Author

Ronan Lucey

Date of Award

2019

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Atmospheric and Earth Science

Committee Chair

Robert E. Griffin

Committee Member

W. Lee Ellenburg

Committee Member

Udaysankar Nair

Subject(s)

Agriculture--Remote sensing, Neural networks (Computer science), Food security, Vegetation mapping--Nepal

Abstract

In the context of global environmental change, food security is a major concern for Nepal. Agricultural interventions have been performed in Nepal with a goal of increasing crop productivity to combat food insecurity. In this study, we compare two techniques for estimating NDVI, serving as a proxy for agricultural productivity, over fields in western Nepal - a Multi-Linear Regression approach and a Random Forest approach. Assessment of agricultural productivity is important for understanding the effectiveness of agricultural interventions. The methodology utilizes space-borne remote sensing and surface meteorology datasets to address three research questions: 1) can these datasets be utilized along with statistical and machine learning models to predict single time values of NDVI, 2) predict time series values of NDVI and, 3) can the resulting NDVI predictions be used to assess the impact of agricultural interventions? Multi-linear regression and Random Forest models were constructed to predict future values of NDVI for both one time step and as a time series. For a single time step NDVI prediction, Random Forest (RF) models showed better overall performance ($R^2$ = 0.67) compared to the Multiple Linear Regression (MLR) model ($R^2$ = 0.15). While the RF model is less skillful at time series prediction ($R^2$ \textless 0.5), it performs better compared to a MLR model making similar predictions. A framework for assessment of impact of interventions was developed, and the methodology was tested with the results of the RF time series prediction. Preliminary investigations show that Artificial Neural Networks (ANN) have the potential to improve upon the results obtained from RF and MLR models for NDVI prediction.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.