Renaro Lab · Independent research laboratory · Indonesia

Computational research that shows its work.

Renaro Lab works across artificial intelligence, data science, computational research and interdisciplinary science. We publish analyses together with the choices that produce them, so that readers can judge what a result supports and what it does not.

  1. 01Artificial intelligenceLearning methods and the design choices that shape what a model can be trusted to do.
  2. 02Data sciencePublic, spatial and remotely sensed data, with provenance and missingness made explicit.
  3. 03Computational researchReproducible pipelines, sensitivity analysis and attribution of model outcomes.
  4. 04Interdisciplinary scienceQuestions between disciplines, approached with quantitative methods.

A model is an argument about what matters.

Rankings, indices and predictions encode decisions about data, weights and thresholds. Our work treats those decisions as part of the result: we make them explicit, test how much they matter, and report the limits that follow.

i. Questions first

Define the question before the instrument

Each model answers a specific question. We name that question before choosing data or methods, and we do not let a more complex model pass as a more correct one.

ii. Assumptions in view

Report the choices with the findings

Weights, exclusions, thresholds and conventions are published alongside results, together with sensitivity analyses that show which conclusions survive plausible alternatives.

iii. Proportionate claims

Size each claim to its evidence

Exploratory results are labelled as exploratory. We state what a study does not show as clearly as what it does, and we distinguish preprints from peer-reviewed work.

Current work · Jakarta, 500 m grid
Jakarta study grid: top-decile screening cellsA map of the five mainland cities of Jakarta drawn as 2,765 cells of 500 m. Grey tones show normalised land surface temperature for the 2,536 cells in the common sample; pale outlined cells are excluded. Highlighted cells are the 254 top-decile cells of the selected screening model. Models M1 and M4 share 141 of these cells.
Showing M4: the 254 top cells when vegetation, the count of children and older adults, and modelled park distance are added. The heat-only model (M1) and the composite model (M4) share 141 of 254 top cells. This is a screening comparison, not a map of health risk.

Heat-screening maps are often read as maps of where to act first. This study asks how much those priorities depend on the indicators and weights chosen. Comparing four models for Jakarta, it attributes changes in the ranking to specific modelling decisions and measures their sensitivity to demographic and park-access assumptions.

141/254

top-priority cells shared by the heat-only and composite models (Jaccard 0.384).

85.5%

of rank movement from M3 to M4 attributed to adding the park-distance term.

0.9995

Spearman correlation between the age count and total population, so the age term adds little new information.

Selected publications

Researchers

Researchers affiliated with Renaro Lab. Each profile lists their contact details and the publications they have contributed to at the laboratory.

  • Muhammad Fariq Faqih

    Renaro Lab, Indonesia

    Publications · 1
    From Thermal Hotspots to Spatial Priorities (2026)
    ORCID
    0009-0006-4081-2638
  • Muhammad Alim Mustofa

    Renaro Lab, Indonesia

    Publications · 1
    From Thermal Hotspots to Spatial Priorities (2026)
    ORCID
    0009-0000-0557-7384
  • Mu’adz Abdulloh

    Renaro Lab, Indonesia

    Publications · 1
    From Thermal Hotspots to Spatial Priorities (2026)
    ORCID
    0009-0005-5412-9058
  • Dio Ramadhany

    Renaro Lab, Indonesia

    ORCID
    0009-0003-2047-1184