GEO Theses and Research Projects

Permanent URI for this collection


Recent Submissions

Now showing 1 - 5 of 30
  • Item
    The application of machine learning algorithms for flood susceptibility assessment for the state of Kansas
    (Wichita State University, 2023-12) Rimal, Prashant; Demissie, Zelalem S.
    Flooding has been a significant problem in the United States (US) over the past century. Since 1996, more than 1,500 flood events have been recorded in Kansas, resulting in more than billions in losses. This project explored the use of machine learning and publicly available data to assess factors affecting flooding and develop a flood susceptibility map for Kansas at multiple resolutions. It aims to explore the major predictor variables or flood-controlling factors and the response of the Stack Generalization across multiple resolutions and scenarios. Six machine learning (ML) algorithms: Logistic Regression (LR); Random Forest (RF); Support Vector Machine (SVM); K-nearest neighbor (KNN); Adaptive Boosting (Ada Boost); Extreme Gradient Boosting (XG Boost) were employed to determine the most important factors influencing the susceptibility of an area to flooding. The learning set for the ML algorithms comprised geospatial datasets of thirteen flood-controlling factors: rainfall, elevation, slope, aspect, flow direction, flow accumulation, Topographic Wetness Index (TWI), distance from the nearest stream, evapotranspiration, land cover, impervious surface, land surface temperature, and hydrologic soil type. A total of 1,528 non-flood inventories were created for two different scenarios, with the only difference being the inclusion of stream buffers for overall analysis. The ML algorithms were compared and used to estimate flood susceptibility for each location in the geodatabase resulting in a flood-susceptibility map for both cases. Overall, testing results showed that the tree-based ensemble algorithms; XGB and RF ML models performed relatively well in both cases over multiple resolutions compared to other models in predicting flooding with an accuracy ranging from 0.82 to 0.97, respectively. Also, variable importance analysis depicted that predictor variables such as distance from the streams, hydrologic soil type, rainfall, elevation, and impervious surfaces significantly affect flood prediction.
  • Item
    Evaluating machine learning models using seismic data sets from different settings
    (Wichita State University, 2023-12) Taskinen, Eldon; Demissie, Zelalem S.
    In prior decades the concept of using mathematical methods to predict earthquakes was considered infeasible. Recent advances in machine learning and predictive modeling offer promising avenues to potentially realize earthquake prediction. In order to test the viability of machine learning methods, experiments were made with earthquake datasets from Kansas and Puerto Rico. The two datasets were chosen for the distinct differences in their tectonic settings. Kansas has few major faults, with a largely inactive subsurface, this produced a smaller dataset with a few large clusters. Puerto Rico is complexly faulted, with an extremely active tectonic setting, this produced a larger dataset with a large number of small clusters. In order to test the effectiveness of these two datasets for machine learning and prediction they were run through four different machine learning algorithms including an LSTM model, Bi-LSTM model, Bi-LSTM model with attention, as well as a transformer algorithm. Not only were the four different machine learning methods compared against each other for accuracy but also the datasets as well. Conclusive findings show that the two different data sets favor different processing methods. The Kansas data performs the best with the Bi-LSTM with attention model, while the Puerto Rico data performs the best with the LSTM model. This is likely due to the tectonic settings of the two regions, since the Kansas dataset has less overall data, and earthquakes are concentrated in a few large clusters, while the Puerto Rico data set has a more even distribution.
  • Item
    Determining an effective replacement for cyanide in the leaching of low-grade gold ores using an emerging industry standard and similar substitutions in the form of thiosulfate
    (Wichita State University, 2023-05) Pastor, Ryan; Swindle, Andrew L.
    In the mining industry, heap leaching is the most common form of extraction for many metals, especially gold, from low-grade ores in open-pit facilities. Cyanide solutions are the most commonly used solutions for the process of leaching low-grade gold ore. These solutions lead to relatively high yield of gold content within low-grade ore. Cyanide solutions pose an environmental risk. Cyanide is a reducing agent that interacts with free oxygen and is hazardous to both people and the environment. A common eco-friendly substitution comes in the form of thiosulfates. These solutions, although better for the environment, have lower yield than the common cyanide solutions. As such, higher concentrations are needed for similar yields relative to cyanide solutions. At the moment, the common thiosulfates used in the industry are sodium thiosulfate and ammonium thiosulfate. Thiosulfate usage and testing has been limited and focused on ammonium thiosulfate. How would changing the cation composition of the thiosulfate compounds affect gold dissolution? Would these varying solutions of thiosulfate lead to higher yields with lower concentrations, more similar to cyanide solutions? This study tests various thiosulfates (sodium, magnesium and calcium thiosulfates), each with varying concentrations, to identify any variations that would produce similar or even better yields than cyanide and ammonium thiosulfate.
  • Item
    Effects of quick vs slow-release fertilizers on nitrate movement into groundwater following precipitation events
    (Wichita State University, 2022-05) Samaniego, Hannah; Swindle, Andrew L.
    This research investigated the effect of quick-release and slow-release fertilizers on the transport of nitrate through the subsurface. The main focus of this study was studying nitrate concentrations in subsurface water samples over time, particularly in the presence of overfertilization. Evaluating nitrate movement allowed for assessment of the potential for contamination of shallow groundwater systems. Rainfall data was also evaluated to see the effects that precipitation had on nitrate concentration trends. Additionally, soil was studied to evaluate nitrate soil water concentrations and organic matter content. It was concluded that nitrate concentrations remained elevated above the Maximum Contaminant Level (MCL) of 10 mg/L nearly a year after initial fertilizer applications for both fertilizer types and moved past the rooting zone of plants. The data also suggested that fertilizer plays a larger role on changes in nitrate concentrations than increased precipitation. The conclusions of this study emphasize the need for fertilizer users to more fully understand their soil and vegetation’s nutritional needs prior to fertilizer application. Over-fertilization can have long term consequences, including the potential to pollute shallow groundwater systems.
  • Item
    The investigation of possible stratospheric origin of ozone presence at the surface at night, using HYSPLIT trajectory analysis
    (Wichita State University, 2021-05) Kosin, Moses Ejiroghene; Reynolds, Nathaniel
    IMPORTANCE: On August 2, 2012, high tropospheric ozone levels, above the EPA limit of 75 parts per billion by mass, were forecast for Wichita, Kansas. Trajectories for ozone transport were examined using two software packages: McIDAS and HYSPLIT. The source of the rise of ozone concentration was examined, namely whether it came from the stratosphere or was a buildup from the ground level. Evidence of stratosphere ozone was lacking. OBJECTIVE: The reason for this research is to investigate the source of high concentrations of ozone in the troposphere. One possibility is the entrance of stratospheric ozone via tropopause folding. Another possible source of high ozone concentration comes from photochemistry in the troposphere, possibly around Wichita or upstream of Wichita. However, the photochemistry details are not part of this study. Instead, the study concerns itself with transportation of ozone pollution to Wichita of either tropospheric ozone produced elsewhere or ozone precursors. METHODS: We use the McIDAS software to analyze the data we collected from the NCAR datasets and use the information to analyze the presence of stratospheric ozone and the ozone concentration in Wichita due to the impact of tropopause folding. Also, we applied the HYSPLIT model to investigate where the polluted air was. The map area is attempting to describe the source and movement of ozone, impacting human health. Therefore, the HYSPLIT trajectory analyzes the ozone movement from August 2, 2012, and August 4, 2012, to determine the wind direction. RESULT: During the period, the NCAR (National Center for Atmospheric Research) dataset was collected, and the ozone mixing ratio at 500 hPa and 850 hPa was run. The 500 hPa map and 850 hPa* map in figure 3 and figure 4, respectively, show the ozone values that are relatively high across the same region. The NCAR data result at 850 hPa* and 500 hPa agreed with the forecast of August 2, 2012, that the tropospheric ozone concentration was high. However, the stratosphere ozone cannot be established during this day because the HYSPLIT trajectory model suggested no wind migrating from the Rocky Mountains to Wichita, Kansas. The investigation reviewed that the data from the Cfsrr in the Cfsr (Climate Forecast System Reanalysis) was the framework for this study. It indicates the increased ozone concentration in the troposphere and shows that ozone expels from the stratosphere to the troposphere cannot be established. The 3-D grid model is used to analyze the data, and the color indicates the increase of ozone. Furthermore, the tongue-like structure suggests the presence of tropopause presence in the Rocky Mountains. CONCLUSIONS: There was no stratospheric ozone because the wind direction was not coming from the Rocky Mountains; instead, we saw the trajectory coming from Texas and local wind-generated within Kansas, Oklahoma, Arkansas, and Missouri. However, that part of the ozone transported from Texas, Arkansas, Missouri, and Oklahoma might have escalated the ozone concentration above 75ppb (part per billion).