Award Date


Degree Type


Degree Name

Master of Science in Electrical Engineering (MSEE)


Electrical Engineering

First Committee Member

Shahram Latifi

Second Committee Member

Henry Selvaraj

Third Committee Member

Sahjendra Singh

Fourth Committee Member

Laxmi Gewali

Number of Pages



Lately, there has been a tremendous increase in the number of climate monitoring

stations in various parts of the country producing abundant climate data. Among climate data parameters, humidity and temperature are the two parameters influencing hydrological and agricultural processes, weather monitoring, and having critical effect on living organisms. As more data is being generated over time, there is a strong need to develop compression methods for efficient transfer and storage of this data.

The main goal of this thesis is to perform compression of humidity and temperature

data via prediction. As these are critical components of climate, it is important that

compression of this data is lossless.

Data for this thesis is collected from ‘Nevada Climate Change Portal’ (NCCP) and

‘United States Geological Survey’ (USGS). Humidity data comprises of 1 and 10-minute

interval data for various sites in Nevada for 2013 and 2014, and temperature data comprises of hourly data from 1999 to 2012. The methodology is based on Artificial Neural Networks (ANN) to predict outputs. Feed-forward ANN model is used, in which learning is facilitated through back propagation. Modeled and observed values for humidity and temperature are compared and accuracy of the model is assessed. Differential encoding is then applied followed by Huffman coding. This is compared with results obtained by directly applying differential encoding and Huffman coding to raw data. Performance of the method is measured by metrics like Compression Ratio (CR) and Root Mean Square Error (RMSE).

Results indicate that the predicted model gives higher compression ratio when

compared to conventional method. In case of humidity, for 1-minute interval data, maximum compression ratio of 6.14 and 2.66 is achieved using proposed and conventional method respectively; and for 10-minute interval data, maximum compression ratio of 5.26 and 2.24 is achieved using proposed and conventional method respectively. In case of hourly temperature data, maximum compression ratio of 4.52 using proposed and 2.95 using conventional method is obtained. Data fluctuations being less for minute level data and values being closer to one another results in better predictions and thus higher compression ratios compared to 10-minute or hourly interval data.


Data compression


Computer Engineering | Electrical and Computer Engineering