Inverse Distance Weighting (IDW) Model Develop in TerrSet (Still Working)
Raster System Development (for TerrSet), Clark University
This project, which I worked solely, intended to add some user specify functions and learn how to use code to realize inverse distance weighting (IDW) interpolation model in TerrSet. First, I compared the old interpolation model in TerrSet with the IDW model in ArcGIS, and learned the basic algorithm behind this method. Then I created IDW model with Delphi Environment, and enabled user to specify number of points be used or search radius. Finally, applied dialog to Dynamic Link Libraries (DLL), and integrated the user interface form into TerrSet.
Computer programming for suitability mapping
Computer programming with python, Clark University
This is my final project for the Computer Programming for GIS course, worked with Tianze Li and Hannah Rush. We created a suitability mapping tool within ArcMap 10.4, which can automatically create suitability map based on input parameters and assess image accuracy. The methods for the Suitability Mapping script tool include: the specification of parameter inputs, the reclassification of raw variable rasters to true probability images, the calculation of the final suitability image, and finally accuracy assessment between the suitability image and a user-specified validation image. The script tool was created using Model Builder in ArcMap and the resulting script served as a template for the suitability mapping procedure and subsequent accuracy assessment.This project allows us to better use python programming in a specific case study.
Evaluation of Landsat 8 and Sentinel Sensors to Identify Seagrass in Coastal Area
Advanced Remote Sensing, Clark University
Aquatic plants, specifically seagrass play an important role in underwater ecosystems and is highly important for the sequestration of carbon. It is for this fact that the importance of identify where seagrass is becoming increasing important, Conservation International (CI) gave me and Jason Ready a project to see if we could find a process of remote sensing to efficiently identify this underwater vegetation and develop a methodology to assist others in identifying oceanic vegetation.
In this investigation, we compared different method of identifying seagrass with the use of Landsat 8, Sentinel 1, and Sentinel 2. Since the study area was mostly covered with cloud, we find a small island inside of the study area which have little cloud for all images. As for Sentinel 1, the RADAR data, it covered with amounts of noise and could not show information under water. Then use Spectral Mixture Analysis (SMA) worked only with Sentinel 2 and Landsat 8. False color composite - red, green, and near inferred (NIR) was used to find a cloud free study area which can visually see underwater vegetation. Then make four endmembers for both sensors: Land, Seagrass, Deep water, and Sediment. Then made signature for all of the endmembers . Finally, use Spectral Mixture Analysis (SMA) worked with both Sentinel 2 and Landsat 8.
After running the SMA for the sentinel 2 and Landsat 8 sensors, we found that Sentinel 2 had increased spatial resolutions, which increased width of the NIR band (744nm - 940nm) and leads to greater range of reluctance within photosynthetic vegetation.. However Landsat 8 sensor had the advantage of being well established with many images, over many dates allowing for more options.
In this investigation, we compared different method of identifying seagrass with the use of Landsat 8, Sentinel 1, and Sentinel 2. Since the study area was mostly covered with cloud, we find a small island inside of the study area which have little cloud for all images. As for Sentinel 1, the RADAR data, it covered with amounts of noise and could not show information under water. Then use Spectral Mixture Analysis (SMA) worked only with Sentinel 2 and Landsat 8. False color composite - red, green, and near inferred (NIR) was used to find a cloud free study area which can visually see underwater vegetation. Then make four endmembers for both sensors: Land, Seagrass, Deep water, and Sediment. Then made signature for all of the endmembers . Finally, use Spectral Mixture Analysis (SMA) worked with both Sentinel 2 and Landsat 8.
After running the SMA for the sentinel 2 and Landsat 8 sensors, we found that Sentinel 2 had increased spatial resolutions, which increased width of the NIR band (744nm - 940nm) and leads to greater range of reluctance within photosynthetic vegetation.. However Landsat 8 sensor had the advantage of being well established with many images, over many dates allowing for more options.
Evaluating major corps suitability under the impacts of climate change on a global scale
Advanced Raster GIS, Clark University
Current and future climate data are acquired from WorldClim website, including monthly total precipitation, monthly minimum average temperature and monthly maximum average temperature. The parameters of corn, wheat, rice, and potatoes are obtained from Ecocrop. The first step is climate data preprocessing. Since temperature data acquired were not real values, temperature values were first divided by 10 in SCALAR modular in TerrSet Software. Also, since Crop Climatic Suitability Modeling in TerrSet require monthly mean temperature and WorldClim does not have it for future climate condition, future monthly mean temperature was calculated by averaged monthly average minimum temperature and monthly average maximum temperature by using OVEERLAY and SCALAR modular. Then Crop Climatic Suitability Modeling was processed in TerrSet, which primarily use fuzzy logic to model current suitability and future suitability of four major crops. For future suitability modeling, each climate model and each RCP scenario across four major crops were applied. Next step is models ensemble and assessment. Principle Component Analysis was used across each climate model for four major crops and four RCP scenarios, and chose principle component 1 as model consensus. The median loadings of component 1 were also evaluated to assess model uncertainty. Final step is the suitability change analysis. The suitability values were rescaled from principle component 1 into 0 to 1 by using SCLAR modular. Suitability differences between current and future climate conditions were evaluated and suitability changes were defined into three levels, low (0-0.25), moderate (0.25-0.5), and high (0.5-1) for both increasing and decreasing by using RECLASS modular.
Assessing the Effects of Public School Quality on Single-family Housing Price in Towns of Massachusetts
Advanced Vector GIS, Clark University
This is my final project for Advanced Vector GIS working with Lei Tang. The objectives to this project is to examines the effects of public school quality on single-family housing price in towns of Massachusetts in 2016. School quality was represented by nine explanatory variables, which are Student Attendance Rate, Percentage of Top Students, Composite Performance Index, SAT Score, Student Stability, Class Size, Student/Teacher Ratio, Teachers' Annual Salary, and Teachers' Proficiency. By conducting Moran's I of single-family housing price, and Getis-Ord Gi * Statistic Hot Spot Analysis of dependent and independent variables, these two results show similar distributions between nine explanatory variables and housing price. Results of Ordinary Least Squares (OLS), which is a global linear regression, showed a moderate strong correlation between school quality and housing price.