Laboratory for Remote Sensing and Geoinformatics 　 Directed by  Dr. Hongjie Xie , Assistant Professor

Graduated Students and Their Theses/Dissertations

     Summer 2008

Newfel Mazari (Master's),

Current address:

PhD student

Environmental Science and Engineering PhD Program

University of Texas at San Antonio

Tel: 210-458-7815

Email: newfel@yahoo.com
　

Thesis Title: Validation of NEXRAD products with rain gauge networks

　

Thesis Committee: Drs. Hongjie Xie (Chair), Alan Dutton, and Hatim Sharif

Abstract:

The purpose of this study is to improve our understanding of Next Generation Weather Radar (NEXRAD) precipitation products by comparing them to high accuracy of rain gauge network, for the estimation of precipitation quantity and spatial rainfall distribution. The first part of the study was the calibration and installation of double-gauge platforms within a single radar cell (1 km by 1°), with some preliminary data analysis based on collected data. The second part was a comparison study of a network of 50 rain gauges with NEXRAD’s Digital Storm Total Precipitation (DSP) product.

            The first part consists of two phases: a testing phase for hardware calibration and development of data processing based on one double-gauge platform on the UTSA campus, and a second phase where a network of four double-gauge platforms within one radar cell is installed in the Government Canyon State Park, near San Antonio.  The preliminary results indicate a correlation coefficient of 85% with a p-value of 0.0001 and a radar underestimation by 23% for paired gauge-radar events within a 6-month period (August 2006 to March 2007). Rain gauge rainfall data collected from the Park (August 2007 to June 2008) did not show a large spatial variation of rainfall distribution, which indicate most of the rainfall events in study area and time period are uniform events, although it is indicated that the “SE’ of the radar cell has relative higher rain rates at certain storms and lower correlations to other three pairs of gauge measurements.

            In the second part, a network of 50 rain gauges rainfall in the Upper Guadalupe River Basin were used to compare the DSP rainfalls from two radars, the KEWX radar at New Braunfels, Texas and the KDFX radar at Laughlin Air Force Base near Del Rio, Texas, for the period of September 2006 to May 2007. The rainfall data comparisons were examined based on different time scales: 6 minutes, 30 minutes, one hour, and storm total, different distances from radar to gauges (from near range, middle range to far range), and different gauge elevations from 200 m to 700 m. It is found that there is a strong radar range dependence as previously found: underestimate in the near range (< 50 km) and far range (>150 km), while overestimate in the middle range (50-150 km). The results found that at the storm total time scale that the DPS product offers a good rainfall estimation and detection, and that the correlation with rain gauges is the highest (KEWX  R²= 0.59 and KDFX  R²= 0.37); and the mean relative differences between gauges observations and radars estimates are the lowest (52% for the KEWX and 59% for the KDFX). It is found there is no gauge elevation dependence in the study area.

            The probability of rainfall detection (POD) is dependant on time scales. It is found that radar PODs are less than gauge PODs in the 6 minutes, 30 minutes and hour time scales, while in the storm total scale, the radar PODs are higher than the gauge PODs. It is 92% for the KEWX and 85% for the KDFX at the storm scale, which are higher than the rain gauges PODs of 84% and 79%, respectively. It is supposed that radar PODs should be always larger than the gauge PODs. The reason for this inconsistencies needs further discussion.

            Overall, it is clear that the DSP product is range dependant; its rainfall estimation carries some errors due to the Z-R relationship power law. This product can be used for storm monitoring and flash flood warning. A longer study period with a different rain gauges network is needed to provide further comparison and analysis between radar DSP product and rain gauges rainfall observations.

　

     Spring 2008

Xianwei Wang (PhD),

Current address:

PostDoc Research Fellow

Department of Earth System Science

University of California-Irvine

Irvine, CA 92697

Tel: 949-842-1571

Email: xianweiw@uci.edu
　

Dissertation Title: Applications of remote sensing and GIS in surface hydrology: snow cover, soil moisture and precipitation

　

Thesis Committee: Drs. Hongjie Xie (Chair), Richard French, Hatim Sharif, Kyle Murray, and Jerome Keating

Abstract:

Studies on surface hydrology can generally be classified into two categories, observation for different components of surface water, and modeling their dynamic movements. This study only focuses on observation part of surface water components: snow cover, soil moisture, and precipitation. Moreover, instead of discussion on the detailed algorithm and instrument technique behind each component, this dissertation pours efforts on analysis of the standard remotely sensed products and their applications under different settings. 

First in Chapter 2, validation of MODIS Terra 8-day maximum snow cover composite (MOD10A2) in the Northern Xinjiang, China, from 2000-2006, shows that the 8-day MODIS/Terra product has high agreements with in situ measurements as the in situ snow depth is larger or equal to 4 cm, while the agreement is low for the patchy snow as the in situ snow depth less than 4 cm. However, the cloud blockage in the 8-day products is still high in most winter times although the clouds are much lower than that in the daily snow cover product. According to the in situ observation, this chapter develops an empirical algorithm to separate the cloud-covered pixels into snow and no snow. This separation generates a new snow cover time series that corrects the dramatic decrease of raw snow cover area caused by MODIS cloud mask, thus yielding a better estimation of the actual snow coverage in a watershed. This solves the problem of cloud blockage frustrating hydrologic modelers. This new cloud free snow cover time series is further used to study the seasonal and inter-annual variation of snow cover at this region. Variation of snow area extent (SAE) at Northern Xinjiang is closely associated with air temperature.  The increase of elevation generally accompanies the decrease of air temperature, resulting in more snow cover extent and longer snow cover duration. During the six hydrologic years from 2000-2001 to 2005-2006, the SAE has a similar pattern, although there is variation at the beginning of snow accumulation and at the end of snow melting in different years. Because of the short duration of MODIS data, the change trend of snow cover is not obvious. Therefore, the continued long-term production of MODIS-type snow cover product is critical to assess water resources of the study area, as well as other larger scale global environment monitoring, offering critical inputs for hydrologic and climatic modeling, forecasting, and climate change analyses.

Terra and Aqua satellites carry the same MODIS instrument and provide two parallel MODIS daily snow cover products at different time (local time 10:30 am and 1:30 pm, respectively). Chapter 3 develops an algorithm and automated scripts to combine the daily MODIS Terra (MOD10A1) and Aqua (MYD10A1) snow cover products, and to automatically generate multi-day Terra-Aqua snow cover image composites, with flexible starting and ending dates and a user-defined cloud cover threshold. The daily combination product (MODMYD10DC) offers much more open space than the daily MODIS Terra or Aqua only since cloud keeps moving in most times. The multi-day Terra-Aqua composite product (MODMYD10MC) provides about 3 times of images as and similar accuracy with the standard 8-day composite products with the 2003-2004 hydrologic year at test sites of Northern Xinjiang, China, and at the Colorado Plateau, USA. The new multi-day composites are significant contributions and complement to the standard MODIS snow cover product series. This approach can also be extended to other optical remote sensing images, e.g., to combine other Terra and Aqua MODIS products, like vegetation indices, etc.

Chapter 4 systematically compares the difference between MODIS Terra and Aqua snow cover products within a hydrologic year of 2003-2004, validates the MODIS Terra and Aqua snow cover products using in situ measurements in Northern Xinjiang, and compares the accuracy among the standard MODIS Terra and Aqua snow cover products, and the new combined daily and multi-day composite from both MODIS Terra and Aqua daily products. Taken the MODIS/Terra daily and 8-day products as references when being compared with MODIS/Aqua daily and 8-day products (MYD10A2), respectively,  the agreement of land classification from MODIS Terra and Aqua daily and 8-day snow cover products is close to 100% in the entire year under clear sky in both morning and afternoon. In contrast, the agreement of snow classification from MODIS Terra and Aqua is high only in the winter months, decreasing in the rest of the year. The snow cover agreement in the daily products is higher than that in the 8-day products. When being compared with the in situ snow depth observations, under the so-called clear sky conditions after removing the cloud data pairs, three daily products (MOD10A1, MYD10A1, and MODMYD10DC) have similar accuracy of snow and land classification. Moreover, in the actual weather/cloud conditions, the daily combination of MODIS Terra and Aqua reduces the cloud blockage and improves the land and snow classification accuracy, although it still has high cloud percentage. Three composite products (MOD10A2, MYD10A2, and MODMYD10MC) also have similar accuracy of snow and land classification after removing the cloud data pairs. Similarly, in the actual weather/cloud conditions, the new multi-day composite product reduces the cloud blockage and improves the snow classification accuracy against either the 8-day MODIS Terra or Aqua snow cover product.

In Chapter 5, utilizing the new cloud-low multi-day composite of MODIS Terra and Aqua snow cover products, several new methods are developed to study the spatiotemporal variation of snow cover conditions from different aspects at the Northern Xinjiang and on the central Tianshan Mountains, mainly in China, partly covering Kazakhstan and Kyrgyzstan. MODIS-derived snow covered days (SCD) has good agreement of 90% with in situ SCD from hydrologic years of 2001-2005. Overall, SCD increases as elevation increases when elevation is less than 4 500 m. The R² value between SCD and DEM is 0.62 on the entire Central Tianshan Mountains.  Snow cover index (SCI) value contains both duration and extent information and is easily used to quantify the variation of the overall snow cover situation. MODIS-derived snow cover onset dates (SCOD) and snow cover melting dates (SCMD) also have good agreement with the major snow cover duration from in situ observation by one week forward and backward shifts because of transient snowfall events in the early and at the end of the snow season, respectively. In the six years, the perennial snow in August 2005 has the least spatial extent (2 386 km²), which is believed to be a most close map of the glacier distribution at this peak area.  More accurate map of glacier distribution can be obtained using Landsat or higher resolution images.

Secondly, Chapter 6 investigates the feasibility to indirectly map root-zone soil moisture using optical remote sensing techniques and in situ measurements. Specifically, covariation of root-zone soil moisture with the normalized difference of vegetation index (NDVI) from MODIS observation is studied at three sites (New Mexico, Arizona, and Texas). The three sites represent two types of vegetation (shrub and grass) and two types of climate conditions: arid/semi-arid (New Mexico and Arizona) and humid (Texas). Results show that the root-zone soil moisture has significant linear correlation with vegetation (NDVI).  Then, a linear regression model is developed to estimate the root-zone soil moisture based on NDVI and in situ soil moisture. The time-series of root-zone soil moisture estimated by NDVI using the linear regression model accounts for 42 - 71% of the observed soil moisture variations for the 3 sites. Thus the point ground measurement of soil moisture can be interpolated to areal soil moisture using NDVI as a proxy of soil moisture variation.

Finally, Chapter 7 validates and compares the NEXRAD Stage III and MPE precipitation products using a high density rain gauge network on the Upper Guadalupe River Basin of the Texas Hill Country in 2001 and 2004.  Because of the point-area representativeness error between rain gauge and radar rainfall estimation, this chapter develops a new method to automatically select uniform rainfall events based on coefficient of variation criteria of 3 by 3 radar cells. Only gauge observations of those uniform rainfall events are used as ground truth to evaluate radar rainfall estimation. Results show that rain gauge observations of uniform rainfall better represent ground truth of a 4 x 4 km² radar cell than non-uniform rainfall; MPE has better performance than Stage III; Stage III tends to overestimate precipitation (19.5%), but MPE tends to underestimate (-7.2%).

     Spring 2008

Misti A. Thueson (Master's),

Current address:

Chemistry Teacher

Email: misti.thueson@gmail.com
　

Thesis Title: Exploring high-albedo event craters in the near-polar permanent ice cap of Mars.

　

Thesis Committee: Drs. Hongjie Xie (chair), Stephen F. Ackley, and Rupali Datta

Abstract: This study focused on the water-frost/ice and carbon dioxide frost interactions in impact craters of the northern seasonal polar ice cap on Mars.  Studying the interaction between water and carbon dioxide frost/ice within the outer permanent ice cap can help us in determining the role of these two materials in the hydrological and geological cycles, as well as the possible biological cycle (if any) on Mars. Using THermal EMission Imaging System (THEMIS) infrared and visible images, Thermal Emission Spectroscopy (TES) Lambert albedo and target temperature data and Mars Orbital Laser Altimeter (MOLA), a detailed study of seven previously identified craters with high albedo events (HAEs) in the period of late spring to late summer was carried out.  Results indicate six of the seven observed impact craters with both AM and PM HAEs are associated with a permanent water-ice body (or patch) covering a part or all of those crater floors, while one impact crater with AM-only HAEs is not associated with a permanent water-ice body. This observation is consistent with a conceptual model proposed by Xie et al. [in press] that an exposed permanent water-ice body is the basis for the AM and PM HAEs in those craters, while a subsurface water ice body or ice-rich regolith in the near subsurface is a necessary condition for the AM-only HAEs crater. Variations in albedo in Phases 2 and 3 of the Martian annual cycle are found in some craters. This is mainly due to the exposed, permanent water-ice body (patch) that is only a small portion of the crater floor, resulting in a limited contribution from the water-ice body to the albedo change of the entire crater floor measured from the TES instrument.   

     December 2006

Lani M. Cabico (Master's),

Current address:

GIS Analyst
Planning and Development
Brooks City-Base
Brooks Development Authority,

San Antonio, Texas
Tel: (210)678-3315
Email: lani.cabico@brookscity-base.com

Thesis Title: Correlation between RAWS LFMC and MODIS vegetation indices and reflectance: A case study in West Texas.

　

Thesis Committee: Drs. Hongjie Xie (chair) and Youn-Min Chou

Abstract: Satellite remote sensing provides spatially and temporally continuous datasets for vegetation analysis at large coverage areas.  In this study, three vegetation indices and band composites derived from Moderate Resolution Imaging Spectroradiometer (MODIS) reflectivity product (MOD9A1: 8-day and 500m) are examined for their correlations with the Remote Automated Weather Stations (RAWS)-derived live fuel moisture content (LFMC) during the time period of February 2000 to December 2004.  The correlation analyses were based on green-up months (April 1^st – October 31^st) between collocated LFMC from a RAWS station and MODIS-derived vegetation indices or band composites from one pixel where the RAWS station is located at the center (near center) of the pixel. Three RAWS together with additional 4 sites (total 7 pixels) were chosen in the West Texas study area.  In situ data were also analyzed to determine the best fit for correlation with MODIS vegetation indices. It is found that, based on RAWS-derived LFMC, the multiple variable regression (7 bands together) approach results in the best correlation in establishing equations to produce LFMC maps for all study sites for the mixed vegetation community of West Texas.  The strongest relationship between RAWS and MODIS was found at Fort Davis site3 in multiple regression analysis (r² = 0.55, p<0.0001). The best correlation (r² = 0.61, p<0.0001) between MODIS indices and in situ measurements was based on Normalized Difference Infrared Index band 6 (NDII₆) at the Fort Davis Mountains within the High Pocosin vegetation fuel class.