Preprint · v1Not peer-reviewed

Supplementary material for

From Thermal Hotspots to Spatial Priorities: Attribution and Sensitivity of Urban Heat Screening in Jakarta

Muhammad Fariq Faqih1, Muhammad Alim Mustofa1, Mu’adz Abdulloh1

1Renaro Lab, Indonesia · Correspondence: [email protected]

This page is an HTML rendering of preprint version 1, produced from the same source as the PDF. The PDF is the reference version; if the two differ, the PDF takes precedence.


Supplementary material

S1. Data quality and missingness

Table S1 reports the quality flags and missingness counts used to define the common comparison sample. The counts are cell counts unless stated otherwise. A missing value is not a zero: route-missing cells are excluded from the common set, and cells without demographic source support retain analytical nulls while their original exported zeros remain in the raw columns.

Table S1. Data-quality and missingness register for the 2,765-cell master grid.
ItemCountMeaning
LST / NDVI missing0All master cells have delivered thermal and vegetation values
Population missing6No valid WorldPop support; analytical null, raw zero preserved
Child / elderly missing6Same six cells; no age support
Park distance missing (baseline)226No route to a candidate park; excluded, not zero
Park distance missing (conservative)407Explicit node/connector restrictions applied
Grid snap to network >250 m66QA flag; retained in baseline, removed in one sensitivity
Coverage <10% of nominal cell65Partial coastal/boundary cells; retained and flagged
Population exactly zero0No supported zero-population cell in the common sample
Children + elderly >105% of population0No age-population inconsistency flagged
LST valid area <50%1JKT_R037_C062; excluded from all models
Grids excluded from common comparison2298.28% of master grid
Common comparison sample2,53691.72% of master grid
Spatial observations after island removal2,535JKT_R013_C019 is a Queen island

S2. Scenario-family sensitivity at three budgets

Table S2 reports the minimum and maximum Jaccard overlap of the M4 top-priority set with the baseline at the 5%, 10% and 20% budgets. Values are from the eligibility-changing comparison; the fixed-anchor comparison is reported in the main text and in the artifact tables. A family with a single scenario has identical minimum and maximum.

Table S2. Scenario-family sensitivity at three top-priority budgets. Each cell is the minimum–maximum Jaccard overlap with the baseline M4 set across scenarios in that family.
Family5% (127)10% (254)20% (508)
MinMaxMinMaxMinMax
Aggregation0.8960.8960.8810.8810.9130.913
Park threshold0.4940.6490.5210.6820.4470.782
Weights0.8810.9540.8680.9540.9170.961
QA exclusions0.9690.9840.9690.9960.9750.994
Missing park distance0.9060.9200.9060.9060.9030.913
Joint factorial (16)0.5300.6820.5830.7280.6680.789
Missing support1.0001.0001.0001.0001.0001.000
Normalisation0.7280.8410.7340.8140.7250.814
Population definition0.2700.9540.3470.9770.4810.996
Route policy0.4191.0000.4201.0000.5281.000
Thermal support1.0001.0001.0001.0000.9921.000
Land cover0.6820.6820.7820.7820.7740.774

S3. City composition of the top deciles

Table S3 reports the number and share of top-decile cells by city for M1–M4, together with the modelled population in the M4 top decile. City assignment follows the majority-area city of each clipped grid cell; it is not a separate raster zonal total within administrative boundaries.

Table S3. Top-decile composition by city. Shares are percentages of each model’s 254-cell top set. Population is the modelled 2024 population in the M4 top decile.
CityM1 nM1 %M2 nM3 nM4 nM4 %M4 population
West Jakarta3915.460626324.8389,483
Central Jakarta20.81425176.7124,689
South Jakarta4818.931273212.6172,959
East Jakarta8834.662738734.3526,819
North Jakarta7730.387675521.7325,841
Total254100.0254254254100.01,539,791

S4. Alternative LISA summaries

Table S4 reports cluster counts for each model under the historical directed/999/BH configuration, retained for comparison. Table S5 repeats the M4 convention sensitivity from the main text; the adopted two-sided/9,999/BH specification is the first row. The counts are labels, not causal findings; the same map can yield different counts under a different tail, permutation count or correction. The per-cell primary labels are in Supplementary_Tables/lisa_m4_primary.csv.

Table S4. LISA cluster counts for M1–M4 under the historical directed/999/BH configuration, retained for comparison. “ns” means not significant at that convention.
ModelHHLLHLLHns
M143336012131,717
M24373751691,698
M34423241151,753
M431932311101,872
Table S5. M4 LISA convention sensitivity. “Changed” counts cells whose significance label differs from the adopted two-sided/9,999/BH convention.
TailPermutationsCorrectionSignificantChanged
Two-sided9,999BH4930
Two-sided999BH49665
Directed999BH663170
Directed9,999BH687194
Two-sided9,999BY234259
Directed9,999BY284209
Directed999BY0493

S5. Exclusion geography and score distributions

Figure S1 maps the 229 excluded cells by primary exclusion reason. Reasons overlap: a cell may be route-missing and also lack demographic support. The primary reason is shown in the order low LST, population, age, route. The route-missing category therefore contains most of the age-unsupported cells that are also unroutable. Figure S2 shows the score distributions across the common sample. Figure S3 shows the candidate-park inventory by evidence state and the route-missing grid.

Figure S1: Excluded cells by primary reason. Route-missing cells dominate the exclusion pattern and are concentrated in the north and east. The map is a disclosure of sample selection, not a statement that excluded areas are unimpo…
Fig. S1Excluded cells by primary reason. Route-missing cells dominate the exclusion pattern and are concentrated in the north and east. The map is a disclosure of sample selection, not a statement that excluded areas are unimportant. Source: eligibility_by_model.csv and the 500 m grid.
Figure S2: Score distributions for M1–M4 on the common 2,536-cell sample. Percentile clipping compresses the upper tail and creates ties, especially in M1. The distributions are descriptive; they are not evidence that one model is …
Fig. S2Score distributions for M1–M4 on the common 2,536-cell sample. Percentile clipping compresses the upper tail and creates ties, especially in M1. The distributions are descriptive; they are not evidence that one model is more accurate than another.
Figure S3: Candidate parks by evidence state (representative points) and route-missing grid cells. Grey points are unknown access, orange points are conditional/unverified, and green points carry a positive OSM tag that remains unv…
Fig. S3Candidate parks by evidence state (representative points) and route-missing grid cells. Grey points are unknown access, orange points are conditional/unverified, and green points carry a positive OSM tag that remains unverified. The map shows why the candidate inventory is a proxy, not a public park census.

S6. Full scenario results

Table S6 lists all 49 executed scenario top-decile comparisons at the 10% budget. Jaccard and shared cells are relative to the M4 top-254 set. The machine-readable table is Supplementary_Tables/full_scenario_results.csv. The three registry entries that are baselines, duplicates or could not be executed are not presented as separate scenarios; the family summary is in S2.

Table S6. Executed scenario top-decile results at the 10% budget. Jaccard and shared cells are relative to the M4 top-254 set.
FamilyScenarioJaccardShared/254
AggregationUniform fractional allocation0.881238
Park thresholdBinary park threshold >1,000 m0.521174
Park thresholdBinary park threshold >500 m0.530176
Park thresholdPark-deficit cap 1,500 m0.655201
Park thresholdPark-deficit cap 750 m0.682206
WeightsEqual weights0.868236
WeightsLST weight x0.80.917243
WeightsLST weight x1.20.917243
WeightsNDVI-deficit weight x0.80.939246
WeightsNDVI-deficit weight x1.20.903241
WeightsPark-deficit weight x0.80.881238
WeightsPark-deficit weight x1.20.924244
WeightsVulnerable-count weight x0.80.946247
WeightsVulnerable-count weight x1.20.954248
QA exclusionExclude coverage <10%0.969248
QA exclusionExclude remote-sensing QA failures0.996253
QA exclusionExclude snap >250 m0.973248
Missing park distanceMissing distance = maximum deficit0.906252
Missing park distanceMissing distance = observed median0.906252
Joint factorialTotal population, cap 1,500 m, clipped, fractional0.649200
Joint factorialTotal population, cap 1,500 m, clipped, hybrid0.655201
Joint factorialTotal population, cap 1,500 m, rank, fractional0.710211
Joint factorialTotal population, cap 1,500 m, rank, hybrid0.728214
Joint factorialTotal population, cap 750 m, clipped, fractional0.682206
Joint factorialTotal population, cap 750 m, clipped, hybrid0.677205
Joint factorialTotal population, cap 750 m, rank, fractional0.608192
Joint factorialTotal population, cap 750 m, rank, hybrid0.583187
Joint factorialVulnerable count, cap 1,500 m, clipped, fractional0.649200
Joint factorialVulnerable count, cap 1,500 m, clipped, hybrid0.655201
Joint factorialVulnerable count, cap 1,500 m, rank, fractional0.710211
Joint factorialVulnerable count, cap 1,500 m, rank, hybrid0.722213
Joint factorialVulnerable count, cap 750 m, clipped, fractional0.677205
Joint factorialVulnerable count, cap 750 m, clipped, hybrid0.682206
Joint factorialVulnerable count, cap 750 m, rank, fractional0.613193
Joint factorialVulnerable count, cap 750 m, rank, hybrid0.592189
Missing supportCorrected demographic support1.000254
Normalisation1st-99th percentile normalisation0.814228
NormalisationPercentile-rank normalisation0.782223
NormalisationPure min-max normalisation0.734215
Population definitionTotal population (not vulnerable count)0.977251
Population definitionVulnerable share (not count)0.347131
Route policyRoute policy: baseline1.000254
Route policyRoute policy: conservative nodes0.842224
Route policyRoute policy: conservative nodes + connectors0.842224
Route policyRoute policy: full extent1.000254
Route policyRoute policy: positive tags0.420150
Thermal supportThermal support 25%1.000254
Thermal supportThermal support 50%1.000254
Land coverTree-probability land cover (NDVI replacement)0.782223

S7. Demographic source support

Table S7 lists the six cells that remain without total-population and age-raster support after the buffered-age remediation. The machine-readable table is Supplementary_Tables/demographic_no_support.csv. Analytical values are null; raw exported zeros are retained in the *_raw fields and are not treated as estimated population.

Table S7. Cells without demographic source support. Analytical values are null; the raw exported zeros are retained but are not treated as estimated population.
Grid IDCityArea (km2)CoverageSupport state
JKT_R003_C031Selatan0.0002660.0011no source support
JKT_R057_C063Utara0.0026430.0106no source support
JKT_R061_C009Utara0.2444580.9778no source support
JKT_R061_C010Utara0.0986870.3947no source support
JKT_R061_C011Utara0.0019810.0079no source support
JKT_R062_C043Utara0.0067360.0269no source support

All six cells have no_source_support for total population and both age groups. Their raw exported zeros are preserved in the *_raw fields; analytical values are null. The six cells are excluded from the M4 common comparison. JKT_R050_C061 is a different case: a supported near-zero estimate, not a no-support cell.

S8. Eligibility reconciliation

Table S8 reconciles the 2,765 master cells with the 2,538-cell benchmark and the 2,536-cell common set. The machine-readable table is Supplementary_Tables/eligibility_reconciliation.csv.

Table S8. Eligibility reconciliation from the 2,765-cell master grid to the 2,536-cell common comparison set.
CategoryCellsNote
Master grid cells2765Five mainland cities; 500 m lattice
Route-only missing222No path to a candidate park in baseline graph
Route + population + age missing4Already outside common set because of route
Population/age missing only2JKT_R003_C031; JKT_R057_C063
Low LST support1JKT_R037_C062; 27.759% valid LST area
Excluded union2298.28% of master grid
Benchmark common set25382,765 - 222 - 4 - 1
Common set (2,536)25362,538 - 2 population-only slivers
Spatial observations after island2535JKT_R013_C019 is a Queen island

S9. Figure and table provenance

Table S9. Provenance of figures and tables. All source files are in the processed analysis dataset; exact commits and hashes are recorded in the reproducibility note.
ItemSourceNote
Main Fig. 1merged_grid.csv, priority_models.csv, grid_500m_jakarta.gpkgGenerated from processed analysis tables
Main Fig. 2Same as Fig. 1Population and park distance
Main Fig. 3Same as Fig. 1M1/M4 top flags
Main Fig. 4Same as Fig. 1Rank displacement
Main Fig. 5rank_change_attribution.csv, robustness_family_summary.csvAttribution and family envelope
Main Fig. 6lisa_m4_primary.csv, lisa_tail_permutation_sensitivity.csvLISA map and counts
Tables 1–11Processed analysis tables listed in the main textValues checked against the analysis dataset
S1data_quality.csv, demographic_support.csvQuality and missingness
S2robustness_family_summary.csvThree budgets
S3kota_summary.csvCity composition
S4–S5lisa_cluster_counts.csv, lisa_tail_permutation_sensitivity.csv, lisa_m4_primary.csvLISA counts and primary per-cell labels
S6Supplementary_Tables/full_scenario_results.csv49 executed scenario comparisons
S7Supplementary_Tables/demographic_no_support.csvSix unsupported demographic cells
S8Supplementary_Tables/eligibility_reconciliation.csvMaster-to-common eligibility
S9This fileProvenance register

Author Contributions

All authors contributed to the preparation and critical revision of the manuscript. All authors reviewed and approved the final version and accept responsibility for its content.

Funding

This research received no external funding.

Conflicts of Interest

The authors declare no competing interests.

Use of AI-Assisted Tools

AI-assisted tools were used for language editing, LaTeX production and consistency checking during the preparation of this manuscript. The authors reviewed and approved all scientific content and accept responsibility for the final article.

License

Copyright © 2026 The Authors. This work is licensed under a Creative Commons Attribution 4.0 International License (CC BY 4.0, https://creativecommons.org/licenses/by/4.0/). The licence applies to the author-created supplementary narrative, tables and original figures. Third-party datasets, software and the LaTeX template remain under their own licences and attribution requirements; the template source remains under GPL-3.0-or-later.

How to cite

APA

Faqih, M. F., Mustofa, M. A., & Abdulloh, M. (2026). From thermal hotspots to spatial priorities: Attribution and sensitivity of urban heat screening in Jakarta [Preprint, version 1; not peer-reviewed]. Renaro Lab.

BibTeX

@misc{faqih2026from,
  title  = {From Thermal Hotspots to Spatial Priorities: Attribution and Sensitivity of Urban Heat Screening in Jakarta},
  author = {Faqih, Muhammad Fariq and Mustofa, Muhammad Alim and Abdulloh, Mu’adz},
  year   = {2026},
  note   = {Preprint, version 1. Not peer-reviewed.},
  doi    = {10.5281/zenodo.24414265},
  url    = {https://renaro.my.id/publications/jakarta-heat-priorities}
}

Licence: CC BY 4.0. Third-party data and software keep their own licences.