In　order　to　satisfy　the　demand　of　air　pollution　monitoring　with　more　spatial　details　and　dense　temporal　series　for　Taiwan　Island,　ground-level　PM2.5　concentrations　were　estimated　with　overall　regression　model　and　seasonal　optimal　regression　models　using　the　100m-resolution　AOD　retrieved　from　Chinese　GF-1　WFV　images.　The　AOD-PM2.5　relationship　was　different　in　different　seasons　because　of　significant　changes　in　aerosol　composition　and　climate　factors.　The　correlation　(R)　and　RMSE　between　the　observed　values　and　the　calculated　PM2.5　results　of　overall　optimal　model　amounted　to　0.551　and　14.30　μg/m3,　respectively,　which　demonstrated　a　good　accuracy.　The　seasonal　optimal　models　showed　better　forecast　results　through　comparing　with　monitoring　PM2.5　values　with　R　values　ranged　from　0.541　to　0.655,　and　RMSE　ranged　from　10.85　μg/m3　to　16.17　μg/m3.　When　the　seasonal　effects　to　the　relationship　of　AOD　and　PM2.5　concentrations　being　considered,　it　may　significantly　improve　the　model　performance.　Therefore　seasonal　models　are　more　suitable　for　PM2.5　estimation　and　GF-1　WFV　images　can　be　served　for　air　monitoring　at　high　spatial　and　temporal　resolution.　©　2019　IEEE.