“A Regression Model for Predicting Percent Built-up Land Cover from Remotely Sensed Imagery of Pucallpa, Peru” by Drake Sprague
Abstract: Accurate information about built-up land cover and population density is essential for sustainable urban growth, especially in lesser developed countries. Unfortunately, this data is often too expensive for planning agencies, prompting use of outdated and unreliable information. As a proxy for estimating population density, a linear regression model is proposed to test the relationship between the percentage of built-up land cover and vegetation in Pucallpa, Peru. Expert knowledge, low-cost moderate-resolution satellite imagery, and high-resolution Google Earth images are used to estimate the percentage of built-up land cover at randomly assigned reference locations. Normalized Difference Vegetation Index (NDVI) data, acquired at each reference point, is the independent variable in a linear regression model constructed to predict the percentage of built-up land cover. The results were successful, with an adjusted R 2 = 0.774 at 95% confidence. Strength and accuracy are further evaluated against zoning maps and population estimates provided by local authorities.