cs.AIOct 11, 2026
SaveGoogle Earth AI Geospatial Reasoning Agent: Capabilities and a Systematic Evaluation
Organizations: Google Research · Google Geo · Spatial Bytes
Abstract
Google Earth AI is a family of geospatial AI models, together with an agent that orchestrates them. We present this agent, Google Earth AI's Geospatial Reasoning agent (GRA), a Gemini-powered research system that combines geospatial data sources, foundation models, and tools to answer questions about our planet. GRA organizes these into modular domain skills and pairs them with upfront geospatial planning to decompose complex queries into multi-stage workflows. Evaluating such systems is hard: real-world problems admit multiple valid workflows, require data discovery across large catalogs, and rely on evolving ground truth. We introduce two benchmarks and release their development splits. GeoExpertQueries poses complex analytical problems grounded in practitioner use cases and refined by geospatial experts; each is paired with a consensus golden workflow from three independent experts and scored by a human-validated, tool-agnostic autorater. GeoKTools consists of templated queries with deterministic reference solvers that systematically vary the number of required tool types. We also use two external benchmarks, CBGB and UnivEARTH; the latter we find to be saturated. To isolate the contribution of GRA's tools and its agentic orchestration, we compare four configurations, each with Gemini 3.1 Pro and 3.8 Flash: a standalone LLM, the LLM with basic tools, the LLM with GRA's skills and tools but without its upfront planning and execution guidance, and GRA itself. With at most basic tools, an advanced model recovers much of the answer on templated queries but satisfies only a small fraction of the expert workflows on open-ended problems; GRA performs well on both. The largest step comes from GRA's skills and tools; on open-ended problems, its planning and execution guidance add a further gain with Gemini 3.8 Flash, letting Flash match or exceed Pro at lower cost.
Figures & tables
Figure 1: The Geospatial Reasoning agent (GRA) and how we evaluate it. The agent plans its approach to a natural-language query, draws on geospatial capabilities across five pillars (Table 1 ), analyzes the retrieved data, and synthesizes a grounded answer. The bottom panel shows how our two benchmarks score the agent’s responses. Greyed-out items are not evaluated in this report.
| Capability Pillar | Specialized Models & Data Assets | Representative Analytical Operations |
|---|---|---|
| Imagery & Earth Observation | AlphaEarth Foundations* (annual satellite embeddings); Copernicus DEM GLO-30; Earth Engine* catalog imagery (e.g., Sentinel-1/2, Landsat, MODIS); Remote Sensing Foundations models* | Embedding-based surface change detection; terrain slope; country-scale open-vocabulary search & classification over overhead imagery |
| Weather, Climate & Hazards | WeatherNext* (ensemble forecast statistics); Google experimental cyclone model*; Google Flood Forecasting* & Google Runoff Reanalysis & Reforecast (GRRR); Google Maps Data; ERA5/ERA5-Land, NASA GPM IMERG, PRISM; NOAA IBTrACS; NEX-GDDP-CMIP6, WRI Aqueduct, JRC GloFAS, USDA Wildfire Risk to Communities; GFPLAIN250m, JRC Global Surface Water | Point weather & cyclone track/wind forecasting; flood forecasts, inundation & gauge return periods; floodplain screening; historical weather statistics & anomalies; climate-hazard projections to 2100 |
| Population, Health & Socioeconomics | Data Commons* (e.g., US Census ACS, CDC, FEMA NRI, World Bank); PDFM* (Population Dynamics embeddings) | Demographic, economic & health statistics across administrative levels; regional similarity search |
| POIs, Boundaries & Mobility | Google Maps Data*; Who’s On First, US Census TIGER/Line & ZCTA | Boundary resolution (countries to census tracts) & spatial filtering; POI search & counting; routes, travel-time matrices & isochrones; rooftop solar potential |
| Prediction & Spatial Compute | Scratch datasets with sandboxed Python; Google Earth Engine* (general-purpose analysis of any catalog dataset); TimesFM*; PDFM*; Planetary Prediction Engine ( Ma et al., 2026 ) , which combines them with AlphaEarth Foundations* embeddings and Data Commons* data | Automated data and model selection ( Ma et al., 2026 ) ; out-of-context data handling & custom Python analysis, including custom Earth Engine scripts; CRS-aware buffering, distances & spatial joins; regional statistics over gridded datasets; time-series forecasting; regional prediction & super-resolution |
Table 1: Overview of the GRA capability ecosystem across five domain pillars (* denotes models or datasets developed at Google).
Figure 2: Overview of GRA. Skills stay on standby and are loaded only when a query needs them. The agent first sets out a plan, then carries it out in a reasoning loop, while intermediate datasets pass between tools outside the model’s context window.
Figure 3: The four evaluated configurations. Each adds one layer to the same Gemini model: basic tools, then GRA’s geospatial skills and tools, then GRA’s upfront planning and execution guidance.
| Benchmark | Type | Queries (dev/test) | Evaluation | Primary Focus |
|---|---|---|---|---|
| GeoExpertQueries | Internal, open-ended | 25 / 76 | Golden workflows + tool-agnostic autorater | Real-world geospatial reasoning |
| GeoKTools | Internal, controlled | 85 / 220 | Ground-truth answers + autorater | Multi-tool orchestration |
| CBGB | External | — / 44 | Deterministic answer match (original criterion) | Quantitative Earth-observation analysis |
| UnivEARTH | External | — / 140 | Ternary answer match | Data discovery and end-to-end reasoning |
Table 2: Summary of the benchmark suite.
| Domain | Example query |
|---|---|
| Natural Hazards & Crisis Resilience | Using SAR Interferometry (InSAR) coherence change detection, estimate the structural damage in the ‘hard-hit’ zones of the 2022 Pakistan floods in the Sindh and Balochistan provinces. Differentiate between ‘inundated’ (water) and ‘destroyed’ (rubble). Overlay with poverty maps to prioritize reconstruction grants. |
| Public Health | Assess the physical accessibility of emergency obstetric care in Sierra Leone by calculating travel times over the Dynamic World land cover friction surface. |
| Site Selection & Urban Planning | For the city of Buenos Aires, determine the ‘Building Footprint’ area vs. available land area. |
Table 3: Example queries from GeoExpertQueries, drawn from the publicly released development split. Appendix B explains each example in plain language.
| Benchmark | LLM | LLM + Basic Tools | LLM + GRA Tools | GRA | ||||
|---|---|---|---|---|---|---|---|---|
| 3.1P | 3.8F | 3.1P | 3.8F | 3.1P | 3.8F | 3.1P | 3.8F | |
| GeoExpertQueries | 0.0 0.0 | 0.0 0.0 | 2.3 0.8 | 3.3 0.8 | 50.8 1.7 | 52.2 5.4 | 47.6 2.7 | 62.7 2.5 |
| GeoKTools | 37.1 1.2 | 40.5 0.7 | 37.8 1.9 | 40.0 0.8 | 84.5 1.4 | 86.0 1.6 | 81.6 1.4 | 85.9 1.5 |
| CBGB | 14.4 3.5 | 12.9 1.3 | 13.6 2.3 | 19.7 2.6 | 84.8 1.3 | 87.9 1.3 | 84.8 1.3 | 86.4 2.3 |
| UnivEARTH | 86.9 1.8 | 93.6 0.7 | 96.7 1.8 | 98.3 0.8 | 97.9 0.7 | 98.8 0.4 | 89.5 0.4 | 98.1 0.4 |
Table 4: Main results (test set). Scores are 0–100, mean std over three runs. For CBGB, the score is the average over all 44 questions, across its three difficulty tiers. Bold marks the best result in each row and results whose ranges overlap with it.
| Domain (n) | LLM + Basic Tools | LLM + GRA Tools | GRA | |||
|---|---|---|---|---|---|---|
| 3.1P | 3.8F | 3.1P | 3.8F | 3.1P | 3.8F | |
| Crisis Resilience (35) | 1.0 1.6 | 1.0 1.6 | 49.4 1.5 | 45.6 3.8 | 50.0 1.4 | 66.0 2.6 |
| Public Health (17) | 0.0 0.0 | 0.0 0.0 | 45.4 8.6 | 42.5 10.2 | 39.2 2.9 | 55.6 3.2 |
| Site Selection (20) | 3.6 0.5 | 6.9 1.3 | 57.2 5.3 | 71.4 5.0 | 48.1 6.3 | 65.6 3.4 |
| Operational (4) | 18.1 4.8 | 19.4 6.4 | 54.2 7.2 | 55.6 6.4 | 59.7 4.8 | 50.0 0.0 |
| All (76) | 2.3 0.8 | 3.3 0.8 | 50.8 1.7 | 52.2 5.4 | 47.6 2.7 | 62.7 2.5 |
Table 5: GeoExpertQueries results by domain (test set). Scores are 0–100, mean std over three runs. The standalone LLM scores 0.0 in every domain with both models and is omitted. Domains: Natural Hazards & Crisis Resilience; Public Health; Site Selection & Urban Planning; Operational & Commercial Utilities. Bold marks the best result in each row and results whose ranges overlap with it.
| LLM | LLM + Basic Tools | LLM + GRA Tools | GRA | |||||
|---|---|---|---|---|---|---|---|---|
| 3.1P | 3.8F | 3.1P | 3.8F | 3.1P | 3.8F | 3.1P | 3.8F | |
| (30) | 46.5 2.5 | 43.7 3.2 | 47.7 5.0 | 51.6 3.4 | 85.0 2.1 | 90.4 3.8 | 86.2 7.1 | 91.2 2.1 |
| (110) | 41.0 1.7 | 44.0 0.1 | 39.1 2.8 | 41.8 1.5 | 85.1 1.9 | 85.3 2.2 | 79.4 1.6 | 85.3 1.1 |
| (50) | 28.9 2.0 | 33.9 1.3 | 35.8 2.6 | 35.3 1.2 | 83.3 4.1 | 86.2 2.2 | 80.1 3.6 | 82.0 2.6 |
| (30) | 27.3 3.8 | 35.6 3.6 | 26.9 2.9 | 29.4 1.4 | 83.6 1.1 | 84.0 3.3 | 87.8 1.6 | 89.2 2.3 |
| All (220) | 37.1 1.2 | 40.5 0.7 | 37.8 1.9 | 40.0 0.8 | 84.5 1.4 | 86.0 1.6 | 81.6 1.4 | 85.9 1.5 |
Table 6: GeoKTools results by number of required tool types, (test set; query counts in parentheses). Scores are 0–100, mean std over three runs. Bold marks the best result in each row and results whose ranges overlap with it.
| Tier (n) | LLM | LLM + Basic Tools | LLM + GRA Tools | GRA | ||||
|---|---|---|---|---|---|---|---|---|
| 3.1P | 3.8F | 3.1P | 3.8F | 3.1P | 3.8F | 3.1P | 3.8F | |
| Easy (15) | 24.4 3.8 | 22.2 3.8 | 22.2 3.8 | 26.7 0.0 | 100.0 0.0 | 100.0 0.0 | 95.6 3.8 | 97.8 3.8 |
| Intermediate (20) | 8.3 5.8 | 11.7 2.9 | 13.3 5.8 | 21.7 2.9 | 78.3 5.8 | 81.7 2.9 | 80.0 0.0 | 81.7 2.9 |
| Difficult (9) | 11.1 0.0 | 0.0 0.0 | 0.0 0.0 | 3.7 6.4 | 74.1 6.4 | 81.5 6.4 | 77.8 0.0 | 77.8 0.0 |
| All (44) | 14.4 3.5 | 12.9 1.3 | 13.6 2.3 | 19.7 2.6 | 84.8 1.3 | 87.9 1.3 | 84.8 1.3 | 86.4 2.3 |
Table 7: CBGB results by difficulty tier (test set; question counts in parentheses). Scores are 0–100, mean std over three runs. Bold marks the best result in each row and results whose ranges overlap with it.
| Configuration | 3.1P | 3.8F |
|---|---|---|
| LLM | 86.9 1.8 | 93.6 0.7 |
| LLM + Basic Tools | 96.7 1.8 | 98.3 0.8 |
| LLM + GRA Tools | 97.9 0.7 | 98.8 0.4 |
| GRA | 89.5 0.4 | 98.1 0.4 |
Table 8: UnivEARTH results (test set, 140 questions). Scores are 0–100, mean std over three runs.
| Benchmark | LLM | LLM + Basic Tools | LLM + GRA Tools | GRA | ||||
|---|---|---|---|---|---|---|---|---|
| 3.1P | 3.8F | 3.1P | 3.8F | 3.1P | 3.8F | 3.1P | 3.8F | |
| GeoExpertQueries | $0.028 | $0.010 | $0.115 | $0.044 | $0.512 | $0.167 | $0.688 | $0.199 |
| GeoKTools | $0.024 | $0.012 | $0.575 | $0.112 | $0.410 | $0.147 | $0.360 | $0.122 |
| CBGB | $0.025 | $0.017 | $0.514 | $0.132 | $0.140 | $0.110 | $0.151 | $0.093 |
| UnivEARTH | $0.006 | $0.002 | $0.029 | $0.010 | $0.116 | $0.046 | $0.271 | $0.079 |
Table 9: Estimated inference cost per query (USD).
| LLM + GRA Tools | GRA | |||
| 3.1P | 3.8F | 3.1P | 3.8F | |
| Imagery & Earth Observation | 1% | 2% | 4% | 3% |
| Weather, Climate & Hazards | 18% | 18% | 23% | 18% |
| Population, Health & Socioeconomics | 21% | 22% | 23% | 20% |
| POIs, Boundaries & Mobility | 83% | 57% | 93% | 89% |
| Prediction & Spatial Compute | 83% | 75% | 93% | 99% |
Table 10: Capability usage on GeoExpertQueries (test set, all three runs). Top: share of queries in which at least one skill from each capability pillar (Table 1 ) is loaded. Bottom: average activity per query. General-purpose Earth Engine skills are counted under Prediction & Spatial Compute, since they serve all domains. Map-display skills are omitted.
Appendix figures & tables1 asset
Supplementary material from the paper’s appendix.
Appendix
| Category | Capability | Analytical Functionality & Operations | Underlying Data Source or Model |
|---|---|---|---|
| Imagery & Earth Observation | Embedding-based surface change detection | Annual surface-similarity maps and vectorized change polygons between two years, from cosine similarity of 64-dimensional satellite embeddings | AlphaEarth Foundations* ( Brown et al., 2025 ) (in Google Earth Engine* ( Gorelick et al., 2017 ) ) |
| Terrain slope | Zonal slope statistics (mean, min, max percent grade) from a 30 m elevation model over administrative or custom geometries | Copernicus DEM GLO-30 ( European Space Agency, 2022 ) | |
| Open-vocabulary overhead imagery search & classification | Zero-shot search for objects or scenes, and open-vocabulary land-cover/scene classification, over high-resolution US imagery (2023–2025) | Remote Sensing Foundations models* ( Barzilai et al., 2025 ) , Google overhead imagery* | |
| Weather, Climate & Hazards | Point weather forecasts | Current, hourly and daily point forecasts (temperature, precipitation, wind, humidity, UV, cloud cover) | Google Maps Data* |
| Ensemble forecast statistics | Regional ensemble mean, spread and percentiles | WeatherNext* ( Rasp et al., 2026 ) | |
| Tropical cyclones | Ensemble track, wind-field and strike-probability forecasts for active storms; historical storm tracks and wind-swath impact areas | Google cyclone model* (FNV3, via Weather Lab) ( Alet et al., 2026 ) ; NOAA IBTrACS v4 ( Knapp et al., 2010 ) ; NOAA NHC strike probabilities |
Table 11: GRA capabilities by pillar (* developed at Google).