Introduction: Urban greening is a climate change-related policy with considerable health benefits. But do these benefits extend to prevention of dementia and, if so, which types of green space matter? Method: Multilevel discrete time-to-event cohort study of incident Alzheimer's disease over 11 years among a baseline recruited between January 1, 2006 and December 31, 2009 (the Sax Institute's 45 and Up Study). Sampled participants for this study (N=109,688) were aged 45 years or older with no record of dementia up to 6 years before baseline, living in the cities of Sydney, Wollongong and Newcastle, Australia. Exposures were percentage total green space, tree canopy and open grass within 1.6-km road network distance buffers at baseline. Outcomes were time-to-first anti-dementia medication prescription (Department of Human Services) or dementia detected during hospitalisation or death up to 31 December 2016 (up to 11 years follow-up). Outcomes were analysed in parallel to triangulate on associations with green space, while testing for bias due to potential under-prescribing of anti-dementia medications. Models were adjusted for baseline person-level factors and area-level socioeconomic disadvantage. Results: Dementia detection varied by case ascertainment method. 1.55% (1,703/109,688) persons were detected using prescribed anti-dementia medications. 3.32% (3,639/109,688) persons were detected during hospitalisation or death via ICD-10 codes. Dementia incidence irrespective of outcome measurement was lower among females, younger participants, those living in couples, with higher qualifications and higher incomes. Dementia risk was lower with more tree canopy when the outcome was measured using hospital and death records (≥30% vs <10% tree canopy incidence hazard ratio (IHR) = 0.86, 95%CI 0.75, 0.99), after adjusting for person-level factors. The opposite association was observed when anti-dementia medications were used to detect dementia (≥30% vs <10% tree canopy IHR = 1.33, 95%CI 1.07, 1.66). Anti-dementia medication-based detection also indicated lower dementia risk with more open grass (≥20% vs <5% IHR = 0.83, 95%CI 0.67, 1.03). Anti-dementia medication prescribing was lower in the highest vs. lowest area-level disadvantage tertile (29.8% vs. 43.7%) among people diagnosed with dementia, indicating potential bias from geographic differences in prescribing practices. Adjusting for area-level disadvantage explained associations between tree canopy, open grass and dementia when detected by anti-dementia medication, but had negligible impact on negative (i.e. potentially protective) association between tree canopy and dementia detected by hospital and death records (≥30% vs <10% tree canopy hazard ratio 0.84, 95%CI 0.72, 0.99). Conclusions: Increasing urban tree canopy cover may help to reduce the risk of dementia. Replication in contrasting contexts and mediation studies to assess pathways are warranted.