Code base for the JustGreen shiny app. This app aims to showcase the relationship between NDVI distributions within a city and health outcomes.
Overall Goal : For the 200 most populated cities within the United States, showcase NDVI values, and preprocess a quantitative assessment of the relationship between NDVI values and health metrics
Population data: https://services.arcgis.com/P3ePLMYs2RVChkJx/arcgis/rest/services/USA_Census_2020_DHC_Population_Households_Place/FeatureServer
NDVI : Generate from Landsat or Sentinel-2 data
Census tracts : US Census
- gather max yearly NDVI values from GEE using 1k buffer geographies of city
- Get the mean and sd of each census tract, buffered to 500m that intersects with the city
- For all census tracts gather the total population 18 years old and older
- NDVI null value is considered 0.1
- For each census tract
- calculated the difference between the NDVI null value and average value for that area
- Apply the dose reponse function to relate current NDVI values to health improvements
- dose function from Garber et al 2024. 0.146 increase in NDVI results in 4.5% change in non-accidental mortality (decrease)
- convert this to a attributiable fraction (define)
- Garber et al 2024 : PAF = 1 - (1 / (RR^(NDVIdiff/0.146))).
- RR -; 0.954
- use the relationship between mortality x/100,000 people, census tract population, and attributable fraction
- garber et al 2024 : crude number of deaths prevented = population * crude mortality rate * PAF * -1 -result; number of deaths prevented by NDVI exposure