Fig 1.
A) Average annual dengue incidence rate (per 100,000 population) by province over the study period.
B) Monthly incidence rate by geographic region (dashed lines) and nationally (solid line). C) Heatmap of z-scored log-transformed monthly incidence rate by province, ordered by geographic region. Horizontal lines separate regions. Z-scores were calculated within each province. Provinces are ordered as in panel A. Names of provinces for each index number are available in Table A in S1 Appendix. Administrative boundaries from geoBoundaries [37], CC BY 4.0.
Table 1.
West-to-east gradient in peak timing by province subset. Spearman rank correlation between province longitude and peak outbreak month (July-June scale) for each epidemic year (2010-2011 to 2023-2024).
Fig 2.
A) Average phase lag (months) of each province’s annual dengue cycle relative to other provinces.
Negative values (blue) indicate earlier peaks; positive values (red) indicate later peaks. B) Median peak outbreak month by province, identified using circular statistics over July-June annual periods. Administrative boundaries from geoBoundaries [37], CC BY 4.0.
Table 2.
Within-region temporal pattern cohesion. Mean pairwise dynamic time warping (DTW) distance between provinces within each geographic region. Lower values indicate more similar temporal dynamics. SD = standard deviation; CV = coefficient of variation. Regions are ranked from most cohesive (lowest mean DTW distance) to least cohesive.
Fig 3.
A) Cluster assignments for all 34 provinces (k = 12). B) Cluster assignments for western Indonesia including Kalimantan (k = 10); non-included provinces shown in grey. C) Cluster assignments for Sumatra and Java-Bali only (k = 4); non-included provinces shown in grey. D) Within-region dissimilarity z-scores based on mean pairwise DTW distance to regional neighbours. Positive values (red) indicate provinces whose temporal dynamics deviate from their region; negative values (blue) indicate high similarity to regional neighbours. Administrative boundaries from geoBoundaries [37], CC BY 4.0.
Fig 4.
A) Time series of the Oceanic Nino Index (ONI) and Dipole Mode Index (DMI).
Horizontal reference lines indicate thresholds for moderate (+-0.5) and strong (+-1.0, + -1.5, + -2.0) events. B-D) Heatmaps of z-scored monthly precipitation B), temperature C), and relative humidity D) by province, ordered by geographic region. Horizontal lines separate regions. Z-scores were calculated within each province. Dashed vertical lines indicate January of each year.
Fig 5.
A) Wavelet-derived phase lag (months) between local climate variables and dengue incidence.
Positive values (red) indicate climate leading dengue; negative values (blue) indicate dengue leading climate. B) Phase coherence (%) quantifying the consistency of the climate-dengue timing relationship over the study period. Higher values indicate more reliable timing relationships. Provinces are ordered by geographic region with horizontal lines separating regions. Values are displayed within cells.
Fig 6.
Forest plots showing cumulative relative risk (RR) with 95% confidence intervals for 90th versus 50th percentile exposure to precipitation A) and temperature B).Only provinces with phase coherence >= 0.85 and positive or zero phase lag are included. Point colour indicates the wavelet-derived lag (months) at which cumulative RR was evaluated. Vertical dashed line indicates RR = 1 (no effect).