A Random Forest Based Spatiotemporal Regression Model for the Impact of Greenspace on Childhood Academic Performance.

Abstract:

The relationship between greenspace exposure and children’s cognitive development has expanded from discussions on biophilia theories to statistically modeling the relationships between remotely sensed variables and academic performance. This association is often modeled using linear regression models on quasi-static covariates pertaining to greenspace access and childhood development. This study explores spatiotemporal relationships between data from the State of California’s Standardized Testing and Reporting Program (STAR), from 2002 to 2013, and the National Oceanic and Atmospheric Administration’s (NOAA) dataset on the Leaf Area Index (LAI) and the Fraction of Absorbed Photosynthetically Active Radiation (FAPAR) using a time-discrete model. Demographic variations across schools of study are integrated via data on the proportion of children eligible for Free or Reduced Price Meals (FRPM). A random-forest model is proposed to explore complex space-time relationships between greenspace variability and the STAR score outcome, while considering the influence of FRPM and static sources of pollution in the area. The results indicate FAPAR and LAI measures maintain a significant positive relationship with academic performance measures. The proposed model also delineates representative lags between the spatio-temporal variations evolution of greenspace and STAR scores. Measuring the degree to which each year’s conditions explained the tested year's variations showcased a need to frame the correlation as a dynamic relationship. Greenspace variability consistently correlated to the variability in STAR scores, highlighting the strong space-time correlations between access to nature and cognitive development. FRPM is also discovered to be an important variable showing that environmental concerns must be discussed within the realm of socio-demographic conditions. These findings suggest that a long-term relationship between greenspace and school scores exists, partially formulated by socio-demographic drivers, indicating the importance of environmental conditions on childhood development.