Skip navigation
Please use this identifier to cite or link to this item: http://arks.princeton.edu/ark:/88435/dsp013t945t88s
Title: Evaluating Forecasting Methods for Precipitation using Weather Data collected on top of Guyot Hall
Authors: Zhang, Tyrone
Advisors: Simons, Frederik J
Department: Geosciences
Class Year: 2021
Abstract: Princeton's climate has four seasons, with strong temperature variations, and precipitation occurring throughout the year. Statistically, precipitation event sequences can be characterized as drawn from exponential distributions in the three variables the precipitation event duration, intensity (the total precipitation divided by the duration), and the non-precipitation duration. The shortest and least intense precipitation events are the most frequent. Analyzing the precipitation measured from 2017 to the present day by a Vaisala WXT530 weather station located on the roof of Guyot Hall, I first summarize the data in terms of exponential distributions and their parameters, by season and by year. Subsequently, I evaluate the skill in predicting the arrival, duration and intensity of precipitation events solely based on this local "climatology'', before including other variables logged by the weather station. Predicting precipitation events using this climatology yielded 2-3 % precipitation accuracy. Thus, we proceeded to use linear regression and decision trees regression to improve the precipitation accuracy. Linear regression yielded 4.7 %, while decision tree yielded 11.9 %. Then neural networks were used in the form of LSTM, where we had hourly and minute inputs. The hourly input resulted in 9.9 %, while the minute inputs resulted in 16.8 %. With each new model, we are able to see improvements in the accuracy of predicting precipitation. However, there are further improvements that can be made with many pathways forward to improve the precipitation accuracy.
URI: http://arks.princeton.edu/ark:/88435/dsp013t945t88s
Type of Material: Princeton University Senior Theses
Language: en
Appears in Collections:Geosciences, 1929-2023

Files in This Item:
File Description SizeFormat 
ZHANG-TYRONE-THESIS.pdf3.95 MBAdobe PDF    Request a copy


Items in Dataspace are protected by copyright, with all rights reserved, unless otherwise indicated.