Prioritization
These are research rankings of where to look first. They are not damage assessments and do not say where damage is.
Cells of the area are ranked by several published methods and combined. The best cells and the recommended target boxes are on the map; the methods below say exactly what each does.
Recommended target boxes
| Rank | Box (W, S, E, N) | Why | Numbers |
|---|---|---|---|
| 1 | -80.9833, 8.0833, -80.9417, 8.1250 | A 4.6 x 4.6 km box centred at 8.1042, -80.9625. About 35,760 people live here, all of them where ShakeMap puts the shaking at MMI VII or more (mean MMI 7.4, peak 7.5). PAGER's country rates give about 3.72 expected deaths (a rank signal; PAGER itself quotes wide ranges). PAGER's economic model gives about US$ 148.0 million. Overture counts 17,374 buildings. USGS ground failure: liquefaction probability up to 0.02, landslide up to 0.00. Rank among 36,017 boxes: exposed_population #5, pager_fatality #1, pager_economic #17, building_damage #40, ground_failure #386. Even the least favourable method (ground_failure) places it in the top 1.1% of boxes; the largest gap between two methods is 385 places, so they agree it is a high-priority box but differ on how high. Chosen because it is the best combined rank among boxes that do not touch a better one. | mmi_mean: 7.36; mmi_max: 7.5; cells_in_shaking_footprint: 25; population: 35760.0; population_mmi_ge7: 35760.0; population_mmi_ge8: 0.0; expected_fatalities: 3.72; expected_loss_usd_million: 148.02; buildings: 17374.0; damage_equivalent_buildings: 331.8; p_liquefaction_max: 0.021; p_landslide_max: 0.002 |
| 2 | -80.4667, 7.3833, -80.4250, 7.4250 | A 4.6 x 4.6 km box centred at 7.4042, -80.4458. About 2,140 people live here, all of them where ShakeMap puts the shaking at MMI VII or more (mean MMI 8.8, peak 9.0). PAGER's country rates give about 2.47 expected deaths (a rank signal; PAGER itself quotes wide ranges). PAGER's economic model gives about US$ 188.7 million. Overture counts 1,328 buildings. USGS ground failure: liquefaction probability up to 0.45, landslide up to 0.00. Rank among 36,017 boxes: exposed_population #470, pager_fatality #16, pager_economic #1, building_damage #5, ground_failure #75. Even the least favourable method (exposed_population) places it in the top 1.3% of boxes; the largest gap between two methods is 469 places, so they agree it is a high-priority box but differ on how high. Chosen because it is the best combined rank among boxes that do not touch a better one. | mmi_mean: 8.82; mmi_max: 9.02; cells_in_shaking_footprint: 25; population: 2136.0; population_mmi_ge7: 2136.0; population_mmi_ge8: 2136.0; expected_fatalities: 2.473; expected_loss_usd_million: 188.65; buildings: 1328.0; damage_equivalent_buildings: 535.9; p_liquefaction_max: 0.454; p_landslide_max: 0.003 |
| 3 | -80.4417, 7.9500, -80.4000, 7.9917 | A 4.6 x 4.6 km box centred at 7.9708, -80.4208. About 37,570 people live here, all of them where ShakeMap puts the shaking at MMI VII or more (mean MMI 7.0, peak 7.1). PAGER's country rates give about 1.68 expected deaths (a rank signal; PAGER itself quotes wide ranges). PAGER's economic model gives about US$ 31.8 million. Overture counts 17,323 buildings. USGS ground failure: liquefaction probability up to 0.20, landslide up to 0.00. Rank among 36,017 boxes: exposed_population #2, pager_fatality #48, pager_economic #197, building_damage #628, ground_failure #29. Even the least favourable method (building_damage) places it in the top 1.7% of boxes; the largest gap between two methods is 626 places, so they agree it is a high-priority box but differ on how high. Chosen because it is the best combined rank among boxes that do not touch a better one. | mmi_mean: 6.96; mmi_max: 7.06; cells_in_shaking_footprint: 25; population: 37571.0; population_mmi_ge7: 37571.0; population_mmi_ge8: 0.0; expected_fatalities: 1.68; expected_loss_usd_million: 31.78; buildings: 17323.0; damage_equivalent_buildings: 67.0; p_liquefaction_max: 0.198; p_landslide_max: 0.0 |
| 4 | -80.3000, 7.7417, -80.2583, 7.7833 | A 4.6 x 4.6 km box centred at 7.7625, -80.2792. About 14,020 people live here, all of them where ShakeMap puts the shaking at MMI VII or more (mean MMI 7.3, peak 7.4). PAGER's country rates give about 1.38 expected deaths (a rank signal; PAGER itself quotes wide ranges). PAGER's economic model gives about US$ 52.0 million. Overture counts 7,710 buildings. USGS ground failure: liquefaction probability up to 0.06, landslide up to 0.00. Rank among 36,017 boxes: exposed_population #104, pager_fatality #75, pager_economic #74, building_damage #215, ground_failure #638. Even the least favourable method (ground_failure) places it in the top 1.8% of boxes; the largest gap between two methods is 564 places, so they agree it is a high-priority box but differ on how high. Chosen because it is the best combined rank among boxes that do not touch a better one. | mmi_mean: 7.32; mmi_max: 7.38; cells_in_shaking_footprint: 25; population: 14018.0; population_mmi_ge7: 14018.0; population_mmi_ge8: 0.0; expected_fatalities: 1.38; expected_loss_usd_million: 52.01; buildings: 7710.0; damage_equivalent_buildings: 131.4; p_liquefaction_max: 0.063; p_landslide_max: 0.002 |
| 5 | -80.5667, 7.7083, -80.5250, 7.7500 | A 4.6 x 4.6 km box centred at 7.7292, -80.5458. About 3,280 people live here, all of them where ShakeMap puts the shaking at MMI VII or more (mean MMI 7.8, peak 7.9). PAGER's country rates give about 0.69 expected deaths (a rank signal; PAGER itself quotes wide ranges). PAGER's economic model gives about US$ 41.7 million. Overture counts 1,619 buildings. USGS ground failure: liquefaction probability up to 0.26, landslide up to 0.01. Rank among 36,017 boxes: exposed_population #265, pager_fatality #173, pager_economic #120, building_damage #401, ground_failure #197. Even the least favourable method (building_damage) places it in the top 1.1% of boxes; the largest gap between two methods is 281 places, so they agree it is a high-priority box but differ on how high. Chosen because it is the best combined rank among boxes that do not touch a better one. | mmi_mean: 7.76; mmi_max: 7.9; cells_in_shaking_footprint: 25; population: 3282.0; population_mmi_ge7: 3282.0; population_mmi_ge8: 3282.0; expected_fatalities: 0.692; expected_loss_usd_million: 41.67; buildings: 1619.0; damage_equivalent_buildings: 92.4; p_liquefaction_max: 0.262; p_landslide_max: 0.006 |
Methods
| Method | Formula | Inputs | Limits | Official reference |
|---|---|---|---|---|
Population exposed to severe shakingexposed_population | score = POP_VII + POP_VIII, where POP_k = sum of people in cells with ShakeMap MMI >= k - 0.5 (MMI rounded to the nearest integer class); people at MMI VIII or more count twice (a convention, not an empirical weight) | ShakeMap grid.xml (MMI), WorldPop 2020 1 km population (or GHSL / any people-per-pixel raster) | MMI is the ShakeMap estimate, not an observation: it carries uncertainty that is not propagated here (grid.xml has no uncertainty field), so a cell at 6.4 and one at 6.6 are treated as different classes. | Population is a 2020 modelled raster at 1 km, redistributed by WorldPop's own covariates; it is not a census, it is not where people are at the hour of the earthquake, and it is blind to building type. | Counting VIII+ twice is a design choice that favours the most strongly shaken ground; the rank is insensitive to it only when the VII and VIII areas coincide. | earthquake.usgs.gov/data/shakemap/ earthquake.usgs.gov/data/pager/background.php doi.org/10.5258/SOTON/WP00647 |
USGS PAGER empirical fatality model, applied per cellpager_fatality | score = sum over cells of POP * nu_c(MMI), nu_c = the country fatality rate PAGER published for this event (losses.json, empirical_fatality.country_fatalities[c].rates, tabulated at MMI 1..10; for this event the table is exactly lognormal, so it is evaluated as the fitted Phi[ln(S/theta)/beta] at the cell's MMI); PAGER's functional form is nu(S) = Phi[ln(S/theta)/beta] (Jaiswal and Wald 2010, eq. 1) | ShakeMap grid.xml (MMI), WorldPop population, PAGER losses.json (country rates by MMI) | PAGER applies these rates to the whole country's exposure; applying them cell by cell is an extension that keeps the expected total but says nothing of which cell inside a country is worse than another beyond what shaking and population say. | The rates are country averages calibrated on past fatal earthquakes; the PAGER fatality alert is quoted by PAGER as an order-of-magnitude range, and the model is blind to local building stock. | Our population raster differs from the one PAGER uses, so the total will not equal PAGER's; the difference is reported under `validation`. | Cells whose country is not identified use the most-exposed PAGER country's rates (flagged in notes). | earthquake.usgs.gov/data/pager/ earthquake.usgs.gov/data/pager/background.php pubs.usgs.gov/of/2009/1136/ doi.org/10.1193/1.3480331 |
USGS PAGER empirical economic-loss model, applied per cellpager_economic | score = sum over cells of POP * v_c(MMI) * r_c(MMI); r_c = PAGER's country economic loss ratio for this event (losses.json, empirical_economic.country_dollars[c].rates), v_c = PAGER's own economic exposure divided by its population exposure in the same MMI class and country (exposures.json), r evaluated like the fatality rate, v interpolated in log between MMI 1..10; PAGER's form is r(s) = Phi[ln(s/theta)/beta] (Jaiswal and Wald 2011) | ShakeMap grid.xml (MMI), WorldPop population, PAGER losses.json, PAGER exposures.json | Dollar value per person is a national average from PAGER; it makes dense cells look as valuable as sparse ones per head, and ignores that urban cores concentrate far more value. | The loss ratio is a national empirical curve (economic loss over economic exposure), not a damage state. | Strongly correlated with pager_fatality because both multiply the same population by increasing functions of MMI: they agree by construction more than independent evidence would. | earthquake.usgs.gov/data/pager/ earthquake.usgs.gov/data/pager/background.php pubs.usgs.gov/of/2011/1116/ |
Building exposure x intensity-damage proxybuilding_damage | score = sum over cells of N_BLDG * r_c(MMI) * m(h), N_BLDG = Overture building footprints in the cell, r_c = the PAGER country economic loss ratio for this event (used as the mean damage ratio of the stock), m(h) = relative vulnerability by mean building height h; m = 1 for every height unless a person supplies a table (--vuln-json) because no source for a HAZUS or GEM height multiplier was fetched | ShakeMap grid.xml (MMI), Overture buildings counted on the lattice (catalogue/priority/inputs), PAGER losses.json, optional building heights raster (GlobalBuildingAtlas LoD1) or Overture height, optional --vuln-json | With the default m = 1 this is building count times a national loss ratio: it separates built-up from empty ground and strong from weak shaking, but cannot tell a masonry block from a steel frame. | A HAZUS-style fragility (lognormal median per building class and code level) and a GEM exposure-model class mix by country would sharpen it; their parameters were NOT fetched from a source, so none are applied. [UNVERIFIED] any height multiplier you supply is your assumption, recorded in the output. | Overture footprints are incomplete in places and most have no height outside North America and Europe; heights, where used, are modelled (3D-GloBFP and GlobalBuildingAtlas report RMSE of metres). | The PAGER economic loss ratio is a value-weighted national curve used here as a damage-ratio proxy; it is not a count of damaged or collapsed buildings. | [UNVERIFIED] height multiplier m(h) for any value other than 1 (HAZUS / GEM class-specific parameters not fetched) | docs.overturemaps.org/ earthquake.usgs.gov/data/pager/ pubs.usgs.gov/of/2011/1116/ www.globalquakemodel.org/product/global-exposure-model www.fema.gov/flood-maps/products-tools/hazus doi.org/10.5194/essd-17-6647-2025 |
Population on ground likely to fail (liquefaction and landslide)ground_failure | score = sum over cells of POP * (1 - (1 - P_liq)(1 - P_ls)); P_liq = Zhu et al. (2017) liquefaction probability, P_ls = Nowicki Jessee et al. (2018) landslide probability, both as published in the event's USGS ground failure product; the union assumes independence | USGS ground-failure zhu_2017_general_model.tif, USGS ground-failure jessee_2018_model.tif, WorldPop population | Probabilities are model outputs on ~250-450 m rasters; USGS itself issues hazard and population alerts on aggregated areas and not for a single cell. | Independence of the two hazards is assumed; where both are high (steep wet river valleys) the union is an upper bound only when they are positively correlated, not always. | It ranks where ground failure may add to shaking damage; it says nothing about the shaking damage itself, so it is a complement to the other four, and a cell can rank high here with little shaking. | earthquake.usgs.gov/data/ground-failure/ doi.org/10.1785/0120160198 doi.org/10.1029/2017JF004494 doi.org/10.5258/SOTON/WP00647 |
Agreement across methods (Borda count)agreement_borda | for each cell or tile and each scoring method with status ok: points = 1 - (rank - 1)/(n - 1), rank 1 = highest score, ties share the best rank; combined = mean of the points; combined_rank 1 = highest combined | the scores of the methods above | The methods share inputs (the same shaking and the same population), so agreement among them is weaker evidence than agreement among independent ones; it is a consensus, not a confidence. | Equal weights are a convention; the panel shows the per-method ranks so a reader can reweigh. | en.wikipedia.org/wiki/Borda_count |
Diversity: targets in different clustersdiversity_selection | greedy by combined rank over tiles of the requested size; a tile is accepted only if its centre is at least D km from every accepted centre (targets: D = 2 x tile size, so none touch; targets_diverse: D = --diverse-km, default 30) | tile scores, --tile-km, --min-sep-km, --diverse-km | Greedy selection is not optimal for the sum of scores; it avoids targeting one cluster, it does not maximise the total. | D is a choice. A single large city can hold several clusters at 30 km or none at 5 km. | en.wikipedia.org/wiki/Greedy_algorithm |
SAR practicality: layover, shadow, next passsar_practicality | for a building of height h seen at grazing angle g (incidence 90 - g): layover length on the ground = h * tan(g), shadow length = h / tan(g); reported for the mean building height of the tile at g = 15, 45 and 70 degrees, the limits of the tasking window this package sends; next pass = earliest SAR pass in catalogue/next_passes.json whose footprint intersects the tile. Reported, never used to change a rank | building height raster or Overture height (optional), catalogue/next_passes.json (optional) | Elementary geometry for a flat-roofed block facing the radar; real layover also depends on look direction relative to the street grid and on roof shape. | Tile mean height hides the tall minority that causes the worst layover; where heights are unknown the fields are null, never assumed. | The pass list is an estimate unless it says otherwise; whether a commercial satellite can be scheduled only the provider's feasibility call knows. | doi.org/10.5194/essd-17-6647-2025 docs.overturemaps.org/ |