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: 18 November 2026.

Data as of: 16 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
Boma MangbetuHaut-Uele518.7 (3.6–20.5)1.13 (0.73–1.7)0.69-0.02 (-0.44–0.41)yes
MangalaIturi46014.0 (4.0–40.8)1.16 (0.56–2.01)0.650.5 (-0.18–1.22)yes
FatakiIturi870.7 (0.2–3.3)1.15 (0.55–2.13)0.630.25 (-0.61–1.06)yes
PawaHaut-Uele13833.8 (17.1–56.0)1.08 (0.71–1.5)0.630.04 (-0.35–0.45)yes
OichaNord-Kivu14113.0 (5.9–27.7)1.09 (0.67–1.62)0.610.24 (-0.29–0.79)yes
MutwangaNord-Kivu1259.9 (4.0–22.5)1.06 (0.62–1.6)0.580.23 (-0.23–0.72)yes
IsiroHaut-Uele12821.4 (7.5–46.7)1.05 (0.56–1.62)0.560.02 (-0.58–0.57)yes
WambaHaut-Uele19026.1 (10.6–49.5)1.03 (0.59–1.54)0.54-0.03 (-0.44–0.41)yes
DamasIturi410.2 (0.0–0.9)1.0 (0.54–1.84)0.50.0 (-0.96–0.87)yes
KyondoNord-Kivu501.3 (0.5–3.4)1.0 (0.64–1.56)0.50.01 (-0.45–0.46)yes
MaserekaNord-Kivu571.6 (0.6–5.0)1.0 (0.57–1.7)0.50.07 (-0.38–0.53)yes
KiloIturi441.1 (0.3–5.0)0.99 (0.51–1.91)0.480.3 (-0.43–1.03)yes
BambuIturi2200.8 (0.2–2.3)0.95 (0.53–1.57)0.430.01 (-0.68–0.7)yes
KalungutaNord-Kivu1235.9 (3.1–11.0)0.95 (0.64–1.34)0.420.01 (-0.38–0.36)yes
MongbwaluIturi7404.0 (0.9–19.2)0.88 (0.35–2.04)0.40.22 (-0.65–1.12)yes
VuhoviNord-Kivu661.9 (0.8–4.0)0.94 (0.62–1.36)0.4-0.0 (-0.35–0.34)yes
NiziIturi8836.0 (2.2–15.2)0.9 (0.47–1.52)0.370.19 (-0.47–0.83)yes
MusieneneNord-Kivu1692.0 (0.8–5.1)0.91 (0.54–1.45)0.37-0.1 (-0.56–0.34)yes
TchomiaIturi860.3 (0.1–1.0)0.88 (0.49–1.48)0.34-0.16 (-0.98–0.66)yes
LitaIturi3523.8 (1.2–11.4)0.85 (0.42–1.51)0.330.09 (-0.63–0.79)yes
ButemboNord-Kivu3563.9 (1.9–7.8)0.89 (0.57–1.33)0.32-0.15 (-0.5–0.19)yes
NyankundeIturi1430.9 (0.3–3.1)0.84 (0.47–1.49)0.31-0.11 (-0.95–0.65)yes
KatwaNord-Kivu6985.7 (2.8–11.4)0.88 (0.57–1.31)0.31-0.14 (-0.49–0.18)yes
BuniaIturi210518.8 (7.8–36.5)0.85 (0.48–1.32)0.30.06 (-0.6–0.71)yes
BeniNord-Kivu90542.8 (28.3–58.7)0.92 (0.67–1.16)0.30.03 (-0.43–0.48)yes
LuberoNord-Kivu460.8 (0.2–2.6)0.83 (0.43–1.48)0.3-0.06 (-0.61–0.44)yes
LolwaIturi370.2 (0.0–0.8)0.78 (0.35–1.57)0.28-0.32 (-1.37–0.6)yes
RwamparaIturi12356.8 (2.6–16.2)0.84 (0.45–1.41)0.28-0.03 (-0.72–0.65)yes
MabalakoNord-Kivu1343.1 (1.3–6.9)0.82 (0.51–1.28)0.25-0.09 (-0.61–0.35)yes
MambasaIturi1606.3 (1.6–22.9)0.67 (0.24–1.59)0.24-0.06 (-1.03–0.92)yes
MandimaIturi63914.8 (3.7–44.5)0.67 (0.23–1.38)0.220.0 (-0.94–0.97)yes
KomandaIturi1950.5 (0.1–1.8)0.69 (0.3–1.35)0.2-0.53 (-1.58–0.46)yes
Nia-NiaIturi3460.9 (0.2–4.5)0.59 (0.21–1.46)0.17-0.32 (-1.4–0.72)yes
WapindaNord-Ubangi653.8 (40.2–74.3)1.18 (0.87–1.5)0.820.0 (-0.0–0.0)no
RunguHaut-Uele112.5 (1.3–4.3)1.11 (0.82–1.4)0.72-0.0 (-0.01–0.0)no
KabondoTshopo2923.6 (16.3–32.1)1.02 (0.8–1.26)0.55-0.0 (-0.0–0.0)no
MangoboTshopo1116.2 (10.0–23.7)1.02 (0.8–1.26)0.550.0 (-0.0–0.0)no
Makiso-KisanganiTshopo4141.7 (33.2–50.1)1.01 (0.79–1.25)0.530.0 (-0.0–0.0)no
BafwasendeTshopo199.8 (5.5–15.8)0.98 (0.78–1.2)0.45-0.0 (-0.0–0.0)no
AlimbongoNord-Kivu60.2 (0.1–0.4)0.93 (0.68–1.22)0.35-0.0 (-0.01–0.0)no
DrodroIturi160.2 (0.1–0.5)0.93 (0.69–1.2)0.340.0 (-0.0–0.0)no
BienaNord-Kivu291.0 (0.6–1.7)0.94 (0.69–1.21)0.34-0.0 (-0.0–0.0)no
KambalaIturi30.1 (0.0–0.2)0.92 (0.67–1.17)0.3-0.0 (-0.01–0.0)no
RimbaIturi290.4 (0.2–0.7)0.92 (0.68–1.17)0.3-0.0 (-0.01–-0.0)no
ManguredjipaNord-Kivu160.4 (0.2–0.8)0.92 (0.68–1.18)0.3-0.0 (-0.0–0.0)no
LogoIturi140.3 (0.1–0.6)0.9 (0.66–1.15)0.25-0.0 (-0.01–0.0)no
MahagiIturi60.1 (0.0–0.2)0.89 (0.65–1.15)0.24-0.0 (-0.01–0.0)no
AruIturi260.3 (0.2–0.7)0.88 (0.63–1.12)0.22-0.0 (-0.01–-0.0)no
AdjaIturi110.1 (0.0–0.1)0.88 (0.62–1.13)0.2-0.0 (-0.0–0.0)no
AungbaIturi120.1 (0.0–0.2)0.87 (0.63–1.12)0.2-0.0 (-0.01–0.0)no
AriwaraIturi80.1 (0.0–0.2)0.86 (0.61–1.11)0.17-0.0 (-0.0–0.0)no
GetyIturi110.2 (0.1–0.4)0.85 (0.62–1.09)0.150.0 (-0.0–0.0)no
BogaIturi20.1 (0.0–0.2)0.82 (0.59–1.08)0.120.0 (-0.0–0.0)no
GomaNord-Kivu10.0 (0.0–0.1)NaN-0.0 (-0.01–0.0)no
KaynaNord-Kivu30.1 (0.0–0.2)NaN-0.0 (-0.01–0.0)no
DunguHaut-Uele10.3 (0.1–0.7)NaN-0.0 (-0.01–0.0)no
GombariHaut-Uele30.5 (0.2–1.1)NaN-0.0 (-0.01–0.0)no
WatsaHaut-Uele100.7 (0.2–1.5)NaN-0.0 (-0.01–-0.0)no
Miti-MurhesaSud-Kivu3100.0 (100.0–100.0)NaN0.0 (0.0–0.0)no
LubungaTshopo11.9 (0.6–4.8)NaN0.0 (-0.0–0.0)no
TshopoTshopo63.9 (1.9–7.3)NaN-0.0 (-0.01–0.0)no
Wanie-RukulaTshopo11.5 (0.5–3.7)NaN-0.0 (-0.0–0.0)no
ButaBas-Uele117.2 (5.7–33.8)NaN-0.0 (-0.0–0.0)no
GangaBas-Uele542.4 (27.1–58.1)NaN0.0 (-0.0–0.0)no
ViadanaBas-Uele639.1 (22.4–58.5)NaN0.0 (-0.0–0.0)no
BuluSud-Ubangi2100.0 (100.0–100.0)NaN0.0 (0.0–0.0)no
YakomaNord-Ubangi446.2 (25.7–59.8)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
Mangala8971355.3 (15.9–183.5)51.0 (20.6–125.1)106.7 (38.3–302.5)96 (34–271)
Beni42885434.4 (17.5–68.8)41.4 (21.3–77.7)76.4 (40.4–142.9)328 (173–613)
Mandima28663720.8 (6.2–61.4)30.9 (13.7–65.6)52.6 (22.3–120.3)151 (64–345)
Mambasa15080716.5 (5.0–56.0)24.3 (10.4–56.2)41.7 (16.5–107.0)63 (25–161)
Nizi12969717.3 (6.5–47.0)19.9 (8.9–45.6)37.8 (16.4–90.8)49 (21–118)
Oicha27499815.8 (6.2–42.0)16.1 (7.3–36.3)32.5 (14.6–76.9)89 (40–212)
Bunia50906013.4 (5.7–30.8)16.6 (8.0–33.6)30.1 (14.8–62.5)153 (75–318)
Lita11336412.5 (4.4–37.0)15.5 (6.4–36.0)28.7 (11.4–72.6)33 (13–82)
Mutwanga25527813.1 (4.7–37.0)13.4 (5.7–32.2)26.7 (11.4–66.3)68 (29–169)
Kalunguta19173210.5 (4.6–24.1)11.9 (5.5–25.4)22.6 (10.5–48.4)43 (20–93)
Rwampara2523249.9 (4.0–25.9)12.4 (5.7–26.4)22.6 (10.4–50.7)57 (26–128)
Mongbwalu1577149.6 (2.5–42.5)12.2 (4.7–32.9)22.5 (7.7–72.7)35 (12–115)
Kilo431399.0 (2.5–40.0)9.9 (3.2–31.0)19.3 (5.9–69.3)8 (3–30)
Wamba2247577.7 (2.7–21.9)7.8 (3.4–18.5)15.6 (6.3–39.0)35 (14–88)
Pawa3017687.5 (2.8–19.6)7.4 (3.3–17.3)15.1 (6.4–36.3)46 (19–110)
Mabalako1971365.6 (2.3–14.4)7.0 (3.1–16.1)12.7 (5.8–29.9)25 (11–59)
Nia-Nia917564.1 (1.1–18.5)7.3 (2.7–19.3)11.6 (4.1–35.4)11 (4–32)
Butembo2893784.7 (2.0–11.3)5.6 (2.6–12.4)10.5 (4.8–22.7)30 (14–66)
Vuhovi1519534.2 (1.6–10.8)4.8 (2.0–11.1)9.1 (3.8–21.2)14 (6–32)
Nyankunde940043.7 (1.3–12.3)4.6 (1.9–12.2)8.5 (3.4–23.2)8 (3–22)
Katwa6122423.3 (1.4–7.6)4.0 (1.9–8.6)7.4 (3.5–15.7)45 (22–96)
Isiro5124012.8 (0.9–7.9)2.8 (1.2–6.7)5.7 (2.2–14.4)29 (11–74)
Biena1361592.5 (1.1–5.5)2.9 (1.3–6.3)5.4 (2.5–11.9)7 (3–16)
Boma Mangbetu2301042.5 (0.8–8.3)2.4 (0.9–6.6)4.9 (1.8–14.2)11 (4–33)
Bambu1361512.1 (0.6–6.8)2.4 (0.8–6.2)4.5 (1.5–12.8)6 (2–17)
Musienene3483922.0 (0.7–5.8)2.4 (1.0–5.7)4.4 (1.8–11.3)15 (6–39)
Manguredjipa680142.0 (0.8–4.8)2.3 (0.9–5.4)4.3 (1.8–10.1)3 (1–7)
Lolwa413141.6 (0.3–7.6)2.2 (0.6–7.3)3.9 (1.0–15.2)2 (0–6)
Masereka2986361.8 (0.6–6.3)2.0 (0.8–5.5)3.8 (1.4–11.9)11 (4–36)
Kyondo2703441.6 (0.6–4.9)1.8 (0.7–4.4)3.4 (1.4–9.2)9 (4–25)
Fataki1580321.6 (0.4–7.6)1.6 (0.5–5.3)3.2 (0.9–12.7)5 (1–20)
Wapinda977121.6 (0.4–6.5)1.5 (0.4–5.2)3.2 (0.9–11.7)3 (1–11)
Kabondo2046291.5 (0.6–3.5)1.5 (0.7–3.4)3.1 (1.4–6.9)6 (3–14)
Tchomia839171.4 (0.4–4.8)1.6 (0.5–4.7)3.0 (1.0–9.3)3 (1–8)
Komanda2199480.9 (0.2–3.2)1.4 (0.5–3.8)2.3 (0.8–6.9)5 (2–15)
Mangobo1886521.1 (0.4–2.8)1.2 (0.5–2.8)2.3 (1.0–5.6)4 (2–11)
Makiso-Kisangani5357981.0 (0.5–2.3)1.1 (0.5–2.3)2.1 (1.0–4.5)11 (5–24)
Bafwasende1322641.0 (0.4–2.5)1.0 (0.4–2.5)2.0 (0.9–5.0)3 (1–7)
Rungu1948040.9 (0.3–2.2)0.8 (0.3–2.1)1.7 (0.6–4.3)3 (1–8)
Lubero3775060.7 (0.2–2.5)0.9 (0.3–2.4)1.7 (0.6–4.7)6 (2–18)
Damas948200.7 (0.1–3.5)0.8 (0.2–2.9)1.5 (0.4–6.3)1 (0–6)
Yakoma1670620.8 (0.2–3.1)0.7 (0.2–2.6)1.5 (0.4–5.7)2 (1–10)
Rimba2222820.6 (0.3–1.5)0.7 (0.3–1.6)1.3 (0.6–3.0)3 (1–7)
Gety1434420.5 (0.2–1.2)0.6 (0.3–1.6)1.2 (0.5–2.8)2 (1–4)
Logo2169440.5 (0.2–1.3)0.6 (0.3–1.5)1.2 (0.5–2.7)3 (1–6)
Boga720280.4 (0.1–1.3)0.5 (0.2–1.5)1.0 (0.3–2.7)1 (0–2)
Drodro1999960.4 (0.2–1.0)0.5 (0.2–1.2)0.9 (0.4–2.2)2 (1–4)
Aru3296670.4 (0.2–0.9)0.5 (0.2–1.1)0.9 (0.4–2.0)3 (1–6)
Tshopo1728850.3 (0.1–0.8)0.3 (0.1–0.8)0.6 (0.2–1.6)1 (0–3)
Gombari1229010.3 (0.1–0.8)0.3 (0.1–0.8)0.5 (0.2–1.6)1 (0–2)
Kambala1409560.2 (0.1–0.6)0.3 (0.1–0.7)0.5 (0.2–1.3)1 (0–2)
Alimbongo2968210.2 (0.1–0.6)0.3 (0.1–0.6)0.5 (0.2–1.2)1 (1–3)
Adja1269330.2 (0.0–0.5)0.2 (0.1–0.6)0.4 (0.1–1.0)0 (0–1)
Mahagi2478930.2 (0.1–0.4)0.2 (0.1–0.5)0.4 (0.1–0.9)1 (0–2)
Lubunga1547120.2 (0.0–0.6)0.2 (0.0–0.6)0.3 (0.1–1.1)1 (0–2)
Aungba2167870.1 (0.0–0.4)0.2 (0.0–0.5)0.3 (0.1–0.9)1 (0–2)
Viadana1878730.1 (0.1–0.3)0.2 (0.1–0.4)0.3 (0.1–0.7)1 (0–1)
Wanie-Rukula1552970.1 (0.0–0.4)0.1 (0.0–0.4)0.3 (0.1–0.8)0 (0–1)
Ganga2354010.1 (0.0–0.3)0.1 (0.0–0.3)0.3 (0.1–0.6)1 (0–1)
Kayna3226280.1 (0.0–0.3)0.1 (0.0–0.4)0.2 (0.1–0.7)1 (0–2)
Bulu1674810.1 (0.1–0.1)0.1 (0.1–0.2)0.2 (0.2–0.3)0 (0–0)
Ariwara3147300.1 (0.0–0.3)0.1 (0.0–0.4)0.2 (0.1–0.7)1 (0–2)
Watsa4915750.1 (0.0–0.3)0.1 (0.0–0.3)0.2 (0.1–0.5)1 (0–3)
Miti-Murhesa3039830.1 (0.1–0.1)0.1 (0.0–0.1)0.1 (0.1–0.2)0 (0–1)
Dungu3077630.1 (0.0–0.2)0.1 (0.0–0.2)0.1 (0.0–0.4)0 (0–1)
Buta2749640.0 (0.0–0.1)0.0 (0.0–0.1)0.1 (0.0–0.3)0 (0–1)
Goma2828380.0 (0.0–0.1)0.0 (0.0–0.1)0.1 (0.0–0.3)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, 2026-11-16.

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 zones1.01815931500.0 / 0.0010.914 / 0.8440.00649 / 0.00749
health zones (frozen)1.01216133600.001 / 0.00.906 / 0.8430.00875 / 0.00749
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
AdjanoNaNNaNNaN1.002178513141.00210441040
AriwaranoNaNNaNNaN1100810891.00214301155
Aruno1.001172614291.003135014311.001381971
AungbanoNaNNaNNaN1.001128916431.003553857
Bambuyes1226914701184417251.00224001398
BoganoNaNNaNNaN1149716341.00210571083
Buniayes1230415351.001235613710.99921681407
Damasyes1.002230614221191915661.00623461361
DrodronoNaNNaNNaN1155615641.00314511082
Fatakiyes1.001210912841228516251.00123341532
GetynoNaNNaNNaN1.001151614751.0031250845
KambalanoNaNNaNNaN1179713661.001574970
Kiloyes1.001226513211.00123251757123481482
Komandayes1.001188114291.005189316501.00421621035
Litayes123851153122241041123621153
Logono1.001169013291.004134313701.002682968
Lolwayes1.001198015251199815971.00122721315
MahaginoNaNNaNNaN1.001132115761597931
Mambasayes1.002197114971.002221314191.00225951306
Mandimayes1.001229115151.001201313471.00425771291
Mangalayes1.00119281534120911638119811262
Mongbwaluyes1.003201610721.0052224163112015837
Nia-Niayes1.001234117631.001202116101.00326231491
Niziyes1.001250313891257918141.00225481582
Nyankundeyes1.003224017251254614801.00221501306
Rimbano1169913851.005111113051.001424781
Rwamparayes0.999211513211.00121451502120991344
Tchomiayes1.003209314221181113691.00126961403
AlimbongonoNaNNaNNaN1221512951.0017561191
Beniyes1251314510.999150814030.99923391425
Bienano1237714881.001188217231.0021622893
Butemboyes1229713110.999225315751.00121981458
GomanoNaNNaNNaN1.001140914951.001785982
Kalungutayes1.001235717211.00324741571123391181
Katwayes1.00123441440123171726123961643
KaynanoNaNNaNNaN1.001189115781.007609975
Kyondoyes1.002168313610.99920701301125571217
Luberoyes1283616081307318061.00128881570
Mabalakoyes1.001197314621.001184917821.00221621735
ManguredjipanoNaNNaNNaN1.00379813601.004738723
Maserekayes126591191122481587124971441
Musieneneyes0.999213914411192713000.99925061421
Mutwangayes1.001210715661.003187611981.00223061482
Oichayes123051456125381548123591224
Vuhoviyes0.99921171236122251613126501032
Boma Mangbetuyes1.003223615871228016421.00122711659
DungunoNaNNaNNaN0.9992493165718241039
GombarinoNaNNaNNaN0.999170617631.0027761151
Isiroyes1.00120561323122871555120481555
Pawayes1.001217414311212616591.00127561550
RungunoNaNNaNNaN1.003147514621.0029641156
Wambayes1.008212714141.002202814881.00523121396
WatsanoNaNNaNNaN1198216801.005336591
Miti-MurhesanoNaNNaNNaN1.00119281981NaNNaNNaN
BafwasendenoNaNNaNNaN1.002109619381.0011069932
Kabondono1263313851183315131.00114361208
LubunganoNaNNaNNaN113731718112211091
Makiso-Kisanganino1.001251814001177017081.0025781039
MangobonoNaNNaNNaN1.002204017441.001820979
TshoponoNaNNaNNaN0.999242119261.0011070961
Wanie-RukulanoNaNNaNNaN1194014951.00214121028
ButanoNaNNaNNaN1.001240016541.00119261088
GanganoNaNNaNNaN0.999281215131.00513661219
ViadananoNaNNaNNaN1.001195917221.00415301074
BulunoNaNNaNNaN1.0021885NaNNaNNaNNaN
WapindanoNaNNaNNaN1.003153911511.00318261358
YakomanoNaNNaNNaN1.003153911511.00318261358
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_level0.707 (0.423–1.03)0.222 (0.0221–0.635)0.98
Deviation half-life (days)22.0 (14.6–33.4)40.6 (15.4–105.0)0.19
Correlation ρ_corr2.05e-5 (1.83e-7–0.000642)0.387 (0.0354–0.897)0
Ascertainment spread σ_ascertainment0.184 (0.0872–0.274)0.139 (0.0468–0.445)0.4
Severity spread σ_severity (logit)0.166 (0.0145–0.402)0.103 (0.0129–0.273)1.46
Within-patch mixing ε_within0.0762 (0.0381–0.141)0.0347 (0.00246–0.129)0.76
Mixing departure τ_mix0.418 (0.0349–0.987)0.313 (0.0285–0.942)1.02
Case composition ρ0.0291 (0.0265–0.0321)0.0348 (0.0269–0.0436)0.32
Death composition ρ_death0.0365 (0.0314–0.0416)0.0451 (0.0338–0.0586)0.38
Drift scale σ_δ (Ituri)0.429 (0.336–0.557)0.108 (0.0585–0.205)1.47
Drift scale σ_δ (Nord-Kivu)0.249 (0.174–0.357)0.111 (0.0587–0.204)1.16
Drift scale σ_δ (Haut-Uele)0.302 (0.189–0.47)0.108 (0.0579–0.196)1.94
Drift scale σ_δ (Sud-Kivu)0.109 (0.0613–0.201)0.111 (0.0583–0.203)0.96
Drift scale σ_δ (Tshopo)0.109 (0.0592–0.208)0.111 (0.0588–0.209)0.96
Drift scale σ_δ (Bas-Uele)0.11 (0.0598–0.196)0.109 (0.064–0.195)1.02
Drift scale σ_δ (Sud-Ubangi)0.109 (0.058–0.204)0.111 (0.0585–0.208)1.04
Drift scale σ_δ (Nord-Ubangi)0.109 (0.0582–0.195)0.115 (0.0604–0.209)0.91
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.