Introduction
A GIS tool that helps water managers, planners, and researchers identify prime locations for groundwater recharge using floodwaters.
I collaborated with Earth Genome to redesign a Flood-MAR App to help planners, researchers and water managers identify high-potential locations for Managed Aquifer Recharge (MAR).
Under the supervision of Mikel Maron (Humanitarian OpenStreetMap Team cofounder, OpenStreetMap board memeber), I redesigned the analytical workflow to reduce cognitive load, improve map usability, and create a more intuitive exploration experience.
The problem
By the time funding reaches affected communities, the damage is already done.
Recharge IQ wasn't short on capability. It held genuinely powerful analytical depth — the kind of scientific detail that makes a decision-support tool worth using. But the interface exposed nearly all of it at once. Users landed on a homepage, moved into a general map, and from there hover, click, and zoom all surfaced overlapping layers of the same information. There was no ramp-up, no sense of what mattered first.
The result was a tool that asked users to split their attention constantly — between map, filters, and data panels, before they'd even decided what they were looking for. For an exploratory decision-support tool, not a GIS editing platform, that's the wrong kind of friction: it's not protecting precision, it's just adding load.
The question I framed the redesign around:
How might we help technical users explore spatial data more efficiently, without hiding any of the analytical depth that makes the tool valuable?
The Context
A heuristic evaluation surfaced five compounding problems.
I ran a full heuristic evaluation of the existing product before touching any screens. Five findings stood out:
Individually, none of these were dramatic. Together, they meant every session started from cognitive overload instead of orientation.
Design goals
Four goals, in order of priority.
A set of rules were established in order to redesign the platform.
Defining the interaction model
The core fix: giving hover, click, and zoom different jobs.
The existing product treated hover, click, and zoom as near-duplicates — three ways to trigger the same data. I redesigned this into a genuine progressive disclosure model, where each interaction earns the user more detail:
Hover → a temporary preview Click → a pinned summary Double-click → full detailed analysis
General Exploration Mode
Built for fast scanning. A compact panel, regional filters, minimal statistics, and a suitability overview, all while keeping the map dominant. The goal here was speed: let users rapidly scan a country without being interrupted by detail they didn't ask for yet.
Detailed Analysis Mode:
Triggered once a user selects a region. The right panel expands into accordion sections, charts, contextual explanations, and comparison metrics. This is where the scientific depth lives, but only once someone has asked for it
Design exploration
I organized iterations by problem
Rather than presenting a string of full-screen redesigns, I broke exploration into four focused tracks.
Interaction model - prototyping hover, click, and double-click against each other to settle the disclosure logic above.
Data visualization - exploring charts, progress bars, grouped statistics, and suitability comparisons to make dense metrics scannable at a glance.
General vs. detailed modes - comparing small cards, large cards, and dynamic panels to find the right density for each mode.
Right panel hierarchy - testing ways to organize scientific metrics while minimizing scroll depth.
Redesigning the right panel
Before: a flat list. After: a hierarchy.
The original panel was one long, equally-weighted list - everything scrolled, and nothing signaled what mattered most. The redesign grouped information deliberately: summary first, details second, with progressive expansion via accordion sections instead of one continuous scroll. Typography and spacing did the rest of the work, establishing a clear read order instead of a wall of metrics.
Color system
Color had a job to do beyond looking good.
I moved from the original palette through several explored options to a final system, and after stakeholder feedback aligned the selected-intervention color directly with the cartographic palette, so a highlighted region on the map and its corresponding chart data read as the same thing at a glance.
Regional filters
Designing for how people actually think about geography.
Users approached the tool geographically before they approached it analytically - they thought in terms of "the Southwest" or "the Pacific Northwest," not in terms of individual filterable regions. So instead of filtering one region at a time, I proposed grouping regions (California, Great Basin, Pacific Northwest, Southwest, etc.).
The flow became: select a region group → map highlights the group → user explores → clicks an individual region → drops into detailed analysis. It's a small structural change, but it matches the tool to the user's mental model instead of the database's.
Challenges
Honest about what didn't work at first.
The biggest lesson from this project wasn't about panels or color - it was about how to present exploration itself. Early on, I shared several visually distinct prototypes side by side. Each one tested a different interaction pattern, but because multiple variables changed between them at once, stakeholders struggled to compare them meaningfully - was a reaction to the color, the layout, or the interaction logic?
After that feedback, I changed approach: instead of full competing redesigns, I iterated on one variable at a time - panel hierarchy in one round, interaction dynamics in the next, color last. Discussions got sharper, and stakeholders could actually evaluate individual decisions instead of reacting to whole concepts.
Outcome
A clearer model for exploration
The redesign establishes an interaction model centered on progressive disclosure: users move from fast, high-level scanning to detailed regional analysis, with the map staying the primary workspace throughout rather than competing with it. The proposed direction simplifies navigation, strengthens visual hierarchy, and gives the product a more scalable foundation as new analytical features are added.
It's worth being clear-eyed here: this is a design proposal, not a shipped, tested result. The real test — whether the hover/click/double-click model actually feels intuitive, and whether regional grouping matches how users search in practice — is still ahead
Reflection
The hard part wasn't the interface, it was learning how to show my thinking.
Solving the disclosure model mattered. But the bigger shift was learning that showing four different "final" concepts at once isn't rigor, it's noise — and that isolating one variable per round of exploration is what actually lets people evaluate a decision on its merits. Project contentPortfolioCreated by youOPAL-case-study (2).md186 linesmdskills-case study.md31 linesmdpdf