{"id":"W4388908869","doi":"10.1016/j.envpol.2023.122914","title":"Using deep transfer learning and satellite imagery to estimate urban air quality in data-poor regions","year":2023,"lang":"en","type":"article","venue":"Environmental Pollution","topic":"Air Quality and Health Impacts","field":"Environmental Science","cited_by":14,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of British Columbia","funders":"Medical Research Council; Northeastern University; Universities Space Research Association; Wellcome Trust; National Aeronautics and Space Administration; University of California, Los Angeles; National Science Foundation","keywords":"Urbanization; Satellite imagery; Air quality index; Satellite; Transfer of learning; Deep learning; Air pollution; Estimation; Computer science; Developing country; Distribution (mathematics); Environmental planning; Remote sensing; Geography; Meteorology; Artificial intelligence; Economic growth; Mathematics","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0005768962,0.0007268857,0.0003399894,0.0006631745,0.0002397127,0.0005333544,0.0007983125,0.0007063049,0.000855563],"category_scores_gemma":[0.001896769,0.0002500475,0.0004464779,0.0007275979,0.0004336995,0.0008176507,0.0009386668,0.0008249175,0.0003086531],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0009059444,"about_ca_system_score_gemma":0.0006658809,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.03989387,"about_ca_topic_score_gemma":0.03422532,"domain_scores_codex":[0.999831,0.00004358645,0.000007667632,0.00005168778,0.00002411684,0.00004202752],"domain_scores_gemma":[0.9996712,0.0001156246,0.00004962833,0.00005107495,0.00008375297,0.00002875874],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0001569385,0.0002453722,0.04019765,0.00005340653,0.0001300998,0.0001718742,0.0001122293,0.8583823,0.002636672,0.0008648889,0.003148347,0.09390022],"study_design_scores_gemma":[0.000007909997,0.00001315953,0.003368397,0.000004195979,0.000006085204,0.00000645642,0.00002911032,0.9951368,0.0005087358,0.0007070091,0.0002081731,0.000004024765],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.8829656,0.0006037698,0.1081694,0.001201357,0.00009259931,0.0000573126,0.0009490578,0.001991404,0.003969512],"genre_scores_gemma":[0.9857726,0.00008195113,0.01203471,0.0001215344,0.00002832436,0.00001949813,0.000928323,0.00002928938,0.0009837927],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.03989387,"threshold_uncertainty_score":0.07932329,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.08593483753603297,"score_gpt":0.3670806612192948,"score_spread":0.2811458236832619,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}