📄 earth_science

Diagnosing Spatial Disparities in Urban Flood Burden through Hydro- Environmental Normalization in Jakarta, Indonesia

This study introduces the Hazard-Normalized Flood Burden Deviation (HNFBDev) framework to diagnose spatial disparities in Jakarta's urban flood burden by quantifying the deviation between observed and environmentally expected outcomes, thereby identifying locations where flood impacts are unusually high or low relative to hydro-environmental conditions.

Original authors: Rifky Husaini, Perdinan Perdinan, Novindra Novindra

Published 2026-08-11

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Rifky Husaini, Perdinan Perdinan, Novindra Novindra

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Great Flood Detective Game

Imagine you are trying to figure out why some neighborhoods get soaked in a rainstorm while others stay dry. In the world of science, this is called "urban flood assessment." Usually, scientists look at two main things: the hazard (how much rain fell or how high the river got) and the exposure (how many people live there). It's like checking the size of a wave and counting the number of swimmers. If the wave is huge and the crowd is big, we assume the disaster is massive.

But here is the tricky part: sometimes a huge wave hits a place with very few people, and the "disaster" feels small. Other times, a medium-sized wave hits a crowded neighborhood, and the chaos is huge. This paper asks a clever question: Is the actual mess we see bigger or smaller than what we would expect, given the weather and the terrain? To answer this, the researchers use a method called "observed-expected diagnostics." Think of it like a teacher grading a test. If a student gets a 90% on a super-hard exam, that's amazing. If they get a 90% on a test where everyone else got a 99%, maybe they didn't study enough. This paper tries to grade Jakarta's neighborhoods not just on how wet they got, but on how wet they got compared to what the weather and ground conditions said they should get.

The Story of Jakarta's Flood Detective

In the bustling, low-lying city of Jakarta, Indonesia, floods are a complicated mix of heavy rain, rising sea levels, sinking ground, and crowded streets. The researchers, Rifky Husaini, Perdinan, and Novindra, wanted to solve a mystery: Why do some neighborhoods suffer more than others, even when the weather looks similar? They didn't just want to map where the water went; they wanted to know if the burden (the trouble caused by the water hitting people) was unusually high or low.

To do this, they created a new tool called HNFBDev (Hazard-Normalized Flood Burden Deviation). It sounds like a robot name, but it's actually a simple "Expectation vs. Reality" checker. Here is how they played the game:

Step 1: The "Expected" Score
First, they built a digital model of Jakarta's "flood personality." They looked at the rain, the slope of the land, how close the neighborhood was to the ocean, how much the ground was sinking, and how many buildings were there. Based on these ingredients, they calculated what the flood burden should be. If a neighborhood is right on the coast and the ground is sinking, the model says, "You should expect a lot of trouble." If a neighborhood is high up and far from the river, the model says, "You should expect very little trouble."

Step 2: The "Observed" Score
Next, they looked at what actually happened. Using satellite eyes (specifically Sentinel-1 radar, which can see through clouds), they tracked major floods between 2020 and 2024. They didn't just count the water; they counted the people standing in the water. They combined the satellite pictures of flooded areas with maps of where people live to create a "burden score."

Step 3: The Big Reveal (The Deviation)
Finally, they subtracted the "Expected" score from the "Observed" score.

Positive Score: The neighborhood got more trouble than the weather and ground predicted.
Negative Score: The neighborhood got less trouble than predicted.
Zero Score: The neighborhood got exactly what was expected.

What They Found

The results were fascinating and showed that the story of Jakarta's floods is more complex than just "rain + people = disaster."

The Weather Explains Some, But Not All
The researchers found that the "flood personality" (the hydro-environmental forcing) did a decent job of predicting the burden. It explained about 22.8% of the differences between neighborhoods. This means that while the rain and the sinking ground are important, they don't tell the whole story. There is a lot of mystery left over.

The "High-High" Hotspots
The map of "Positive Scores" (places that got more trouble than expected) wasn't random. It formed big clusters, especially in the northern and central parts of Jakarta. The neighborhoods of Tangki, Kota Bambu Utara, Jembatan Besi, Duri Selatan, and Duri Utara had the highest "deviation" scores, ranging from 3.56 to 3.83. These places were getting hit much harder than the weather and terrain alone would suggest.

The "Low-Low" Cool Zones
Conversely, some areas in the south and west had "Negative Scores." These neighborhoods got less trouble than expected. This could mean they have great drainage, better protection, or perhaps the satellites missed some of the water there.

The Pattern is Real
The researchers checked if these patterns were just a fluke of their math. They tried changing the weights of their ingredients (giving rain more importance, or giving ground-sinking more importance) and the results stayed almost the same. The correlation was incredibly high (0.990 with a standard weighting and 0.970 with equal weighting). This suggests the "extra trouble" or "extra luck" in these neighborhoods is a real, physical thing, not just a mistake in the math.

What This Means (And What It Doesn't)

The authors are very careful to tell us what this tool is not. They explicitly state that a high score does not mean the local government is bad, the infrastructure is broken, or the people are vulnerable. It also doesn't tell us why the burden is high. It just flags the neighborhood as "suspicious."

Think of HNFBDev as a smoke alarm. If the alarm goes off, it doesn't tell you if there is a fire, a burnt piece of toast, or a steamy shower. It just tells you, "Hey, something is happening here that doesn't match the usual rules."

The paper suggests that these "High-High" clusters are places where we need to look closer. Maybe there is a hidden drainage bottleneck, a blocked river, or a specific type of building layout that traps water. The "Low-Low" areas might be places where local defenses are working well, or maybe the satellites just didn't see the water.

The Takeaway

This study gives us a new way to look at flood maps. Instead of just saying "This area is wet," we can now say, "This area is wetter than it should be, given the rain and the ground." By using this "observed-expected" detective work, city planners can stop guessing and start investigating the specific neighborhoods where the flood burden is acting strangely. It's a tool for finding the clues, not for solving the whole mystery all at once. The authors emphasize that this is a screening tool to help prioritize where to send the real experts to investigate, ensuring that resources go to the places where the flood burden is truly out of the ordinary.

Technical Summary: Diagnosing Spatial Disparities in Urban Flood Burden through Hydro-Environmental Normalization in Jakarta, Indonesia

Problem Statement
Urban flood assessment traditionally focuses on identifying locations with high hazards, exposure, or vulnerability. However, existing frameworks often fail to distinguish whether a realized flood burden is unusually high or low relative to the specific hydro-environmental conditions under which it occurs. In complex coastal and deltaic cities like Jakarta, flood outcomes result from interacting drivers (pluvial, fluvial, coastal, groundwater, and land-surface processes) rather than single mechanisms. Consequently, a high observed burden in a highly flood-prone area may be consistent with environmental expectations, while a moderate burden in a low-risk area might indicate a failure of protection or a data gap. Current methods lack a geospatial diagnostic that explicitly separates represented hydro-environmental forcing from realized urban flood burden to identify spatial disparities that warrant further investigation.

Methodology
The study develops and applies the Hazard-Normalized Flood Burden Deviation (HNFBDev), an observed–expected geospatial diagnostic framework. The analysis was conducted at the kelurahan (urban village) level across 261 administrative units in Jakarta, Indonesia, covering the period 2020–2024.

Hydro-Environmental Forcing Baseline: A compound forcing surface was constructed to represent the environmental conditions under which flood burden is expected. This baseline integrated inland factors (maximum daily precipitation, subsidence-adjusted elevation, slope, proximity to surface water, built-up intensity, imperviousness) and coastal factors (subsidence-adjusted elevation, coastline proximity, river/estuarine connectivity). Indicators were normalized to a 0–1 scale and integrated using a fuzzy algebraic sum to create a compound forcing score.
Population-Exposure-Weighted Observed Burden: Using Sentinel-1 SAR imagery, flood extents for selected major events (2020–2024) were mapped. These extents were intersected with gridded population data (WorldPop) to calculate an annual, population-exposure-weighted flood-burden proxy for each kelurahan. This metric represents the intersection of inundation and exposed population, rather than direct economic loss or flood depth.
Environmental Expectation Modeling: An environmental expectation model was developed to estimate the burden anticipated under the represented hydro-environmental forcing. Given the non-negative and right-skewed nature of the data, three model specifications (Tweedie GLM, log-transformed OLS, and two-part hurdle models) were evaluated. The final model (selected based on adequacy and stability) estimated the expected burden ( $\hat{B}_{it}$ ) for each unit.
HNFBDev Calculation: The deviation between observed burden ( $B_{it}$ ) and expected burden was calculated as the annual residual ( $e_{it}$ ). These residuals were aggregated by taking the median across the study period for each kelurahan. The final HNFBDev value was derived by standardizing these median residuals ( $HNFBDev_i = \frac{\tilde{e}_i - \mu_{\tilde{e}}}{\sigma_{\tilde{e}}}$ ).
Spatial Diagnostics and Robustness: Global and Local Moran's I statistics were used to test for spatial autocorrelation and identify clusters (High-High, Low-Low) and outliers. Robustness was assessed by comparing results under literature-informed weighting versus equal weighting of forcing components, and by evaluating alternative flood-mask assumptions.

Key Results

Forcing vs. Burden: The compound hydro-environmental forcing showed a statistically significant positive relationship with the population-exposure-weighted burden proxy. However, the forcing model explained only 22.8% of the variation in observed burden, indicating that the hydro-environmental baseline provides a reference but does not fully determine realized outcomes.
Spatial Disparities: The HNFBDev surface revealed substantial intra-urban heterogeneity. Positive deviations (higher-than-expected burden) were concentrated in northern and central Jakarta, while negative deviations (lower-than-expected burden) appeared more frequently in southern and western areas.
Specific Hotspots: The highest positive deviations (HNFBDev values ranging from 3.56 to 3.83) were identified in kelurahan such as Tangki, Kota Bambu Utara, Jembatan Besi, Duri Selatan, and Duri Utara.
Spatial Organization: HNFBDev values exhibited significant spatial autocorrelation (Global Moran's I = 0.355, p = 0.001). Local analysis identified High-High clusters (29 units) where neighboring areas consistently experienced higher-than-expected burden, and Low-Low clusters (44 units) where burden was lower than expected. Spatial outliers were less common.
Robustness: The HNFBDev rankings and spatial patterns were highly stable under alternative analytical assumptions, showing a Pearson correlation of 0.990 between literature-informed and equal-weight forcing schemes.

Significance and Claims
The paper positions HNFBDev as a transferable observed–expected geospatial diagnostic designed to complement, not replace, existing hydrodynamic models, vulnerability assessments, or infrastructure audits.

Diagnostic Utility: The primary contribution is the ability to identify locations where flood outcomes are disproportionately high or low relative to the environmental setting. This allows planners to distinguish between areas where high burden is "expected" due to severe forcing and areas where high burden is "unexpected," suggesting potential drainage bottlenecks, infrastructure underperformance, or unmeasured exposure factors.
Screening Tool: The authors explicitly state that HNFBDev is a screening and prioritization tool. It identifies where further hydraulic, infrastructure, institutional, and local verification is warranted.
Interpretation Boundaries: The study maintains a modest scope regarding its claims. HNFBDev is not a direct measure of:
- Adaptation performance or resilience.
- Governance quality or institutional effectiveness.
- Infrastructure effectiveness.
- Social vulnerability, flood depth, or causal mechanisms.
- Direct economic loss.

The authors caution that spatial clustering of deviations does not automatically validate specific mechanisms; rather, it highlights areas where the "observed–expected" gap requires multi-source triangulation (e.g., field verification, drainage network analysis) to understand the underlying causes. The framework is presented as a step toward evidence-based prioritization in urban climate resilience planning, particularly for data-constrained environments where detailed hydrodynamic or infrastructure data may be incomplete.