n
            ), "cases"
        ),
        _with_series(_zone_weekly_deaths, "deaths"),
        _with_series(
            weekly_quantile_table(
                zone_inputs.days, _zone_case_ppc.predictive, zone_map_keys;
                cutoff = obs.cutoff, n = zone_inputs.n
            ), "cases_fit"
        );
        cols = :union
    )
)
open(joinpath(dashboard_dir, "zone_headline.md"), "w") do io
    print(io, zone_headline(zone_estimates; top = 10))
end
# The zone estimates and time series into the release outputs, so each
# results release carries the dashboard map's zone inputs at its cut-off.
output_dir = get(
    ENV, "BVD_OUTPUT_DIR",
    joinpath(pkgdir(BVDOutbreakSize), "output")
)
mkpath(output_dir)
for f in ("zone_estimates.csv", "zone_timeseries.csv")
    cp(joinpath(dashboard_dir, f), joinpath(output_dir, f); force = true)
end

The full analysis code, data and model definitions are in the epiforecasts/BVDOutbreakSize repository.