The Development of a Cloud-Based Application for Estimating Air Temperature Using Satellite Image and Crowd-Sourced Weather Data (2017)

Name: Thannarot Kunlamai
ID: 117753
Title: The Development of a Cloud-Based Application for Estimating Air Temperature Using Satellite Image and Crowd-Sourced Weather Data
Type: Thesis


Air temperature (T ) is one of the most sensitive parameter of dynamical processes. Moreover, Ta can be estimated from the relation between Land Surface Temperature (LST) satellite data product and surface air temperature from ground-based methodological stations. With the availability of crowd-sourced personal weather observation devices, the referenced air temperature can be easily accessed via the Internet. However, the main problem of crowd-sourced data is the reliability of the data, it is needed to develop algorithm to filter outlier observation from crowd sources. Additionally, the technology of cloud computing is commonly used as a service infrastructure for processing task without installing software and it can help to reduce the time to develop web processing service. The overall objective of this study is the development of a cloud-based application for estimating T . Firstly, a data collection system was developed to collect weather data from a distributed sensor network. Then, a suitable algorithm to filter weather data from crowd sources was applied for estimating air temperature with different land cover type and spatial distribution of weather station. The crowdsourced weather data were requested by using Wundergroud weather API. In this study, we proposed the approach for estimating T using the data at the same date and time the satellite passes the study area. We found that our approaches that are the estimating T based on land cover type and based on spatial distribution give the better accuracy than the conventional method that conducted with the data in one year period. And also by filtering the outlier observation from nearby area within 10 kilometers, we found the high difference value of RMSE was obtained around 1.00 °C when comparing with the estimating Ta using all crowdsourced weather data without data filtering. According to the results of comparing the performance between the standard weather stations which the small number of weather station and crowdsourced weather stations have been indicated that the number of weather station affected the accuracy of the estimated air temperature map. The estimating air temperature using crowdsourced weather data from 228 weather stations gives the lower RMSE than using the standard weather data from 10 ASOS stations with approximately 1.00 °C. The application was developed with JavaScript, Google Map API, Google Visualization API, and Google Earth Enginethat used to retrieve and process geospatial data on Google’s infrastructure. The cloud-based application was configured to provide the air temperature information on the Internet and allow the user can export the estimated air temperature image in TIFF file through the web browser. The estimated air temperature data from this study can be as preliminary data for further analysis such as weather prediction and weather forecasting application.

Keywords: Estimating Air Temperature, Cloud Computing, Crowdsourced Data, Google Earth Engine.