Skip to content

Health-zone forecast evaluation ​

How the health-zone split on the health-zone forecasts page has scored against what each zone went on to report. The patch totals are scored as in the national forecast evaluation and the zone split against two persistence rules of its own. The same scoring by province is on the province forecast evaluation page. The in-sample zone checks are on the health-zone in-sample checks page.

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 zone fit melded from the frozen joint, loaded from the cache here.
frozen_local = load_fit("local_frozen_validation");

Summary ​

The overall bullets come first, then a table of each province's scores, from the sections below. The split columns give the log score of each rule's allocation of the week's cases, which is higher for a better allocation. The total columns score the zone model's projected provincial total.

  • Split: the zone model allocates the week better than both persistence rules in 3 of 5 provinces scored, over 472 observed confirmed cases.
ProvinceSplit, zone modelSplit, share persistenceSplit, naive persistenceObserved casesBias of the totalTotal within 90%
Ituri-49.17-130.63-74.482200.5yes
Nord-Kivu-33.18-105.55-36.192090.34yes
Haut-Uele-11.39-10.81-10.8331-0.23yes
Tshopo-7.25-6.93-10.7890.32yes
Nord-Ubangi-2.97-6.24-4.163-0.82yes

Across releases

  • 37 of 64 zones scored beat the persistence baseline, from the forecast made on 2026-09-27 onwards.

Forecast by health zone ​

The one-week-ahead zone forecast comes from the health-zone model melded from the frozen fit, run a week past the frozen cut-off. The observed count is the change in each zone's cumulative confirmed cases between the frozen cut-off and the vintage a week later, clamped at zero. The week is scored only when the zone tables carry a vintage on its last day. The split and the patch total are scored against the persistence rules of forecast scoring against a persistence baseline.

Zone forecast against observed
julia
# `frozen_zone_stage_inputs` is defined in the shared setup, so the zone
# estimates page reads the same frozen inputs.
frozen_zone_inputs = frozen_zone_stage_inputs(; forecast = true);
# The week is scored when the current zone tables carry a vintage a week
# past the frozen cut-off. Otherwise the section shows a note in place of
# its outputs.
zone_validation_horizon = frozen_zone_inputs.model_data.forecast.horizon
zone_truth = zone_forecast_truth(
    obs, frozen_zone_inputs;
    made_date = frozen_local.o.cutoff, horizon = zone_validation_horizon
)
_zone_validation_missing = Markdown.parse(
    "_The zone tables carry no vintage a week after the frozen cut-off, so " *
        "the zone forecast is not scored in this build._"
)
if any(!ismissing, zone_truth)
    zone_validation_forecast = zone_forecast(
        frozen_local.chn, frozen_zone_inputs
    )
    zone_validation_table = zone_forecast_vs_truth(
        zone_validation_forecast, frozen_zone_inputs; truth = zone_truth
    )
    zone_validation_scores = zone_forecast_scores(
        zone_validation_forecast, frozen_zone_inputs; truth = zone_truth
    )
    zone_validation_fig = plot_zone_forecast(
        zone_validation_table;
        patch_labels = frozen_zone_inputs.patch_labels,
        xlabel = "New confirmed cases over $(zone_validation_horizon) days",
        title = "Zone forecast from $(frozen_local.o.cutoff) against observed"
    )
else
    # The frames stay frames, since the release assets below write them.
    zone_validation_table = DataFrame()
    zone_validation_scores = DataFrame()
    zone_validation_fig = _zone_validation_missing
end;
# The scores rounded for display. Share persistence forecasts the split
# alone, so its total columns are blank. No patch is scored when every
# observed total is zero or incomplete, which leaves an empty frame
# without the columns.
zone_validation_scores_table = isempty(zone_validation_scores) ?
    _zone_validation_missing :
    let s = zone_validation_scores
        fmt(x, d) = ismissing(x) || isnan(x) ? "" :
        string(round(x; digits = d))
        DataFrame(
            "Province" => s.patch, "Rule" => s.method,
            "Observed total" => s.observed,
            "Log score of the split" => [fmt(x, 2) for x in s.log_score],
            "CRPS of the total" => [fmt(x, 1) for x in s.crps],
            "Log CRPS" => [fmt(x, 2) for x in s.log_crps],
            "Dispersion" => [fmt(x, 1) for x in s.dispersion],
            "Bias" => [fmt(x, 2) for x in s.bias],
            "Total within 90%" => [
                ismissing(x) ? "" : string(x)
                for x in s.coverage_90
            ]
        )
end;
zone_validation_table_display = isempty(zone_validation_table) ?
    _zone_validation_missing :
    zone_validation_table;

The fifteen zones with the largest forecast medians against what each went on to report, with every zone in the fold below the scores.
ProvinceRuleObserved totalLog score of the splitCRPS of the totalLog CRPSDispersionBiasTotal within 90%
Iturizone model220-49.1734.20.1320.80.5true
Iturishare persistence220-130.63
Iturinaive persistence220-74.4826.20.1125.30.12true
Nord-Kivuzone model209-33.1823.90.117.80.34true
Nord-Kivushare persistence209-105.55
Nord-Kivunaive persistence209-36.1923.30.1123.30.04true
Haut-Uelezone model31-11.394.40.163.8-0.23true
Haut-Ueleshare persistence31-10.81
Haut-Uelenaive persistence31-10.836.90.282.4-0.7true
Tshopozone model9-7.252.60.22.10.32true
Tshoposhare persistence9-6.93
Tshoponaive persistence9-10.781.20.121.20.02true
Nord-Ubangizone model3-2.971.70.660.3-0.82true
Nord-Ubangishare persistence3-6.24
Nord-Ubanginaive persistence3-4.161.60.580.2-0.88true
Scores of each province's split and total for the model and the persistence rules. Share persistence forecasts the split alone, so its total columns are blank.
Zone forecast against observed, every zone
zonepatchcentral_estimatelower_90lower_50upper_50upper_90observedwithin_90
AdjaIturi000010true
AriwaraIturi000020true
AruIturi000190true
AungbaIturi000020true
BambuIturi1003142true
BogaIturi000030true
BuniaIturi5118347111263true
DamasIturi000050true
DrodroIturi000030true
FatakiIturi000040true
GetyIturi000060true
KambalaIturi000000true
KiloIturi000042true
KomandaIturi40111270true
LitaIturi80316359true
LogoIturi000050true
LolwaIturi000181true
MahagiIturi000020true
MambasaIturi1427254822true
MandimaIturi108477814420358true
MangalaIturi30192611true
MongbwaluIturi1003135true
Nia-NiaIturi4019245true
NiziIturi803163816true
NyankundeIturi1004163true
RimbaIturi000053true
RwamparaIturi24514386519true
TchomiaIturi1004151true
Patch totalIturi274156220338441220true
AlimbongoNord-Kivu000031true
BeniNord-Kivu105507913519299true
BienaNord-Kivu000292true
ButemboNord-Kivu1216214112true
GomaNord-Kivu000000true
KalungutaNord-Kivu111620399true
KatwaNord-Kivu18410295217true
KaynaNord-Kivu000020true
KyondoNord-Kivu1004141true
LuberoNord-Kivu2006172true
MabalakoNord-Kivu20411325516true
ManguredjipaNord-Kivu000190true
MaserekaNord-Kivu0002114true
MusieneneNord-Kivu4019245true
MutwangaNord-Kivu1215224222true
OichaNord-Kivu21413346019true
VuhoviNord-Kivu3007200true
Patch totalNord-Kivu241140196299405209true
Boma MangbetuHaut-Uele201481true
DunguHaut-Uele000000true
GombariHaut-Uele000010true
IsiroHaut-Uele2014116true
PawaHaut-Uele61311197true
RunguHaut-Uele000023false
WambaHaut-Uele1338203714true
Patch totalHaut-Uele26716386631true
Miti-MurhesaSud-Kivu000020true
Patch totalSud-Kivu000020true
BafwasendeTshopo200370true
KabondoTshopo201374true
LubungaTshopo000020true
Makiso-KisanganiTshopo5139174true
MangoboTshopo201480true
TshopoTshopo000131true
Wanie-RukulaTshopo000020true
Patch totalTshopo123719369true
ButaBas-Uele000010true
GangaBas-Uele000130true
ViadanaBas-Uele000120true
Patch totalBas-Uele100250true
BuluSud-Ubangi000020true
Patch totalSud-Ubangi000020true
WapindaNord-Ubangi000130true
YakomaNord-Ubangi000013false
Patch totalNord-Ubangi000143true

Forecast by health zone across releases ​

The archived zone forecast of each release is scored against what each zone went on to report. The scores are those of the province forecast evaluation, with each zone scored as its own stream against a persistence baseline built from its own history. A week is left out when the zone tables carry no vintage on its last day, when it holds a harmonisation-break day, or when its province moved unallocated cases into named zones during it. The figures show the ten zones with the most observed cases over the scored weeks, and the folds hold every zone.

Load and summarise the zone forecast scores
julia
# Written by `scripts/score_releases.jl` from each release's
# `zone_forecast.csv`. Missing files read as empty tables.
zone_release_scores_df = _release_data(
    joinpath("zone", "forecast_scores.csv"),
    (;
        release = String, made_date = Date, stream = String, horizon = Int,
        target_date = Date, fit = String, crps = Float64,
        log_crps = Float64, dispersion = Float64, overprediction = Float64,
        underprediction = Float64, coverage_50 = Float64,
        coverage_90 = Float64,
        bias = Float64, n_samples = Int,
        log_rel_to_baseline = Float64,
    )
)
zone_release_overlay_df = scored_overlay(
    _release_data(
        joinpath("zone", "forecast_overlay.csv"),
        (;
            release = String, made_date = Date, stream = String,
            horizon = Int, target_date = Date, fit = String,
            observed = Float64, median = Float64, lo30 = Float64,
            hi30 = Float64, lo60 = Float64, hi60 = Float64, lo90 = Float64,
            hi90 = Float64,
        )
    )
)
# Every scored zone, ordered by its observed cases over the scored weeks.
zone_release_keys = let o = zone_release_overlay_df[
        zone_release_overlay_df.fit .== JOINT_FIT, :,
    ]
    tot = Dict{String, Float64}()
    for r in eachrow(o)
        k = zone_score_key(r.stream)
        k === nothing || (tot[k] = get(tot, k, 0.0) + r.observed)
    end
    sort!(collect(keys(tot)); by = k -> (-tot[k], k))
end
zone_release_labels = let lab = Dict(
        zip(frozen_zone_inputs.zone_keys, frozen_zone_inputs.zone_labels)
    )
    [get(lab, k, k) for k in zone_release_keys]
end
zone_release_top = first(zone_release_keys, 10)
_zone_release(tbl, keys = zone_release_keys) = zone_score_rows(
    tbl, keys, zone_release_labels[indexin(keys, zone_release_keys)]
)
# Zones in the order of `zone_release_keys`, under a `zone` column.
_zone_release_display(tbl) = let t = drop_degenerate_fit_column(
        drop_individual_fit_columns(tbl)
    )
    t = t[sortperm(indexin(t.stream, zone_release_labels)), :]
    DataFrame([(c == "stream" ? "zone" : c) => t[!, c] for c in names(t)])
end
zone_release_by_horizon = forecast_score_by_horizon(
    _zone_release(zone_release_scores_df, zone_release_top)
)
zone_release_by_release = forecast_score_by_release(
    _zone_release(zone_release_scores_df, zone_release_top)
)
zone_release_overview_display = _zone_release_display(
    forecast_score_overview(_zone_release(zone_release_scores_df))
)
zone_release_by_release_display = _zone_release_display(
    forecast_score_by_release(_zone_release(zone_release_scores_df))
);

_zone_release_empty = "No zone forecast has been scored yet. Scores " *
    "appear from the first release that archives the zone forecast onwards.";
# A plain statement above the tables: when nothing is scored yet, and
# otherwise the first cut-off scored.
zone_release_note = Markdown.parse(
    size(zone_release_scores_df, 1) == 0 ? _zone_release_empty :
        string(
            "Scores run from the forecast made on ",
            minimum(zone_release_scores_df.made_date), "."
        )
);

Scores run from the forecast made on 2026-09-27.

julia
zone_release_crps_fig = plot_forecast_crps_by_horizon(
    zone_release_by_horizon;
    title = "CRPS decomposition, by health zone",
    empty_message = _zone_release_empty
);

The mean CRPS split into its width, overprediction and underprediction, one panel per zone.
julia
zone_release_by_cutoff_fig = plot_forecast_skill_by_cutoff(
    zone_release_by_release;
    title = "Relative skill against the baseline by release, by health zone",
    empty_message = _zone_release_empty
);

Relative skill against the baseline release by release, on a log scale with the reference line at one.
julia
zone_release_overlay_fig = plot_forecast_overlay(
    _zone_release(zone_release_overlay_df, zone_release_top);
    empty_message = _zone_release_empty, ncol = 3
);

Each release's zone forecasts against what each zone went on to report, by the cut-off each was made from. Each forecast shows its median and 90% predictive interval beside the persistence baseline and the observed count.
Zone scores across releases, every zone
zonencrpsrel_to_baselinelog_crpslog_rel_to_baselinedispersionoverpredictionunderpredictioncoverage_50coverage_90bias
Mandima613.931.440.171.289.753.750.420.8310.18
Beni610.11.730.1221.528.381.7200.8310.22
Bunia611.581.050.1980.96.821.193.570.831-0.14
Katwa68.650.760.3270.743.671.233.760.50.83-0.06
Rwampara69.331.070.3360.94.923.341.070.510.28
Nizi65.191.050.3440.973.060.281.850.671-0.24
Butembo65.40.830.3270.792.90.492.010.831-0.12
Mambasa66.881.080.6991.692.410.324.140.670.83-0.42
Mabalako64.390.970.2740.993.020.680.70.671-0.01
Oicha63.580.720.3670.922.350.031.190.51-0.31
Wamba62.941.060.2380.911.860.061.020.671-0.16
Mangala63.640.790.430.892.010.011.610.671-0.33
Nia-Nia64.580.880.330.863.141.360.080.8310.25
Lita62.521.130.2781.112.220.010.2911-0.15
Kalunguta63.22.260.2471.722.560.6300.8310.25
Musienene63.880.750.4470.791.820.091.970.671-0.22
Mutwanga63.591.040.5941.131.1702.410.331-0.5
Pawa62.71.460.2741.111.890.8100.6710.32
Komanda63.660.950.5660.992.131.010.520.3310.27
Bambu62.060.730.4350.811.20.040.830.831-0.25
Mongbwalu63.231.010.5781.031.970.880.370.8310.32
Lubero61.51.160.3991.40.8600.640.671-0.31
Masereka61.581.270.4171.231.020.050.510.671-0.18
Kabondo61.840.860.480.920.3401.490.330.83-0.5
Rimba62.551.031.0031.330.0602.4900.83-0.83
Kyondo61.411.110.4041.160.900.50.831-0.27
Vuhovi62.420.660.5440.771.470.70.250.6710.25
Makiso-Kisangani61.881.330.4140.991.040.810.030.6710.31
Isiro61.370.940.4420.910.60.150.610.51-0.13
Nyankunde61.451.280.5671.540.5100.950.51-0.51
Aru61.890.650.5780.90.0901.80.50.83-0.33
Bafwasende62.180.680.6460.710.330.211.640.50.830.08
Boma Mangbetu60.960.980.2920.870.670.210.07110.2
Mangobo61.461.090.6191.240.450.390.620.170.830.03
Gety61.062.020.5622.080.100.960.331-0.45
Biena61.020.590.4270.580.530.040.450.671-0.02
Rungu60.90.730.4720.750.0500.850.330.67-0.48
Kilo60.810.70.4160.790.0700.740.51-0.3
Lolwa60.61.240.3361.490.300.30.831-0.31
Ganga60.560.670.3030.670.1800.380.671-0.01
Tchomia60.590.370.240.550.4200.18110.11
Tshopo60.430.630.2580.650.0900.340.671-0.21
Fataki60.430.570.2310.660.1300.30.831-0.07
Manguredjipa60.490.810.2310.690.400.09110.16
Alimbongo50.350.190.220.30.1100.240.81-0.13
Kayna60.340.380.1810.370.0900.260.8310.02
Adja60.160.390.1080.40.0100.150.830.83-0.1
Logo60.190.560.1180.520.0700.120.8310.03
Mahagi60.180.630.1160.60.0300.150.830.83-0.06
Buta60.0720.960.04718.840.0700110.25
Viadana60.210.360.1350.430.2100110.41
Dungu60.012.280.0062.160.0100110.08
Gombari60.010.570.0080.520.0100110.1
Ariwara60.020.330.0110.260.0200110.1
Aungba60.020.440.0110.310.0200110.1
Boga60.030.970.0120.630.0300110.1
Damas60.10.260.0450.220.100110.2
Drodro60.080.590.0410.460.0800110.2
Kambala60.011.180.0070.80.0100110.08
Goma60.024.270.013.050.0200110.09
Wapinda20.430.240.2440.260.360.070110.52
Miti-Murhesa60.110.0680.1100110.28
Bulu60.060.40.0380.380.0600110.22
Lubunga60.1139.010.07434.80.1100110.31
Wanie-Rukula60.0624.450.04220.680.0600110.24
Zone scores by release, every zone
made_datezonencrpsrel_to_baselinelog_crpslog_rel_to_baselinedispersionoverpredictionunderpredictioncoverage_50coverage_90bias
2026-09-27Mandima19.993.340.1462.948.51.490110.27
2026-10-02Mandima110.680.910.1560.958.1302.5411-0.32
2026-10-06Mandima110.081.40.131.19100.080110.07
2026-10-10Mandima111.172.710.1132.3410.620.550110.14
2026-10-20Mandima111.34.010.1083.5610.660.640110.17
2026-10-25Mandima130.361.040.3670.9310.6219.740010.78
2026-09-27Beni16.93.150.1082.796.190.710110.22
2026-10-02Beni19.3320.1291.577.671.660110.26
2026-10-06Beni18.071.120.1050.988.010.060110.05
2026-10-10Beni114.697.070.1785.699.235.460010.53
2026-10-20Beni112.030.740.1140.589.572.460110.29
2026-10-25Beni19.583.770.0983.779.5800110
2026-09-27Bunia119.531.010.3530.855.15014.3801-0.78
2026-10-02Bunia110.920.890.1890.856.983.940110.44
2026-10-06Bunia17.760.940.1160.787.6500.1111-0.07
2026-10-10Bunia110.391.010.1810.96.1804.2211-0.45
2026-10-20Bunia111.291.490.1621.068.073.220110.38
2026-10-25Bunia19.571.160.1851.086.8802.6911-0.33
2026-09-27Katwa118.761.10.712.62016.1400-0.93
2026-10-02Katwa110.071.080.3311.073.6806.3901-0.63
2026-10-06Katwa16.030.40.1920.444.641.390110.34
2026-10-10Katwa110.530.530.4080.584.725.810010.72
2026-10-20Katwa13.640.980.1620.623.460.180110.15
2026-10-25Katwa12.890.830.1670.72.8900110.02
2026-09-27Rwampara19.81.440.3951.233.3506.4501-0.64
2026-10-02Rwampara19.071.160.2590.976.112.960110.44
2026-10-06Rwampara112.941.050.4815.147.80010.76
2026-10-10Rwampara15.951.060.1740.695.690.260110.12
2026-10-20Rwampara113.890.940.5270.925.438.460010.75
2026-10-25Rwampara14.330.850.1790.513.780.550110.23
2026-09-27Nizi13.630.770.3530.992.2801.3511-0.4
2026-10-02Nizi16.51.290.4381.292.7903.7101-0.56
2026-10-06Nizi17.061.020.3680.832.9204.1401-0.61
2026-10-10Nizi14.640.910.2080.834.130.510110.22
2026-10-20Nizi14.921.20.3040.783.741.190110.36
2026-10-25Nizi14.361.140.391.12.4901.8711-0.43
2026-09-27Butembo111.150.750.7740.651.5209.6301-0.85
2026-10-02Butembo14.891.190.2791.292.6902.211-0.51
2026-10-06Butembo15.090.920.2460.843.781.310110.4
2026-10-10Butembo15.850.790.2710.724.251.60110.43
2026-10-20Butembo12.991.320.1991.682.7900.211-0.17
2026-10-25Butembo12.390.480.1940.662.3900110.02
2026-09-27Mambasa16.0415.531.45832.770.2105.8300-0.92
2026-10-02Mambasa13.071.090.5111.41.501.5811-0.48
2026-10-06Mambasa14.450.770.3160.93.4401.0111-0.33
2026-10-10Mambasa116.911.031.331.471.7015.2101-0.89
2026-10-20Mambasa16.42.540.3251.954.481.920110.44
2026-10-25Mambasa14.390.430.2510.393.1601.2311-0.36
2026-09-27Mabalako15.330.810.5380.951.6103.7201-0.69
2026-10-02Mabalako12.362.630.173.182.2300.1311-0.15
2026-10-06Mabalako14.282.580.2092.23.480.80110.3
2026-10-10Mabalako16.970.980.3540.843.833.140010.59
2026-10-20Mabalako13.770.940.1850.963.4400.3311-0.2
2026-10-25Mabalako13.650.530.1890.573.530.120110.12
2026-09-27Oicha11.970.340.5410.810.7601.2101-0.58
2026-10-02Oicha13.711.930.6422.821.0902.6101-0.65
2026-10-06Oicha12.760.720.2570.772.1800.5811-0.28
2026-10-10Oicha13.234.420.223.333.050.180110.15
2026-10-20Oicha15.920.380.3520.353.1602.7601-0.52
2026-10-25Oicha13.92.080.1872.163.890.020110.04
2026-09-27Wamba11.460.80.170.631.460011-0.01
2026-10-02Wamba11.951.110.210.851.580.380110.33
2026-10-06Wamba14.560.910.3830.811.4903.0701-0.63
2026-10-10Wamba15.021.360.3551.362.0502.9601-0.56
2026-10-20Wamba12.471.060.1541.072.4700110
2026-10-25Wamba12.191.110.1540.922.1200.0611-0.11
2026-09-27Mangala16.551.210.541.062.204.3601-0.61
2026-10-02Mangala12.160.330.2780.542.1100.0411-0.07
2026-10-06Mangala12.250.380.2690.532.170.080110.14
2026-10-10Mangala12.980.930.4080.841.9301.0511-0.38
2026-10-20Mangala13.721.170.4771.262.0501.6711-0.45
2026-10-25Mangala14.181.260.6091.211.6202.5601-0.59
2026-09-27Nia-Nia16.651.680.2511.44.721.930110.45
2026-10-02Nia-Nia18.240.550.4340.5844.240010.65
2026-10-06Nia-Nia13.331.380.2551.582.9700.3611-0.19
2026-10-10Nia-Nia14.870.930.4090.863.151.720110.48
2026-10-20Nia-Nia12.811.360.3170.82.520.290110.24
2026-10-25Nia-Nia11.550.690.3140.91.4500.111-0.14
2026-09-27Lita12.561.120.2891.022.0300.5311-0.32
2026-10-02Lita12.091.050.2510.832.0500.0411-0.08
2026-10-06Lita12.421.190.2861.32.1300.2911-0.23
2026-10-10Lita12.931.450.3421.452.0900.8411-0.35
2026-10-20Lita12.681.210.2481.12.630.050110.11
2026-10-25Lita12.40.870.2541.12.3900.0211-0.04
2026-09-27Kalunguta12.392.380.2331.451.980.410110.26
2026-10-02Kalunguta13.53.970.2713.012.520.980110.38
2026-10-06Kalunguta13.13.750.2192.62.860.240110.19
2026-10-10Kalunguta15.235.450.373.763.122.120010.53
2026-10-20Kalunguta12.650.790.1860.622.6500110.02
2026-10-25Kalunguta12.311.550.2041.582.260.050110.09
2026-09-27Musienene18.910.790.9720.681.0607.8501-0.9
2026-10-02Musienene15.50.940.5611.011.6703.8301-0.65
2026-10-06Musienene13.190.540.2890.572.710.480110.29
2026-10-10Musienene12.260.450.250.612.260011-0.03
2026-10-20Musienene11.790.960.2870.981.720.070110.14
2026-10-25Musienene11.631.360.3251.451.500.1311-0.17
2026-09-27Mutwanga12.283.380.814.950.3201.9601-0.77
2026-10-02Mutwanga14.971.61.0862.420.5704.3901-0.84
2026-10-06Mutwanga11.060.260.3150.460.8900.1811-0.25
2026-10-10Mutwanga10.920.160.2950.220.890.020110.16
2026-10-20Mutwanga15.991.060.641.411.5704.4201-0.71
2026-10-25Mutwanga16.294.760.426.842.7803.5101-0.56
2026-09-27Pawa12.551.940.1861.732.180.370110.25
2026-10-02Pawa12.532.470.1751.952.350.190110.15
2026-10-06Pawa11.851.410.1570.931.770.080110.15
2026-10-10Pawa13.580.770.4340.721.811.770010.69
2026-10-20Pawa14.263.430.5142.271.82.460010.74
2026-10-25Pawa11.430.920.1810.621.430011-0.03
2026-09-27Komanda14.660.550.340.543.221.440110.42
2026-10-02Komanda14.50.930.5650.832.631.880010.64
2026-10-06Komanda12.341.460.3161.421.8600.4811-0.27
2026-10-10Komanda13.951.290.6421.371.2902.6601-0.6
2026-10-20Komanda13.831.650.5941.272.41.430010.6
2026-10-25Komanda12.680.950.9380.971.371.310010.82
2026-09-27Bambu11.830.860.3620.971.3700.4611-0.34
2026-10-02Bambu14.831.040.810.94103.8301-0.73
2026-10-06Bambu11.990.930.3591.161.6300.3711-0.27
2026-10-10Bambu11.790.330.3590.441.580.210110.28
2026-10-20Bambu11.030.780.340.770.9400.0911-0.17
2026-10-25Bambu10.890.620.3770.920.6800.2111-0.28
2026-09-27Mongbwalu14.21.430.2940.823.220.980110.36
2026-10-02Mongbwalu13.210.970.3760.842.330.880110.45
2026-10-06Mongbwalu13.420.810.4780.822.121.310110.49
2026-10-10Mongbwalu13.230.840.6730.921.991.240110.58
2026-10-20Mongbwalu12.550.90.8171.121.670.890110.74
2026-10-25Mongbwalu12.771.30.8281.610.5202.2501-0.71
2026-09-27Lubero10.451.180.2371.310.4300.0111-0.21
2026-10-02Lubero10.631.820.2551.60.630011-0.03
2026-10-06Lubero12.360.90.5630.980.9401.4201-0.56
2026-10-10Lubero12.873.70.6045.131.0301.8401-0.58
2026-10-20Lubero11.771.490.4242.51.2500.5211-0.35
2026-10-25Lubero10.920.380.3130.610.8700.0511-0.15
2026-09-27Masereka11.842.580.6033.280.4901.3501-0.68
2026-10-02Masereka11.290.990.2610.831.2100.0811-0.15
2026-10-06Masereka11.831.390.3111.21.610.220110.23
2026-10-10Masereka11.591.360.3041.311.520.060110.12
2026-10-20Masereka10.841.210.3021.150.8400110.06
2026-10-25Masereka12.080.920.7240.920.4601.6201-0.65
2026-09-27Kabondo10.266.190.1565.260.2600110.43
2026-10-02Kabondo10.40.640.2560.590.2500.1511-0.33
2026-10-06Kabondo10.890.570.430.540.3100.5801-0.7
2026-10-10Kabondo11.350.840.5130.930.3101.0401-0.73
2026-10-20Kabondo16.721.021.1191.210.4806.2400-0.98
2026-10-25Kabondo11.410.580.4081.050.4600.9501-0.66
2026-09-27Rimba10.731.250.4951.250.0600.6701-0.72
2026-10-02Rimba15.371.011.6821.080.0205.3400-0.95
2026-10-06Rimba13.991.711.3612.580.0703.9101-0.9
2026-10-10Rimba10.710.170.4830.430.0600.6601-0.72
2026-10-20Rimba12.421.21.091.30.0502.3701-0.88
2026-10-25Rimba12.076.830.90710.870.1201.9501-0.8
2026-09-27Kyondo11.040.890.3620.810.7400.311-0.34
2026-10-02Kyondo11.191.010.4140.920.7500.4411-0.36
2026-10-06Kyondo11.321.80.2981.841.320011-0.05
2026-10-10Kyondo13.171.090.7261.40.9402.2301-0.66
2026-10-20Kyondo11.040.90.3281.131.0100.0311-0.1
2026-10-25Kyondo10.681.510.2981.380.6500.0211-0.12
2026-09-27Vuhovi12.143.220.5774.370.6801.4601-0.63
2026-10-02Vuhovi11.630.570.2390.471.5900.0411-0.09
2026-10-06Vuhovi11.982.810.2681.791.920.060110.1
2026-10-10Vuhovi14.5410.9070.812.022.520010.78
2026-10-20Vuhovi12.320.190.5620.291.410.920110.58
2026-10-25Vuhovi11.892.340.7121.771.210.680110.74
2026-09-27Makiso-Kisangani10.861.280.1630.980.8600110.01
2026-10-02Makiso-Kisangani10.921.360.1841.240.890.020110.18
2026-10-06Makiso-Kisangani10.960.750.1930.670.7700.1911-0.28
2026-10-10Makiso-Kisangani13.271.140.7910.981.351.920010.85
2026-10-20Makiso-Kisangani14.139.880.9664.861.32.830010.92
2026-10-25Makiso-Kisangani11.120.430.1850.21.050.070110.18
2026-09-27Isiro11.041.20.3231.040.750.290110.48
2026-10-02Isiro11.251.160.6211.270.620.640010.78
2026-10-06Isiro10.540.610.2240.520.4600.0811-0.27
2026-10-10Isiro12.31.120.7681.060.3501.9401-0.86
2026-10-20Isiro10.941.230.2381.330.7500.1911-0.28
2026-10-25Isiro12.150.70.4790.610.6901.4601-0.64
2026-09-27Nyankunde11.241.990.642.420.1501.0901-0.72
2026-10-02Nyankunde11.472.160.6013.040.4601.0111-0.53
2026-10-06Nyankunde11.811.630.7681.820.2301.5801-0.72
2026-10-10Nyankunde10.710.540.3040.730.6500.0611-0.17
2026-10-20Nyankunde12.340.990.6791.010.7601.5801-0.57
2026-10-25Nyankunde11.151.640.4071.70.7900.3611-0.34
2026-09-27Aru10.030.70.0150.520.0300110.13
2026-10-02Aru10.030.810.0150.50.0300110.12
2026-10-06Aru10.761.050.5171.060.0400.7201-0.77
2026-10-10Aru11.62.110.862.730.0401.5601-0.86
2026-10-20Aru18.592.481.92550.0808.5100-0.94
2026-10-25Aru10.30.020.1340.050.300110.35
2026-09-27Bafwasende10.242.750.1522.690.2400110.44
2026-10-02Bafwasende10.212.890.1312.80.2100110.41
2026-10-06Bafwasende11.90.820.7590.760.2201.6801-0.91
2026-10-10Bafwasende18.430.911.6570.840.2608.1700-0.99
2026-10-20Bafwasende11.330.20.6550.330.590.750010.78
2026-10-25Bafwasende10.971.350.5221.260.480.490110.74
2026-09-27Boma Mangbetu11.011.380.2751.460.6400.3711-0.43
2026-10-02Boma Mangbetu10.70.930.1980.780.6500.0611-0.18
2026-10-06Boma Mangbetu11.080.520.3370.560.750.330110.49
2026-10-10Boma Mangbetu10.970.770.3120.760.690.290110.46
2026-10-20Boma Mangbetu11.352.660.41.680.770.580110.6
2026-10-25Boma Mangbetu10.621.240.230.730.530.090110.29
2026-09-27Mangobo11.470.940.5671.050.3201.1601-0.76
2026-10-02Mangobo12.470.920.80810.2902.1900-0.92
2026-10-06Mangobo10.695.210.3186.90.2900.411-0.53
2026-10-10Mangobo11.060.30.5550.380.530.530010.76
2026-10-20Mangobo11.8621.660.85114.990.721.130010.86
2026-10-25Mangobo11.213.340.61410.340.540.660010.78
2026-09-27Gety11.531.10.8181.120.0501.4801-0.84
2026-10-02Gety11.531.110.8271.140.0501.4801-0.81
2026-10-06Gety11.4110.040.74915.170.0901.3201-0.75
2026-10-10Gety11.5413.60.82520.70.0501.501-0.84
2026-10-20Gety10.213.870.0822.40.2100110.27
2026-10-25Gety10.142.250.0681.690.1400110.26
2026-09-27Biena11.010.870.5050.850.2800.7401-0.61
2026-10-02Biena12.051.010.6861.090.5301.5201-0.68
2026-10-06Biena10.920.340.4110.360.710.220110.61
2026-10-10Biena10.710.250.3010.260.670.040110.52
2026-10-20Biena10.50.790.2340.650.500110.47
2026-10-25Biena10.940.90.4280.830.5100.4311-0.41
2026-09-27Rungu11.570.850.830.830.0401.5400-0.94
2026-10-02Rungu10.040.020.0280.030.0400110.19
2026-10-06Rungu10.630.830.4321.410.0500.5701-0.72
2026-10-10Rungu10.620.920.4290.920.0500.5701-0.72
2026-10-20Rungu10.070.10.0420.090.0700110.23
2026-10-25Rungu12.451.551.0721.990.0402.400-0.94
2026-09-27Kilo12.41.771.0852.20.0402.3601-0.88
2026-10-02Kilo10.710.540.4871.080.0500.6601-0.72
2026-10-06Kilo10.120.050.0610.060.1200110.24
2026-10-10Kilo10.070.10.0320.080.0700110.18
2026-10-20Kilo10.060.330.0290.250.0600110.17
2026-10-25Kilo11.511.310.8031.40.0701.4401-0.81
2026-09-27Lolwa10.660.870.3291.160.4400.2111-0.31
2026-10-02Lolwa10.621.370.3521.660.300.3211-0.41
2026-10-06Lolwa10.631.410.3411.680.3500.2811-0.38
2026-10-10Lolwa10.681.490.4392.210.1300.5501-0.61
2026-10-20Lolwa10.381.320.1541.020.3800110.36
2026-10-25Lolwa10.641.240.41.30.1700.4711-0.54
2026-09-27Ganga10.134.240.0874.070.1300110.34
2026-10-02Ganga11.260.770.6410.740.1201.1401-0.84
2026-10-06Ganga11.270.760.6420.720.1201.1501-0.88
2026-10-10Ganga10.250.150.1580.180.2500110.45
2026-10-20Ganga10.3110.070.1979.620.30.010110.5
2026-10-25Ganga10.143.640.0943.590.1400110.35
2026-09-27Tchomia10.650.820.2770.690.6500110.49
2026-10-02Tchomia10.30.840.1240.650.300110.32
2026-10-06Tchomia10.150.490.070.420.1500110.26
2026-10-10Tchomia10.130.450.0580.360.1300110.24
2026-10-20Tchomia11.660.870.6361.810.6101.0511-0.51
2026-10-25Tchomia10.650.110.2750.210.650011-0.12
2026-09-27Tshopo10.0510.820.0310.010.0500110.2
2026-10-02Tshopo10.590.670.4070.660.0700.5201-0.67
2026-10-06Tshopo10.10.120.070.120.100110.31
2026-10-10Tshopo10.10.120.0680.120.100110.3
2026-10-20Tshopo11.221.590.6291.980.1201.101-0.81
2026-10-25Tshopo10.50.680.3421.150.100.411-0.6
2026-09-27Fataki10.651.010.3841.450.2400.4111-0.49
2026-10-02Fataki11.481.540.7881.730.0601.4101-0.81
2026-10-06Fataki10.20.20.0920.20.200110.3
2026-10-10Fataki10.10.080.0470.090.100110.21
2026-10-20Fataki10.10.270.0420.240.100110.2
2026-10-25Fataki10.060.160.0320.160.0600110.18
2026-09-27Manguredjipa10.260.190.1390.180.2600110.38
2026-10-02Manguredjipa10.413.340.22.680.4100110.44
2026-10-06Manguredjipa10.910.760.4340.710.4200.4911-0.48
2026-10-10Manguredjipa10.652.650.2812.150.6300.0211-0.15
2026-10-20Manguredjipa10.390.770.1810.560.3900110.42
2026-10-25Manguredjipa10.332.210.1481.560.3300110.37
2026-10-02Alimbongo10.590.210.3840.450.1300.4611-0.59
2026-10-06Alimbongo10.160.030.0790.050.1600110.28
2026-10-10Alimbongo10.150.170.0760.130.1500110.28
2026-10-20Alimbongo10.061.810.0281.780.0600110.17
2026-10-25Alimbongo10.780.990.5360.990.0300.7501-0.79
2026-09-27Kayna11.570.850.8440.850.0301.5401-0.87
2026-10-02Kayna10.10.060.0470.050.100110.21
2026-10-06Kayna10.160.090.0830.080.1600110.29
2026-10-10Kayna10.188.050.0895.980.1800110.3
2026-10-20Kayna10.021.670.0121.510.0200110.12
2026-10-25Kayna10.021.590.011.230.0200110.1
2026-09-27Adja100.060.0020.08000110.04
2026-10-02Adja10.911.170.6261.20.0100.900-0.92
2026-10-06Adja10.020.030.0090.020.0200110.09
2026-10-10Adja10.010.010.0030.010.0100110.06
2026-10-20Adja100.10.0020.09000110.04
2026-10-25Adja10.010.240.0060.270.0100110.08
2026-09-27Logo10.050.610.0260.430.0500110.16
2026-10-02Logo10.771.420.5271.440.0300.7401-0.79
2026-10-06Logo10.070.120.0370.10.0700110.2
2026-10-10Logo10.040.060.0190.040.0400110.15
2026-10-20Logo10.111.430.050.910.1100110.22
2026-10-25Logo10.11.380.050.990.100110.23
2026-09-27Mahagi10.041.770.0181.080.0400110.13
2026-10-02Mahagi10.021.090.010.670.0200110.1
2026-10-06Mahagi10.041.870.0181.30.0400110.14
2026-10-10Mahagi10.042.790.0161.650.0400110.12
2026-10-20Mahagi10.91.140.6261.15000.900-0.92
2026-10-25Mahagi10.020.020.0080.010.0200110.08
2026-09-27Buta10.0623.160.04120.020.0600110.23
2026-10-02Buta10.0625.860.04223.190.0600110.24
2026-10-06Buta10.0917.760.057160.0900110.28
2026-10-10Buta10.1123.720.06821.290.1100110.3
2026-10-20Buta10.0821.460.05519.910.0800110.28
2026-10-25Buta10.0313.670.02112.330.0300110.17
2026-09-27Viadana10.270.10.1680.130.2700110.46
2026-10-02Viadana10.20.30.1250.290.200110.4
2026-10-06Viadana10.172.180.1072.130.1700110.37
2026-10-10Viadana10.274.120.1623.760.2700110.44
2026-10-20Viadana10.253.550.1563.340.2500110.44
2026-10-25Viadana10.143.550.0943.360.1400110.35
2026-09-27Dungu10.013.510.0063.410.0100110.09
2026-10-02Dungu10.012.310.0092.230.0100110.11
2026-10-06Dungu10.014.380.0064.220.0100110.09
2026-10-10Dungu10.012.90.0082.670.0100110.1
2026-10-20Dungu10.011.310.0041.20.0100110.07
2026-10-25Dungu100.320.0010.31000110.03
2026-09-27Gombari10.020.690.0110.630.0200110.12
2026-10-02Gombari10.010.360.0060.320.0100110.09
2026-10-06Gombari10.020.690.0130.640.0200110.13
2026-10-10Gombari10.010.730.0080.660.0100110.1
2026-10-20Gombari10.010.80.0080.740.0100110.1
2026-10-25Gombari10.010.280.0040.260.0100110.07
2026-09-27Ariwara10.050.720.0230.560.0500110.15
2026-10-02Ariwara10.030.330.0130.250.0300110.12
2026-10-06Ariwara10.020.260.0110.210.0200110.1
2026-10-10Ariwara10.020.250.0080.220.0200110.09
2026-10-20Ariwara10.010.190.0060.170.0100110.08
2026-10-25Ariwara10.010.180.0060.160.0100110.08
2026-09-27Aungba10.030.820.0140.520.0300110.12
2026-10-02Aungba10.020.30.010.230.0200110.1
2026-10-06Aungba10.040.840.020.620.0400110.14
2026-10-10Aungba10.030.410.0120.290.0300110.11
2026-10-20Aungba10.010.130.0030.10.0100110.06
2026-10-25Aungba10.010.170.0030.130.0100110.06
2026-09-27Boga10.020.610.0090.320.0200110.08
2026-10-02Boga10.030.790.0130.520.0300110.11
2026-10-06Boga10.020.670.0080.440.0200110.08
2026-10-10Boga10.010.540.0060.390.0100110.08
2026-10-20Boga10.032.340.0141.380.0300110.11
2026-10-25Boga10.041.950.0191.480.0400110.15
2026-09-27Damas10.240.620.1140.530.2400110.33
2026-10-02Damas10.050.160.0260.140.0500110.17
2026-10-06Damas10.090.280.0350.210.0900110.18
2026-10-10Damas10.040.10.0170.080.0400110.13
2026-10-20Damas10.090.150.040.120.0900110.19
2026-10-25Damas10.080.330.0360.250.0800110.18
2026-09-27Drodro10.110.670.0510.480.1100110.22
2026-10-02Drodro10.090.490.0450.410.0900110.22
2026-10-06Drodro10.140.720.0690.60.1400110.27
2026-10-10Drodro10.090.690.0470.560.0900110.23
2026-10-20Drodro10.060.530.0280.380.0600110.17
2026-10-25Drodro10.020.250.0090.180.0200110.09
2026-09-27Kambala10.010.950.0070.730.0100110.09
2026-10-02Kambala10.021.220.0070.760.0200110.09
2026-10-06Kambala10.032.190.0161.480.0300110.12
2026-10-10Kambala10.021.480.0091.010.0200110.09
2026-10-20Kambala100.350.0020.23000110.05
2026-10-25Kambala100.50.0020.32000110.04
2026-09-27Goma10.022.590.0112.070.0200110.11
2026-10-02Goma10.023.850.013.180.0200110.11
2026-10-06Goma10.059.020.0236.370.0500110.15
2026-10-10Goma10.037.40.0134.530.0300110.11
2026-10-20Goma100.850.0020.59000110.04
2026-10-25Goma100.460.0010.34000110.02
2026-10-20Wapinda10.560.20.3030.230.420.140110.57
2026-10-25Wapinda10.30.350.1850.310.300110.48
2026-09-27Miti-Murhesa10.070.0440.0700110.24
2026-10-02Miti-Murhesa10.150.1010.1500110.37
2026-10-06Miti-Murhesa10.150.0950.1500110.35
2026-10-10Miti-Murhesa10.210.1320.2100110.4
2026-10-20Miti-Murhesa10.030.020.0300110.16
2026-10-25Miti-Murhesa10.020.0160.0200110.15
2026-09-27Bulu10.040.050.0280.050.0400110.19
2026-10-02Bulu10.043.070.0253.010.0400110.19
2026-10-06Bulu10.031.450.0171.410.0300110.15
2026-10-10Bulu10.063.560.0373.390.0600110.22
2026-10-20Bulu10.1414.170.09213.240.1400110.35
2026-10-25Bulu10.053.940.033.80.0500110.2
2026-09-27Lubunga10.1136.030.07332.030.1100110.32
2026-10-02Lubunga10.1150.780.07447.40.1100110.32
2026-10-06Lubunga10.1227.380.0824.480.1200110.33
2026-10-10Lubunga10.251.830.1345.10.200110.41
2026-10-20Lubunga10.0963.930.06157.710.0900110.29
2026-10-25Lubunga10.0417.40.02715.520.0400110.19
2026-09-27Wanie-Rukula10.0623.210.03819.90.0600110.22
2026-10-02Wanie-Rukula10.0723.190.04420.720.0700110.25
2026-10-06Wanie-Rukula10.0718.670.04616.080.0700110.25
2026-10-10Wanie-Rukula10.0736.540.04630.260.0700110.24
2026-10-20Wanie-Rukula10.0729.720.04223.570.0700110.23
2026-10-25Wanie-Rukula10.0521.240.03317.860.0500110.21

Saving zone forecast outputs ​

The frozen zone fit's one-week-ahead forecast against the observed zone increments, and its scores against the two persistence rules, are written as release assets.

Write zone forecast outputs
julia
output_dir = get(
    ENV, "BVD_OUTPUT_DIR",
    joinpath(pkgdir(BVDOutbreakSize), "output")
)
mkpath(output_dir)
CSV.write(
    joinpath(output_dir, "zone_forecast_validation.csv"),
    zone_validation_table
)
CSV.write(
    joinpath(output_dir, "zone_forecast_scores.csv"),
    zone_validation_scores
)
"/home/runner/work/BVDOutbreakSize/BVDOutbreakSize/output/zone_forecast_scores.csv"
Write the summary bullets
julia
# The bullets under the summary heading at the top of the page. They read
# the scores built further down, so they are written here and read back
# when the site is assembled.
evaluation_forecast_zone_summary = let s = zone_validation_scores
    if isempty(s)
        string(
            "- **Zone forecasts:** the zone tables end at or before the ",
            "frozen cut-off, so no zone forecast is scored in this build."
        )
    else
        patches = unique(s.patch)
        rule(p, m) = let r = s[(s.patch .== p) .& (s.method .== m), :]
            size(r, 1) == 0 ? nothing : first(eachrow(r))
        end
        scored = filter(p -> rule(p, "zone model") !== nothing, patches)
        beats(p) = let m = rule(p, "zone model")
            m === nothing ? false :
                all(
                    r -> r === nothing || m.log_score >= r.log_score,
                    [rule(p, "share persistence"), rule(p, "naive persistence")]
                )
        end
        observed_total = sum(s[s.method .== "zone model", :observed])
        overall = [
            string(
                "- **Split:** the zone model allocates the week better ",
                "than both persistence rules in ", count(beats, scored),
                " of ", length(scored), " provinces scored, over ",
                observed_total, " observed confirmed cases."
            ),
        ]
        # One row per province the zone model scored: the split's log
        # score under each rule, then the zone model's total.
        unscored = setdiff(patches, scored)
        isempty(unscored) || push!(
            overall,
            string("- **Not scored:** ", join_names(unscored), ".")
        )
        score(p, m) = let r = rule(p, m)
            r === nothing ? missing : finite_round(r.log_score)
        end
        model(p) = rule(p, "zone model")
        table = DataFrame(
            "Province" => scored,
            "Split, zone model" => [score(p, "zone model") for p in scored],
            "Split, share persistence" =>
                [score(p, "share persistence") for p in scored],
            "Split, naive persistence" =>
                [score(p, "naive persistence") for p in scored],
            "Observed cases" => [model(p).observed for p in scored],
            "Bias of the total" => Union{Missing, Float64}[
                finite_round(model(p).bias) for p in scored
            ],
            "Total within 90%" => [
                ismissing(model(p).coverage_90) ? "n/a" :
                    model(p).coverage_90 ? "yes" : "no"
                    for p in scored
            ],
        )
        join(
            vcat(
                [join(overall, "\n")],
                isempty(scored) ? String[] :
                    [markdown_table(table; missing_text = "n/a")]
            ), "\n\n"
        )
    end
end
dashboard_dir = joinpath(
    pkgdir(BVDOutbreakSize), "docs", "src", "summary_assets"
)
mkpath(dashboard_dir)
write(
    joinpath(dashboard_dir, "evaluation_forecast_zone.md"),
    evaluation_forecast_zone_summary
);
Add the across-release scores to the summary
julia
# A block on the zone forecasts scored across releases, after the
# per-province blocks above.
zone_release_summary = let ov = forecast_score_overview(
        _zone_release(zone_release_scores_df)
    )
    rows = filter(r -> !ismissing(r.rel_to_baseline), ov)
    size(rows, 1) == 0 ?
        "**Across releases**\n\n- $(_zone_release_empty)" :
        string(
            "**Across releases**\n\n- ", count(<(1), rows.rel_to_baseline),
            " of ", size(rows, 1), " zones scored beat the persistence ",
            "baseline, from the forecast made on ",
            minimum(zone_release_scores_df.made_date), " onwards."
        )
end
open(joinpath(dashboard_dir, "evaluation_forecast_zone.md"), "a") do io
    write(io, "\n\n", zone_release_summary)
end;