{
"$type": "site.standard.document",
"bskyPostRef": {
"cid": "bafyreif3g4ifwrdgc77n4jd2hejialskk3uxjvbqj66tovcy44jok4fsxu",
"commit": {
"cid": "bafyreidsgulgeiviz5c5p6264eqbjv6z6z24uuroxyij2ggj33qk3vvrbe",
"rev": "3mnssylyzy72l"
},
"uri": "at://did:plc:k2h7zvrve2cfizg3q5ig7z37/app.bsky.feed.post/3mnssyluhzs2z",
"validationStatus": "valid"
},
"content": {
"$type": "pub.leaflet.content",
"pages": [
{
"$type": "pub.leaflet.pages.linearDocument",
"blocks": [
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "Europe already has the flood data. The maps exist. The models are good enough. And yet, every year, floods surprise people they shouldn't surprise.",
"textSize": "large"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "The problem isn't model accuracy. It's the last mile — the distance between a number on a map and a person taking action. That kilometre is almost entirely unaddressed.",
"textSize": "large"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.horizontalRule"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "I want to start with a confession."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "I've spent years building flood models. I know how to tune a HEC-RAS simulation. I know how to validate against gauge records, how to argue about Manning's n coefficients, and how to fight over which DEM resolution is \"good enough.\""
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "And I'm here to argue that none of that is the problem we should be solving."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "Right now, in Europe, there are flood hazard maps covering hundreds of thousands of kilometres of coastline and river systems. Produced under the EU Floods Directive. Updated. Peer-reviewed. Publicly available. "
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#italic"
}
],
"index": {
"byteEnd": 29,
"byteStart": 24
}
}
],
"plaintext": "Good enough to tell you today which properties are at serious risk."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "Nobody is using them."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "Not the homeowner who just bought a house in a floodplain because the estate agent didn't mention it. Not the farmer who will lose next season's harvest to a flood that was mapped, modelled, and predicted — and never communicated."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "The data exists. The knowledge exists. And yet, every year, floods surprise people they shouldn't surprise."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "So why are we still talking about building better models?"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.horizontalRule"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.header",
"level": 2,
"plaintext": "The seduction of precision"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.image",
"aspectRatio": {
"height": 1544,
"width": 2720
},
"fullBleed": true,
"image": {
"$type": "blob",
"ref": {
"$link": "bafkreibga57luckxn7feliaqjfmhyju4so2huiyivbwstptt7bwnwgvmj4"
},
"mimeType": "image/webp",
"size": 198744
}
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "Look at how commercial flood data providers market their products. The language is striking in its consistency. One promises tools to simulate and analyse flood events with \"unmatched precision and confidence.\" Another advertises \"market-leading, high-resolution flood maps and catastrophe models\" as \"best-in-class.\" Another headlines a product launch with \"a new type of flood inundation model leads to unprecedented accuracy.\""
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "Unmatched. Market-leading. Unprecedented. Most accurate. The vocabulary is nearly identical across competitors who are otherwise fighting for the same clients. Accuracy is not just a feature — it is the entire value proposition. What you will do with that accuracy once you have it is, apparently, your problem."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "There is something deeply seductive about this framing. As engineers and scientists, we are trained to believe that better data leads to better decisions. It's almost a moral instinct. More decimal places means more rigour. Higher resolution means closer to the truth."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#highlight",
"color": {
"$type": "pub.leaflet.theme.color#rgb",
"b": 195,
"g": 205,
"r": 255
}
}
],
"index": {
"byteEnd": 78,
"byteStart": 45
}
}
],
"plaintext": "But here's the question we almost never ask: better for whom, and by how much?"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "Consider a concrete example. You have a flood model with a 5-metre spatial resolution. A property is predicted to flood to 0.4 metres depth. You invest two years and significant funding improving the model — new lidar, better bathymetry, improved friction parameterisation. Now the model says 0.37 metres."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "You can see it on the picture. The thin red line is your improved prediction."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.image",
"aspectRatio": {
"height": 1530,
"width": 2720
},
"fullBleed": true,
"image": {
"$type": "blob",
"ref": {
"$link": "bafkreiapl2atqb3i3dt2hs5hdutgvhuchjsct7mj4ogsuwov66p5igvkoe"
},
"mimeType": "image/webp",
"size": 943246
}
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#italic"
}
],
"index": {
"byteEnd": 25,
"byteStart": 0
}
}
],
"plaintext": "Did you just save anyone?"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "The answer depends entirely on what decision that number is attached to. If the decision is \"do we evacuate this neighbourhood,\" the answer is no. The decision is the same. The action is the same. You moved a decimal point."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#footnote",
"contentFacets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#bold"
}
],
"index": {
"byteEnd": 126,
"byteStart": 18
}
},
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#italic"
},
{
"$type": "pub.leaflet.richtext.facet#link",
"uri": "https://doi.org/10.48550/arXiv.2509.04772"
}
],
"index": {
"byteEnd": 169,
"byteStart": 128
}
}
],
"contentPlaintext": "Liu et al. (2025) FloodVision: Urban Flood Depth Estimation Using Foundation Vision-Language Models and Domain Knowledge Graph https://doi.org/10.48550/arXiv.2509.04772",
"footnoteId": "019ea97a-36ff-7886-88f1-573b04450911"
}
],
"index": {
"byteEnd": 100,
"byteStart": 99
}
}
],
"plaintext": "The research backs this up. A 2025 study combining vision-language AI with a domain knowledge graph* achieved a mean absolute error of 8.17 cm in flood depth estimation, reducing a GPT-4o baseline of 10.28 cm by 20.5%. That is a genuine technical achievement. It is also a 2 cm improvement — a difference that changes no operational decision a homeowner, farmer, or emergency manager will ever make."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "The big accuracy leap — from multi-metre DEM errors down to sub-metre — already happened, when LiDAR-derived elevation data replaced conventional survey products. That transition reduced mean absolute error in flood depth from roughly 57 cm to under 25 cm. Most European operational flood models have already made that jump."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "What remains is incremental. And incremental accuracy improvements, in most real-world contexts, do not change decisions."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "What does change decisions is whether people know their risk at all. "
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "That gap — between 0.4 metres and zero awareness — is the gap worth closing. "
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#bold"
},
{
"$type": "pub.leaflet.richtext.facet#italic"
}
],
"index": {
"byteEnd": 29,
"byteStart": 15
}
}
],
"plaintext": "It has a name: the last mile."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.horizontalRule"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.header",
"level": 2,
"plaintext": "What the last mile actually looks like"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.image",
"aspectRatio": {
"height": 1530,
"width": 2720
},
"fullBleed": true,
"image": {
"$type": "blob",
"ref": {
"$link": "bafkreie4x4rcw2rpibhwpiwww2z6niegsls2rkflrzgok2sxgw5lquurou"
},
"mimeType": "image/webp",
"size": 292188
}
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "Open a public flood risk portal. What you find is a web map with layers you can toggle, a legend with return periods, colour gradients representing depth and probability, maybe a PDF report you can download."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "All technically correct. All essentially useless to the person trying to decide whether to buy flood insurance or move their car before Thursday's storm."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#footnote",
"contentFacets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#bold"
}
],
"index": {
"byteEnd": 84,
"byteStart": 21
}
}
],
"contentPlaintext": "Meyer et al. (2012) Recommendations for the user-specific enhancement of flood maps https://doi.org/10.5194/nhess-12-1701-2012, 2012",
"footnoteId": "019ea975-d293-7886-88ef-cbccf62ade67"
}
],
"index": {
"byteEnd": 44,
"byteStart": 43
}
}
],
"plaintext": "A 2012 study from the EU's RISK MAP project* found that flood maps are frequently seen as information tools rather than communication tools — that local knowledge isn't incorporated, that contents don't match end-user needs, and that maps are visualised in ways that residents at risk cannot easily understand. That finding is now more than a decade old. The situation has not materially changed."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "The last mile problem is not a data problem. \nIt is a translation problem — and translation operates at three levels."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.unorderedList",
"children": [
{
"$type": "pub.leaflet.blocks.unorderedList#listItem",
"content": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#bold"
}
],
"index": {
"byteEnd": 20,
"byteStart": 0
}
}
],
"plaintext": "Language translation is the most obvious. \"1-in-100-year flood event\" means almost nothing to most people. \"Your street flooded in 1995, 2010, and 2021 — and the risk is increasing\" means everything. The underlying information is the same. The frame is entirely different."
}
},
{
"$type": "pub.leaflet.blocks.unorderedList#listItem",
"content": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#bold"
}
],
"index": {
"byteEnd": 19,
"byteStart": 0
}
}
],
"plaintext": "Context translation is subtler. Flood risk data is almost always presented as a universal map. But risk is experienced as an individual reality. The question isn't \"is this area at risk?\" It's \"is my house at risk? My parents' house? The road I take to school?\" Tools that answer the individual question — rather than the abstract one — are the tools that drive action."
}
},
{
"$type": "pub.leaflet.blocks.unorderedList#listItem",
"content": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#bold"
}
],
"index": {
"byteEnd": 20,
"byteStart": 0
}
}
],
"plaintext": "Decision translation is the hardest and most neglected. Data presented without a connected decision is just information. What do I do with the fact that my property has a 2% annual flood probability? Do I call my insurer? Raise my floor? Buy sandbags? The translation from data to decision is where emergency management lives — and it is almost entirely absent from the tools we build."
}
}
]
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "The last mile is the distance between a number on a map and a person taking action. It is the hardest kilometre in flood risk management."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.horizontalRule"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.header",
"level": 2,
"plaintext": "A framework for last-mile tools"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "What does a last-mile flood risk tool actually look like? Three principles define the approach."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#bold"
}
],
"index": {
"byteEnd": 27,
"byteStart": 0
}
}
],
"plaintext": "Translation over precision. ",
"textSize": "large"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "An interface should be judged not on the accuracy of the underlying model, but on how well it converts that model into something a non-specialist can act on. Plain language risk statements. Local, historical framing. Complexity hidden without losing substance. A good last-mile tool is not a data portal — it is an interpreter."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#bold"
}
],
"index": {
"byteEnd": 26,
"byteStart": 0
}
}
],
"plaintext": "Context over completeness. ",
"textSize": "large"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "The instinct of most data-driven people is to show everything. Every layer, every scenario, every data source. The right instinct is the opposite. A homeowner assessing purchase risk needs different information than a municipal engineer designing evacuation routes. A resident checking conditions on a Wednesday needs a different interface than an emergency manager coordinating response on a Thursday night when the river is rising. Design for the decision, not for the dataset."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#bold"
}
],
"index": {
"byteEnd": 22,
"byteStart": 0
}
}
],
"plaintext": "Trust over technology. ",
"textSize": "large"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "People don't act on data they don't trust. And trust is not built by citing your validation methodology or displaying confidence intervals. It is built by being right about the small events before the large one arrives — and by staying present when it does. The tools that perform best in real flood events are almost never the most technically sophisticated. They are the ones people already have a relationship with. The ones they have used before. The ones that have been right before."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "These principles are not in competition with good modelling. They depend on it. But they are the layer above the model — and right now, that layer barely exists."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.horizontalRule"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.header",
"level": 2,
"plaintext": "The Last Mile in Practice"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "The last mile is not completely unaddressed. A few examples demonstrate that the approach is both achievable and effective."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#bold"
}
],
"index": {
"byteEnd": 37,
"byteStart": 0
}
}
],
"plaintext": "Google's Flood Forecasting Initiative makes a different but equally important point. Google sends flood alerts through Google Search, Google Maps, and Android notifications. If you search the weather in a region that is about to flood, you see a warning. No app to download. No account to create. No map to interpret."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "The information arrives in the place people are already looking."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "That design decision — asking where the person is when they need this information, and bringing the information there — is one of the most underused ideas in flood risk communication. Most flood risk tools assume people will come to them. They assume citizens will navigate to a portal, select layers, and interpret a map. The best tools invert that assumption entirely. They go to the person. They meet people mid-thought."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.image",
"aspectRatio": {
"height": 1422,
"width": 2540
},
"fullBleed": true,
"image": {
"$type": "blob",
"ref": {
"$link": "bafkreia2fliysmlddxxkufot2ijrzyjo77kmqzp3ba4fbt5l5f3yniwnoi"
},
"mimeType": "image/webp",
"size": 72720
}
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#bold"
}
],
"index": {
"byteEnd": 15,
"byteStart": 0
}
}
],
"plaintext": "Overstroomik.nl, the Dutch portal whose name translates simply as \"Will I flood?\", operates on the same logic. Not \"view the national flood hazard map.\" Not \"explore inundation scenarios by return period.\" You enter your address. You get a plain-language answer based on your specific location, your elevation relative to flood scenarios, and what kind of flooding is most relevant to you. The technical complexity underneath is considerable. The interface is almost shockingly simple."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.image",
"aspectRatio": {
"height": 439,
"width": 800
},
"fullBleed": true,
"image": {
"$type": "blob",
"ref": {
"$link": "bafkreicwp5n44iulrzm2scbe2kwsezeaecjmcgdb5xsaqhln6e5hfywd3u"
},
"mimeType": "image/webp",
"size": 29010
}
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#bold"
}
],
"index": {
"byteEnd": 13,
"byteStart": 0
}
}
],
"plaintext": "FloodReady.dk is a property flood risk assessment for Danish homeowners currently in development. When you enter your address, you don't get a map. You don't get a return period. You don't get a confidence interval. You get answers to the questions a homeowner actually asks: how high does the water typically get near my property? How high did it reach during the last major storm? What is the worst-case scenario I should plan for? When were the last significant events in my area, and how frequent are they becoming?"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "The assessment ends with a to-do list — not recommendations, not guidance notes, but concrete measures: install a backflow prevention valve, consider barriers for ground-floor access points, check whether your insurance covers surface water flooding. Actions ranked by relevance to the specific property and risk level."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "FloodReady.dk does not own a single dataset and has not commissioned a single model. Everything behind the assessment is drawn from data that Danish public authorities already produce and publish — flood maps from the coastal authority, storm surge records, bluespot analyses from municipal planning documents. It is all out there. It has always been out there. What is being built is not the data. It is the translation layer — the interface that takes what authorities already know and delivers it as an answer to the question a homeowner is actually asking."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "All of these examples made a deliberate choice to optimise for the human outcome, not the technical output. They accepted a simpler interface as a feature, not a compromise. They asked \"what decision does this person need to make?\" before they asked \"what data do we have?\" None of them required a breakthrough in flood modelling. They required a breakthrough in priorities."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.horizontalRule"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.header",
"level": 2,
"plaintext": "Who builds the last mile?"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "The flood risk community — researchers, emergency managers, data scientists, engineers, designers — is very good at the things it can measure. Model skill scores. Computational efficiency. Data coverage. It is much less comfortable with the things that are harder to measure: comprehension, trust, behavioural change."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "But those unmeasured things are exactly where lives are saved or lost."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "There are three asks worth making here."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#bold"
}
],
"index": {
"byteEnd": 15,
"byteStart": 4
}
}
],
"plaintext": "For researchers: the next time you write a grant proposal or a paper abstract, ask whether it improves a model or improves a decision. Both have value. But only one is systematically underfunded, and it is not the model."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#bold"
}
],
"index": {
"byteEnd": 28,
"byteStart": 4
}
}
],
"plaintext": "For developers and engineers: build something a mayor can use. Build something a 65-year-old in a coastal village can use at 11pm when the storm surge warning goes out. If your product requires a GIS specialist to interpret, you have built for yourself, not for the people who need it."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"facets": [
{
"features": [
{
"$type": "pub.leaflet.richtext.facet#bold"
}
],
"index": {
"byteEnd": 22,
"byteStart": 4
}
}
],
"plaintext": "For emergency managers: you are the bridge. You are the translators between the technical systems and the communities you serve. Demand better tools. Reject interfaces that require a manual. Push vendors and agencies to prioritise usability alongside accuracy. The power to drive that change is already in the conversation — it only needs to be used."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "The last mile is not a technical problem. It is a prioritisation problem. Resources keep flowing into the last decimal because that is what institutions reward and what professional training tells us to value. Changing that requires a conscious, collective decision to value the human outcome over the technical output."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.horizontalRule"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.header",
"level": 2,
"plaintext": "The real breakthrough"
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "There is a version of the future worth working toward."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "In it, flood risk information works the way good medical communication works. A doctor doesn't show the raw MRI data and ask the patient to interpret it. They translate. They contextualise. They connect what they know to what the patient needs to do. And they build, over time, a relationship of trust that means the patient acts on their advice when it matters."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "That is what the last mile looks like when it works. Not a better model. A better conversation."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "The next breakthrough in flood risk management will not come from a satellite with a higher resolution sensor, or a neural network with a lower RMSE, or a simulation that runs ten times faster. It will come from the tools and the people that sit down with a community before the flood arrives, and communicate — in plain language, with relevant context — exactly what is coming and exactly what to do about it."
}
},
{
"$type": "pub.leaflet.pages.linearDocument#block",
"block": {
"$type": "pub.leaflet.blocks.text",
"plaintext": "That is the work worth doing. Everything else should be built in service of it."
}
}
],
"id": "019ea74f-ed4b-7dd0-92c4-d02058fe2856"
}
]
},
"coverImage": {
"$type": "blob",
"ref": {
"$link": "bafkreie4x4rcw2rpibhwpiwww2z6niegsls2rkflrzgok2sxgw5lquurou"
},
"mimeType": "image/webp",
"size": 292188
},
"description": "Europe's flood data is good enough. Getting it to the right people isn't.",
"path": "/3mnssyemuvk2w",
"publishedAt": "2026-06-08T23:18:58.087Z",
"site": "at://did:plc:k2h7zvrve2cfizg3q5ig7z37/site.standard.publication/3m3rw6bocsc2o",
"tags": [
"floodrisk",
"risk-communication",
"floods",
"GreySky",
"floodmodelling",
"riskcomm"
],
"title": "Stop Building Better Flood Models: Why the Last Mile Matters More Than the Last Decimal"
}