Spatial and temporal scale sensitivity in NDVI-based greenness exposure assessment for Alzheimer’s disease incidence: a Miami-Dade County Medicare Cohort Study
- 1SAGE GeoAI Lab, Department of Geography, University of Florida, Gainesville, FL 32611, USA
- 2University of Miami Miller School of Medicine, FL 33136, USA
- 3SAGE GeoAI Lab, Department of Geography, University of Florida, Gainesville, FL 32611, USA; Corresponding author.
Abstract
Alzheimer’s disease (AD) is a leading cause of dementia, and neighborhood greenness may represent a protective environmental exposure. However, greenness-health associations may depend on how vegetation exposure is defined across spatial and temporal scales. We examined the influence of the modifiable areal unit problem (MAUP) and modifiable temporal unit problem (MTUP) on associations between Landsat-derived Normalized Difference Vegetation Index (NDVI) and incident AD among Medicare beneficiaries in Miami-Dade County, Florida. NDVI exposure was evaluated at four resolutions (30, 120, 250, and 500 m), using single-date imagery and seasonal composites, baseline 2011 and combined 2011+2016 exposure specifications, and census-block-level mean and range metrics. The primary exposure specification was baseline 2011 seasonal mean NDVI; single-date imagery, 2011+2016 NDVI, and NDVI range were evaluated as sensitivity analyses. Generalized estimating equation logistic regression models estimated associations with AD incidence from 2012 to 2016 after adjustment for demographic and neighborhood characteristics. Model performance was compared using the Quasi-likelihood under the Independence Model Criterion. Higher mean NDVI was consistently associated with lower AD incidence, although effect magnitude, precision, and model fit varied across spatial and temporal specifications. Seasonal composites produced stronger and more stable associations than single-date imagery, indicating that temporal aggregation influences greenness exposure assessment. Seasonal mean NDVI at 250 m provided the strongest relative model performance within this comparison. NDVI range was positively associated with AD incidence but was less stable and more sensitive to spatial resolution, supporting its interpretation as an exploratory measure of within-block spectral variability rather than a primary greenness exposure metric. These findings demonstrate that finer spatial resolution does not necessarily improve epidemiologic inference and highlight the need for scale-aware exposure specification in environmental health research and urban planning for healthy aging.
Related
- Spatial–temporal heterogeneity in tuberculosis incidence and its association with socioeconomic factors in South Sulawesi
- An Alzheimer’s disease-associated mutant of C. elegans displays mechanosensory sensitivity following exposure to lavender extracts
- High sensitivity-C-reactive protein variability and risk of cancer incidence: A longitudinal prospective cohort study cancer
- Tobacco-attributable burden of alzheimer’s disease and other dementias in China and ASEAN : Temporal trends and projections – an observational study based on GBD 2021
- Leprosy incidence in the state of Piauí, Brazil: spatial distribution and temporal trend, 2013–2023
BibTeX
@article{chaudhary2026spatial,
title = {{Spatial and temporal scale sensitivity in NDVI-based greenness exposure assessment for Alzheimer’s disease incidence: a Miami-Dade County Medicare Cohort Study}},
author = {Anusha Chaudhary and Julius R. Dewald and Jane Southworth and Ruixuan Ma and Mohammad Safaei and Jose Szapocznik and Scott C. Brown},
journal = {Environmental Advances},
year = {2026},
month = oct,
doi = {10.1016/j.envadv.2026.100765},
url = {https://doi.org/10.1016/j.envadv.2026.100765},
note = {CC BY},
}
RIS
TY - JOUR
TI - Spatial and temporal scale sensitivity in NDVI-based greenness exposure assessment for Alzheimer’s disease incidence: a Miami-Dade County Medicare Cohort Study
AU - Anusha Chaudhary
AU - Julius R. Dewald
AU - Jane Southworth
AU - Ruixuan Ma
AU - Mohammad Safaei
AU - Jose Szapocznik
AU - Scott C. Brown
PY - 2026
DA - 2026/10/30
JO - Environmental Advances
DO - 10.1016/j.envadv.2026.100765
UR - https://doi.org/10.1016/j.envadv.2026.100765
AB - Alzheimer’s disease (AD) is a leading cause of dementia, and neighborhood greenness may represent a protective environmental exposure. However, greenness-health associations may depend on how vegetation exposure is defined across spatial and temporal scales. We examined the influence of the modifiable areal unit problem (MAUP) and modifiable temporal unit problem (MTUP) on associations between Landsat-derived Normalized Difference Vegetation Index (NDVI) and incident AD among Medicare beneficiaries in Miami-Dade County, Florida. NDVI exposure was evaluated at four resolutions (30, 120, 250, and 500 m), using single-date imagery and seasonal composites, baseline 2011 and combined 2011+2016 exposure specifications, and census-block-level mean and range metrics. The primary exposure specification was baseline 2011 seasonal mean NDVI; single-date imagery, 2011+2016 NDVI, and NDVI range were evaluated as sensitivity analyses. Generalized estimating equation logistic regression models estimated associations with AD incidence from 2012 to 2016 after adjustment for demographic and neighborhood characteristics. Model performance was compared using the Quasi-likelihood under the Independence Model Criterion. Higher mean NDVI was consistently associated with lower AD incidence, although effect magnitude, precision, and model fit varied across spatial and temporal specifications. Seasonal composites produced stronger and more stable associations than single-date imagery, indicating that temporal aggregation influences greenness exposure assessment. Seasonal mean NDVI at 250 m provided the strongest relative model performance within this comparison. NDVI range was positively associated with AD incidence but was less stable and more sensitive to spatial resolution, supporting its interpretation as an exploratory measure of within-block spectral variability rather than a primary greenness exposure metric. These findings demonstrate that finer spatial resolution does not necessarily improve epidemiologic inference and highlight the need for scale-aware exposure specification in environmental health research and urban planning for healthy aging.
SN - 2666-7657
ER -