SURF_Poster_REU2016Team3
•
0 likes•18 views
This document summarizes research using dimensionality reduction techniques like Sliced Inverse Regression (SIR) and the Nadaraya-Watson Estimator (NWE) to more accurately predict precipitation levels in the agriculturally important Missouri River Basin region using large climate data sets spanning 57 years. The techniques improved predictions by accounting for the semi-continuous nature of precipitation data and reduced the number of covariates from thousands to just a few meaningful factors. Parallelizing the SIR and NWE algorithms across multiple processors significantly improved computational efficiency, allowing the full dataset to be processed up to 16 times faster.
Report
Share
Report
Share
Download to read offline
![Dimensionality Reduction Using Sliced Inverse
Regression In Modeling Large Climate Data
UMBC REU Site: Interdisciplinary Program in High Performance Computing
Ross Flieger-Allison1
, Lois Miller2
, Danielle Sykes3
, Pablo Valle4
RAs: Sai K. Popuri3
, Nadeesri Wijekoon3
Faculty: Nagaraj K. Neerchal3
Client: Amita Mehta5
1
Williams College 2
DePauw University 3
UMBC 4
Kean University 5
JCET
• The Missouri River Basin (MRB) is a
significant agricultural region that is
not irrigated and thus dependent on
rainfall
• Precipitation predictions are histor-
ically inaccurate due to the semi-
continuous nature of observed precip-
itation data
Figure 1: The Missouri Rivier Basin
Background
• Precipitation data is semi-continuous,
meaning many observations are equal
to 0 and all other observations are
positive and follow a continuous dis-
tribution
• Predictions are calculated using ob-
served data and data from MIROC5,
a Global Climate Model (GCM),
across 57 years (1949-2005) from over
270,000 locations
• Precipitation at any given location s
depends on a large number of co-
variates: current and past values of
monthly precipitation, sea-level pres-
sure, relative humidity, and max-
imum/minimum temperatures at s
and its neighboring locations
Data
• Sliced Inverse Regression (SIR) is a
data-analytic tool that can be used to
reduce the dimensionality of a large
set of covariates
• Nadaraya-Watson Estimator (NWE)
is a simple non-paramentric smooth-
ing regression method:
ˆmx =
n
i=1 Kh(x−xi)yi
n
i=1 Kh(x−xi)
• The NWE is adapted to account for
the semi-continous nature of predic-
tions
Methodology
• The following figures show a comparison
between observed precipitation data, and
their predictions
Figure 2: Observed and predicted monthly precipitation, including
the proportion of 0 values for both
Figure 3: Mean Squared Error from positive predictions for months
with positive rainfall (true positives)
Results
Table 1: Performance Study Results
Processes Subregion∗
MRB
1 01:14:24 N/A
4 00:19:07 09:54:22
8 00:11:04 05:20:31
16 00:05:58 02:49:49
32 00:03:13 04:25:51
64 N/A 03:41:26
∗ denotes a region from latitutude -101 to -97 and longitude 39.25 to 42.95
Figure 4: Observed speedup (Sp = (T1/Tp)) of parallelization
Parallelization
• We have successfully demonstrated
that SIR and NWE methods can
be implemented to work on a large
dataset
• Parallelization of SIR and NWE code
greatly increases computational effi-
ciency on the subregion, and also im-
proves efficiency for the entire MRB
region, up to 16 processes
Conclusions
[1] National Integrated Drought Information
Systems. https://www.drought.gov/
drought/dews/missouri-river-basin/
about.
[2] S. K. Popuri, K. P. Adragni, et al., “Spatio-
Temporal Analysis of Daily Precipitation via a
Sufficient Dimension Reduction.” (In Progress).
[3] Full Technical Report: HPCF-2016-13 hpcf.
umbc.edu > Publications
REU Site: hpcreu.umbc.edu
NSF, NSA, DOD, UMBC, HPCF, CIRC
References & Acknowledgments](https://image.slidesharecdn.com/e6737b7f-4fde-4c5c-9c98-1a62d9572384-170119153045/85/SURF_Poster_REU2016Team3-1-320.jpg)
Recommended
SURF_Poster_REU2016Team1
1) Researchers used a technique called Minimum Average Deviance Estimation (MADE) to reduce the dimensionality of historical climate and rainfall data from the Missouri River Basin in order to retrospectively predict rainfall.
2) They tested MADE on simulated data and found that its prediction error decreased as the number of observations and covariates increased. They also found MADE outperformed other estimators when the number of observations was highest.
3) The researchers then applied MADE to over 18,000 observations of climate variables and precipitation measurements from 1949-2005 in the Missouri River Basin. Using a subset of 5 years of data from a single location, MADE was able to reduce the data dimensions from 5 to 1 while achieving a mean
Absorption factor method of radiation
This document describes the absorption factor method of radiation. It involves calculating the absorption factor Bij, which is the fraction of emitted energy from surface i that is absorbed by surface j. An algebraic equation is given relating Bij and the emitted energy Ei. The equations for determining Bij are linear and can be solved using various matrix methods. Once calculated, the absorption factors do not depend on surface temperature and can be used to determine radiative transport for a given geometry under different thermal conditions by writing new energy balance equations. However, the absorption factor method requires solving a larger system of equations than the more commonly used radiosity method.
Leaf Area Index (LAI) in the quantification of vegetation disturbance in Iris...
Leaf Area Index (LAI) in the quantification of vegetation disturbance in Iris...Environmental Protection Agency, Ireland
Dr Jerome O Connell - presentation made at various conferences throughout Europe as part of PhD which was funded by the EPA under the STRIVE Research Programme 2007-2013 (2007-PhD-ET-2)Presentation Boers - Trends in Cloudiness
This document discusses trends in cloudiness and challenges in detecting trends from observational data. It notes that precision in monitoring cloud cover has increased over time but accuracy of long-term trends is difficult to determine due to changes in observation methods, locations, and instruments. Global satellite data shows small or no statistically significant changes in total cloud cover over various time periods. Trends over the Netherlands cannot be reliably assessed due to gaps in surface observer records when instrumentation changed. Reconstructing aerosol and fog trends requires long records of relative humidity and visibility. The public may have difficulty distinguishing demonstrated climate trends from spurious trends caused by changes in observations. Successful experts in climate monitoring require persistence, attention to detail, and focus on continuity of measurements over
Spatial presentation of prognosis models in plant protection
This document discusses the development and use of risk maps for plant protection models. Models are enhanced using geostatistics to interpolate meteorological data across 1,367 virtual stations, allowing risk maps with grids of 1 km2. These maps ease the interpretation of model results across large regions. New technologies bring high resolution spatial meteorological data and mobile internet access to enable location-based decision support services online. Risk maps and online access optimize the provision of millions of model runs to farmers and advisors.
GENERATING FINE RESOLUTION LEAF AREA INDEX MAPS FOR BOREAL FORESTS OF FINLAND...
This document summarizes research generating fine-resolution leaf area index (LAI) maps for boreal forests in Finland using satellite imagery. Two methods were used: 1) an empirical model based on the reduced simple ratio vegetation index and 2) inversion of a forest reflectance model. Field measurements of LAI from over 1000 plots were used to develop and validate the models. The empirical model using scene-specific parameters produced LAI maps for 2000 and 2006 with realistic patterns, though some overestimation. Comparison to MODIS LAI showed generally good agreement but some systematic differences likely due to phenological and calibration issues. The empirical method was selected for producing the LAI maps.
EcoTas13 BradEvans e-MAST
eMAST aims to integrate data from TERN and other sources to model ecosystems at all scales in Australia from 2013-2015. This will be done using data assimilation, model evaluation and optimization tools to further ecosystem science and help address questions about topics like carbon, water, climate change, fire, and biodiversity. Key products being delivered include high resolution climate and productivity datasets as well as tools for data analysis, interpolation and modeling. Progress includes the development and delivery of ANUClimate climate datasets and the ePiSaT model for estimating primary productivity across Australia using flux tower and satellite data.
MarkSim GCM: generating plausible weather data for future climates
MarkSim GCM: generating plausible weather data for future climatesCCAFS | CGIAR Research Program on Climate Change, Agriculture and Food Security
CCAFS workshop titled "Using Climate Scenarios and Analogues for Designing Adaptation Strategies in Agriculture," 19-23 September in Kathmandu, Nepal. Recommended
SURF_Poster_REU2016Team1
1) Researchers used a technique called Minimum Average Deviance Estimation (MADE) to reduce the dimensionality of historical climate and rainfall data from the Missouri River Basin in order to retrospectively predict rainfall.
2) They tested MADE on simulated data and found that its prediction error decreased as the number of observations and covariates increased. They also found MADE outperformed other estimators when the number of observations was highest.
3) The researchers then applied MADE to over 18,000 observations of climate variables and precipitation measurements from 1949-2005 in the Missouri River Basin. Using a subset of 5 years of data from a single location, MADE was able to reduce the data dimensions from 5 to 1 while achieving a mean
Absorption factor method of radiation
This document describes the absorption factor method of radiation. It involves calculating the absorption factor Bij, which is the fraction of emitted energy from surface i that is absorbed by surface j. An algebraic equation is given relating Bij and the emitted energy Ei. The equations for determining Bij are linear and can be solved using various matrix methods. Once calculated, the absorption factors do not depend on surface temperature and can be used to determine radiative transport for a given geometry under different thermal conditions by writing new energy balance equations. However, the absorption factor method requires solving a larger system of equations than the more commonly used radiosity method.
Leaf Area Index (LAI) in the quantification of vegetation disturbance in Iris...
Leaf Area Index (LAI) in the quantification of vegetation disturbance in Iris...Environmental Protection Agency, Ireland
Dr Jerome O Connell - presentation made at various conferences throughout Europe as part of PhD which was funded by the EPA under the STRIVE Research Programme 2007-2013 (2007-PhD-ET-2)Presentation Boers - Trends in Cloudiness
This document discusses trends in cloudiness and challenges in detecting trends from observational data. It notes that precision in monitoring cloud cover has increased over time but accuracy of long-term trends is difficult to determine due to changes in observation methods, locations, and instruments. Global satellite data shows small or no statistically significant changes in total cloud cover over various time periods. Trends over the Netherlands cannot be reliably assessed due to gaps in surface observer records when instrumentation changed. Reconstructing aerosol and fog trends requires long records of relative humidity and visibility. The public may have difficulty distinguishing demonstrated climate trends from spurious trends caused by changes in observations. Successful experts in climate monitoring require persistence, attention to detail, and focus on continuity of measurements over
Spatial presentation of prognosis models in plant protection
This document discusses the development and use of risk maps for plant protection models. Models are enhanced using geostatistics to interpolate meteorological data across 1,367 virtual stations, allowing risk maps with grids of 1 km2. These maps ease the interpretation of model results across large regions. New technologies bring high resolution spatial meteorological data and mobile internet access to enable location-based decision support services online. Risk maps and online access optimize the provision of millions of model runs to farmers and advisors.
GENERATING FINE RESOLUTION LEAF AREA INDEX MAPS FOR BOREAL FORESTS OF FINLAND...
This document summarizes research generating fine-resolution leaf area index (LAI) maps for boreal forests in Finland using satellite imagery. Two methods were used: 1) an empirical model based on the reduced simple ratio vegetation index and 2) inversion of a forest reflectance model. Field measurements of LAI from over 1000 plots were used to develop and validate the models. The empirical model using scene-specific parameters produced LAI maps for 2000 and 2006 with realistic patterns, though some overestimation. Comparison to MODIS LAI showed generally good agreement but some systematic differences likely due to phenological and calibration issues. The empirical method was selected for producing the LAI maps.
EcoTas13 BradEvans e-MAST
eMAST aims to integrate data from TERN and other sources to model ecosystems at all scales in Australia from 2013-2015. This will be done using data assimilation, model evaluation and optimization tools to further ecosystem science and help address questions about topics like carbon, water, climate change, fire, and biodiversity. Key products being delivered include high resolution climate and productivity datasets as well as tools for data analysis, interpolation and modeling. Progress includes the development and delivery of ANUClimate climate datasets and the ePiSaT model for estimating primary productivity across Australia using flux tower and satellite data.
MarkSim GCM: generating plausible weather data for future climates
MarkSim GCM: generating plausible weather data for future climatesCCAFS | CGIAR Research Program on Climate Change, Agriculture and Food Security
CCAFS workshop titled "Using Climate Scenarios and Analogues for Designing Adaptation Strategies in Agriculture," 19-23 September in Kathmandu, Nepal. vectorrasterQGIS
This document discusses the differences between raster and vector data in QGIS. Raster data represents images as a grid of pixels, while vector data represents geographic features as points, lines, or polygons. The document then summarizes a project that used raster data to model potential flooding in Jacksonville, Florida under different levels of sea level rise predicted by the IPCC. Raster data was suitable for this project because it allowed easy transformation of elevation data but resulted in a pixelated map at the edges.
Chris Spry
This document presents a storm water modeling project for the Town Creek Watershed in Salisbury, NC using the Curve Number Runoff Method. Land use in the watershed was modeled between 1978 and 2009 using aerial imagery and satellite data. Results found forest and open space decreased by 3% and 5% respectively, while urban area increased by 8%. Storm water runoff increased from 78,592 cubic feet in 1978 to 92,131 cubic feet in 2009 based on a 4 inch rainfall. While the Curve Number method provided useful insights, it may be better suited to areas with greater elevation changes and less urbanization.
Chris Spry
This document presents a storm water modeling project for the Town Creek Watershed in Salisbury, NC using the Curve Number Runoff Method. Land use in the watershed was modeled between 1978 and 2009 using aerial imagery and satellite data. Results found forest and open space decreased by 3% and 5% respectively, while urban area increased by 8%. Storm water runoff increased from 78,592 cubic feet in 1978 to 92,131 cubic feet in 2009 based on a 4 inch rainfall. While the Curve Number method provided useful insights, it may be better suited to areas with greater elevation changes and less urbanization.
Chris Spry
This document summarizes a storm water modeling project for the Town Creek Watershed in Salisbury, NC that used the Curve Number Runoff Method to model land use change between 1978 and 2009 and calculate the resulting changes in storm water runoff. Land use data from aerial imagery and satellite images from 1978 and 2009 were incorporated into a GIS system along with digital elevation models and soil data. The results found increases in urban land use of 8% along with decreases in forest and open space, and a corresponding increase in storm water runoff accumulation at a point in the watershed. The project reinforced the impact of urbanization in increasing storm water runoff but encountered issues applying the method in an already urbanized environment.
FR4.L10.2: A MICROWAVE SCATTERING MODEL OF VEGETATED SURFACES BASED ON BOR/DD...
This document describes a microwave scattering model for vegetated surfaces being developed for NASA's Soil Moisture Active Passive (SMAP) mission. The model uses numerical solutions of Maxwell's equations to simulate scattering from both rough soil surfaces and vegetation volumes. It represents surfaces and vegetation as 3D "data cubes" that allow fast interpolation. Comparison to experimental data from a grassland site shows good agreement between modeled and measured backscatter values. The model will be used to retrieve soil moisture from SMAP radar measurements by inverting the physical scattering model.
3D Analyst - Lake Lorelindu by GRASS
This document describes using the open source GIS software GRASS to calculate the volume of Lore Lindu Lake in Central Sulawesi, Indonesia. It details the steps taken, which include importing polygon and depth point data, creating a bathymetry map using inverse distance weighted interpolation, generating contour and slope maps, viewing the results in 3D, and calculating the lake volume as 143449194.374 cubic meters using the r.volume command. The document concludes that GRASS provides an accessible and flexible way to perform this hydrological analysis and 3D volume calculation.
CSpryFinal5
This document presents a storm water modeling project for the Town Creek Watershed in Salisbury, NC using the Curve Number Runoff Method. Land use in the watershed was mapped from aerial imagery from 1978 and satellite imagery from 2009. The Curve Number Runoff equation was used to calculate storm water runoff for a 4-inch rainfall event based on the different land uses and soil types. Results found an 8% increase in urban land use from 1978 to 2009, corresponding increases in storm water runoff, and that while the Curve Number method provided useful insights, it may be better suited to less urbanized areas with greater elevation changes.
3D Analyst - Watershed Lorelindu
This document summarizes steps taken to create a geographic information system (GIS) model of watershed areas in Lore Lindu National Park, Indonesia. The steps included: (1) creating a digital elevation model from satellite data, (2) removing errors, (3) generating flow direction and accumulation maps, (4) mapping stream channels based on areas with over 500 cells of accumulated flow, (5) generating stream links, (6) delineating watersheds, (7) visualizing the 3D model, and (8) comparing generated streams to topography.
Undergraduate Modeling Workshop - Southeastern US Rainfall Working Group Fina...
Undergraduate Modeling Workshop - Southeastern US Rainfall Working Group Fina...The Statistical and Applied Mathematical Sciences Institute
In this project the group members will play with daily rainfall data collected in Gulf coast (535stations in total) from 1949 to 2017. The purposes of this exercise are to:
1) to give students an idea of a typical example of a climate data set (spatio-temporal data) and someassociated scientific questions (e.g. how rainfall extremes vary in space and time and how that mightbe affected by other things like greenhouse gases or temperatures).
2) to get students familiar with data analysis using R including data manipulation, data visualization, and data summary.
3) to introduce some statistical methods (e.g. time series analysis, spatial statistics, extreme value analysis) to analyze this kind of data to "answer" (perform statistical inference) the questions of interest.
Group members: Lin Ge, Jianan Jang, Jessica Robinson, Erin Song, Seth Temple, Adam Wu
IGARSS2011(OkiKazuo110726).ppt
This study estimates river discharge using MODIS satellite images. Due to the coarse spatial resolution of MODIS, reflectance values are extracted from a pixel near the river mouth rather than directly over the measurement point. Regression analysis is performed between reflectance and in situ discharge measurements taken on the same days from the Naka and Monobe Rivers in Japan in 2004. Results show the method can effectively estimate discharge in narrow rivers from MODIS data, with root mean square errors of 213 and 199 m3/s for the two rivers. Monthly and annual averages are also estimated with reasonable accuracy.
3D Analyst - Lake, Jatiluhur
This document summarizes finding the volume of water in Jatiluhur Reservoir using 3D Analyst in ArcGIS. It details importing elevation data, creating an area of interest, cropping the data to the reservoir boundaries, and using 3D Analyst tools to calculate volume statistics. The volume of water held below the reservoir surface was computed to understand how much water it is holding.
Change detection of forest fire in los angeles
This document summarizes a study that used Landsat-5 TM satellite imagery to detect changes in forest areas caused by a large wildfire in Los Angeles, California in 2009. The methodology used the Normalized Difference Vegetation Index (NDVI) and both unsupervised and supervised image classification techniques to analyze imagery from before and after the fire. Change detection analysis between 2007 and 2009 clearly identified locations affected by the forest fire and showed that multi-spectral Landsat data can effectively delineate forest fire impacts.
ÖNCEL AKADEMİ: INTRODUCTION TO SEISMOLOGY
The document discusses earthquake occurrence and catalogs. It introduces the Gutenberg-Richter law which states that the number of earthquakes decreases exponentially with magnitude. Complete earthquake catalogs are desirable as they are homogeneous, complete, cover a long duration, and source material is available online. Stable continental regions have low seismicity that follows the Gutenberg-Richter relationship, with an earthquake of magnitude 6.5 expected once a decade for a particular region. Students are assigned homework to analyze earthquake data from Saudi Arabia and establish the Gutenberg-Richter relationship for the region.
Rainfall Erosivity in China
Mr Yun Xie, Beijing Normal University, China. Global Symposium on Soil Erosion (GSER19), 15 - 17 May 2019 at FAO HQ.
Nichols, Meredith_Validation of Precipitation and Lightning Observations
This study analyzes the performance of five satellite-derived precipitation estimation products relative to rain-gauge and radar observations over the continental United States from 2010-2014. The products generally overestimate precipitation during the summer and CMORPH underestimates during winter. Infrared-based products tend to overestimate the most, while CMORPH and Hydro-Estimator have higher correlations. Future research will use lightning data to help validate satellite estimates in convective and non-convective regions.
Use of the German Weather Services KLAM Model to Investigate the Cold Air Dra...
The document discusses the use of the German Weather Services KLAM_21 model to simulate cold air drainage flows in valleys. The single-layer model predicts the depth, heat deficit, velocity and direction of cold air flows based on surface heat loss, advection, momentum equations and optional tracer dispersion. Key inputs are terrain, land use and local heat loss rates. The model is best suited for idealized simulations under fair weather conditions due to simplifications but has been validated and applied for air quality and frost protection assessments.
Fire Poster
This document summarizes the use of ArcMap and FARSITE software to model a 2010 wildfire in Kentucky's Red River Gorge under both historical conditions with rain and a hypothetical scenario without rain. Key inputs to FARSITE included landscape data from LANDFIRE on aspect, elevation, slope, canopy cover and fuel models. Weather data from Kentucky Mesonet was also used. The historical simulation showed fire stagnation during a rain event, while the no-rain scenario resulted in faster fire spread over a larger 2.12 square mile area compared to 0.62 square miles with rain. While the model has limitations, it demonstrates that larger fires are possible in the future if hotter and drier conditions occur in the Daniel Bo
Gutierrez_E_Lab2
This document describes procedures for analyzing terrain and solar insolation for different vegetation types in the Santa Monica Mountains using GIS tools. The key findings are that chaparral exists at the highest elevations and steepest slopes, receiving the lowest solar insolation, while urban agriculture exists at the lowest elevations and receives the highest insolation. Annual grass has a relatively high slope given its low mean elevation. The analysis finds patterns between vegetation types and terrain/solar factors as expected.
Flood Risk Zonation for Karaj, Ezatollah GHANAVATI
6th International Disaster and Risk Conference IDRC 2016 Integrative Risk Management - Towards Resilient Cities. 28 August - 01 September 2016 in Davos, Switzerland
TREE BIOMASS ESTIMATION USING AIRBORNE LASER SCANNING
The document summarizes a study that used airborne laser scanning (ALS) to estimate tree biomass. It describes the methodology which involved: (1) preprocessing ALS data to classify ground and aboveground points and create a digital terrain model, (2) detecting individual trees, (3) calculating point cloud metrics to describe tree crowns, and (4) developing linear regression models to predict aboveground biomass using the metrics. The results showed the metrics had strong correlations with field-measured biomass, and the models improved biomass estimation accuracy with adjusted R2 values from 0.74-0.86.
Comparison of Spatial Interpolation Methods for Precipitation in Ningxia, China
Comparison of Spatial Interpolation Methods for Precipitation in Ningxia, ChinaInternational Journal of Science and Research (IJSR)
As basic data, the reliability of precipitation data makes a significant impact on many results of environmental applications. In order to obtain spatially distributed precipitation data, measured points are interpolated. There are many spatial interpolation schemes, but none of them can perform best in all cases. So criteria of precision evaluation are established. This study aims to find an optimal interpolation scheme for rainfall in Ningxia. The study area is located in northwest China. Meteorological stations distribute at a low density here. Six interpolation methods have been tested after exploring data. Cross-validation was used as the criterion to evaluate the accuracy of various methods. The best results were obtained by cokriging with elevation as the second variable, while the inverse distance weighting (IDW) preform worst. Three types of model in cokriging were compared, and Gaussian model is the best.
Coweeta ppt cd_ms
From all comments received by the LTER Network Office, this year's meeting in Estes Park was a complete success! The 2015 LTER All Scientists Meeting was held from August 30 through the evening of September 2. The Conference was organized around the theme: " From Long-Term Data to Understanding: Toward a Predictive Ecology". Almost 600 people attended the meeting. There were over 300 poster presentations and more than 75 formal and ad-hoc working group meetings. Drs. James Olds, Diana Wall, Knute Nadelhoffer, Ned Gardener and Christine O'Connell provided excellent plenary presentations to highlight the meeting. Chloe Wardropper (NTL) won 1st place in the student poster competition with Alexandra Conway (BNZ), Shinjini Goswami (HBR), Hafsah Nahrawi (GCE) and Bonnie McGill (KBS) winning runner-up awards.
More Related Content
What's hot
vectorrasterQGIS
This document discusses the differences between raster and vector data in QGIS. Raster data represents images as a grid of pixels, while vector data represents geographic features as points, lines, or polygons. The document then summarizes a project that used raster data to model potential flooding in Jacksonville, Florida under different levels of sea level rise predicted by the IPCC. Raster data was suitable for this project because it allowed easy transformation of elevation data but resulted in a pixelated map at the edges.
Chris Spry
This document presents a storm water modeling project for the Town Creek Watershed in Salisbury, NC using the Curve Number Runoff Method. Land use in the watershed was modeled between 1978 and 2009 using aerial imagery and satellite data. Results found forest and open space decreased by 3% and 5% respectively, while urban area increased by 8%. Storm water runoff increased from 78,592 cubic feet in 1978 to 92,131 cubic feet in 2009 based on a 4 inch rainfall. While the Curve Number method provided useful insights, it may be better suited to areas with greater elevation changes and less urbanization.
Chris Spry
This document presents a storm water modeling project for the Town Creek Watershed in Salisbury, NC using the Curve Number Runoff Method. Land use in the watershed was modeled between 1978 and 2009 using aerial imagery and satellite data. Results found forest and open space decreased by 3% and 5% respectively, while urban area increased by 8%. Storm water runoff increased from 78,592 cubic feet in 1978 to 92,131 cubic feet in 2009 based on a 4 inch rainfall. While the Curve Number method provided useful insights, it may be better suited to areas with greater elevation changes and less urbanization.
Chris Spry
This document summarizes a storm water modeling project for the Town Creek Watershed in Salisbury, NC that used the Curve Number Runoff Method to model land use change between 1978 and 2009 and calculate the resulting changes in storm water runoff. Land use data from aerial imagery and satellite images from 1978 and 2009 were incorporated into a GIS system along with digital elevation models and soil data. The results found increases in urban land use of 8% along with decreases in forest and open space, and a corresponding increase in storm water runoff accumulation at a point in the watershed. The project reinforced the impact of urbanization in increasing storm water runoff but encountered issues applying the method in an already urbanized environment.
FR4.L10.2: A MICROWAVE SCATTERING MODEL OF VEGETATED SURFACES BASED ON BOR/DD...
This document describes a microwave scattering model for vegetated surfaces being developed for NASA's Soil Moisture Active Passive (SMAP) mission. The model uses numerical solutions of Maxwell's equations to simulate scattering from both rough soil surfaces and vegetation volumes. It represents surfaces and vegetation as 3D "data cubes" that allow fast interpolation. Comparison to experimental data from a grassland site shows good agreement between modeled and measured backscatter values. The model will be used to retrieve soil moisture from SMAP radar measurements by inverting the physical scattering model.
3D Analyst - Lake Lorelindu by GRASS
This document describes using the open source GIS software GRASS to calculate the volume of Lore Lindu Lake in Central Sulawesi, Indonesia. It details the steps taken, which include importing polygon and depth point data, creating a bathymetry map using inverse distance weighted interpolation, generating contour and slope maps, viewing the results in 3D, and calculating the lake volume as 143449194.374 cubic meters using the r.volume command. The document concludes that GRASS provides an accessible and flexible way to perform this hydrological analysis and 3D volume calculation.
CSpryFinal5
This document presents a storm water modeling project for the Town Creek Watershed in Salisbury, NC using the Curve Number Runoff Method. Land use in the watershed was mapped from aerial imagery from 1978 and satellite imagery from 2009. The Curve Number Runoff equation was used to calculate storm water runoff for a 4-inch rainfall event based on the different land uses and soil types. Results found an 8% increase in urban land use from 1978 to 2009, corresponding increases in storm water runoff, and that while the Curve Number method provided useful insights, it may be better suited to less urbanized areas with greater elevation changes.
3D Analyst - Watershed Lorelindu
This document summarizes steps taken to create a geographic information system (GIS) model of watershed areas in Lore Lindu National Park, Indonesia. The steps included: (1) creating a digital elevation model from satellite data, (2) removing errors, (3) generating flow direction and accumulation maps, (4) mapping stream channels based on areas with over 500 cells of accumulated flow, (5) generating stream links, (6) delineating watersheds, (7) visualizing the 3D model, and (8) comparing generated streams to topography.
Undergraduate Modeling Workshop - Southeastern US Rainfall Working Group Fina...
Undergraduate Modeling Workshop - Southeastern US Rainfall Working Group Fina...The Statistical and Applied Mathematical Sciences Institute
In this project the group members will play with daily rainfall data collected in Gulf coast (535stations in total) from 1949 to 2017. The purposes of this exercise are to:
1) to give students an idea of a typical example of a climate data set (spatio-temporal data) and someassociated scientific questions (e.g. how rainfall extremes vary in space and time and how that mightbe affected by other things like greenhouse gases or temperatures).
2) to get students familiar with data analysis using R including data manipulation, data visualization, and data summary.
3) to introduce some statistical methods (e.g. time series analysis, spatial statistics, extreme value analysis) to analyze this kind of data to "answer" (perform statistical inference) the questions of interest.
Group members: Lin Ge, Jianan Jang, Jessica Robinson, Erin Song, Seth Temple, Adam Wu
IGARSS2011(OkiKazuo110726).ppt
This study estimates river discharge using MODIS satellite images. Due to the coarse spatial resolution of MODIS, reflectance values are extracted from a pixel near the river mouth rather than directly over the measurement point. Regression analysis is performed between reflectance and in situ discharge measurements taken on the same days from the Naka and Monobe Rivers in Japan in 2004. Results show the method can effectively estimate discharge in narrow rivers from MODIS data, with root mean square errors of 213 and 199 m3/s for the two rivers. Monthly and annual averages are also estimated with reasonable accuracy.
3D Analyst - Lake, Jatiluhur
This document summarizes finding the volume of water in Jatiluhur Reservoir using 3D Analyst in ArcGIS. It details importing elevation data, creating an area of interest, cropping the data to the reservoir boundaries, and using 3D Analyst tools to calculate volume statistics. The volume of water held below the reservoir surface was computed to understand how much water it is holding.
Change detection of forest fire in los angeles
This document summarizes a study that used Landsat-5 TM satellite imagery to detect changes in forest areas caused by a large wildfire in Los Angeles, California in 2009. The methodology used the Normalized Difference Vegetation Index (NDVI) and both unsupervised and supervised image classification techniques to analyze imagery from before and after the fire. Change detection analysis between 2007 and 2009 clearly identified locations affected by the forest fire and showed that multi-spectral Landsat data can effectively delineate forest fire impacts.
ÖNCEL AKADEMİ: INTRODUCTION TO SEISMOLOGY
The document discusses earthquake occurrence and catalogs. It introduces the Gutenberg-Richter law which states that the number of earthquakes decreases exponentially with magnitude. Complete earthquake catalogs are desirable as they are homogeneous, complete, cover a long duration, and source material is available online. Stable continental regions have low seismicity that follows the Gutenberg-Richter relationship, with an earthquake of magnitude 6.5 expected once a decade for a particular region. Students are assigned homework to analyze earthquake data from Saudi Arabia and establish the Gutenberg-Richter relationship for the region.
Rainfall Erosivity in China
Mr Yun Xie, Beijing Normal University, China. Global Symposium on Soil Erosion (GSER19), 15 - 17 May 2019 at FAO HQ.
Nichols, Meredith_Validation of Precipitation and Lightning Observations
This study analyzes the performance of five satellite-derived precipitation estimation products relative to rain-gauge and radar observations over the continental United States from 2010-2014. The products generally overestimate precipitation during the summer and CMORPH underestimates during winter. Infrared-based products tend to overestimate the most, while CMORPH and Hydro-Estimator have higher correlations. Future research will use lightning data to help validate satellite estimates in convective and non-convective regions.
Use of the German Weather Services KLAM Model to Investigate the Cold Air Dra...
The document discusses the use of the German Weather Services KLAM_21 model to simulate cold air drainage flows in valleys. The single-layer model predicts the depth, heat deficit, velocity and direction of cold air flows based on surface heat loss, advection, momentum equations and optional tracer dispersion. Key inputs are terrain, land use and local heat loss rates. The model is best suited for idealized simulations under fair weather conditions due to simplifications but has been validated and applied for air quality and frost protection assessments.
Fire Poster
This document summarizes the use of ArcMap and FARSITE software to model a 2010 wildfire in Kentucky's Red River Gorge under both historical conditions with rain and a hypothetical scenario without rain. Key inputs to FARSITE included landscape data from LANDFIRE on aspect, elevation, slope, canopy cover and fuel models. Weather data from Kentucky Mesonet was also used. The historical simulation showed fire stagnation during a rain event, while the no-rain scenario resulted in faster fire spread over a larger 2.12 square mile area compared to 0.62 square miles with rain. While the model has limitations, it demonstrates that larger fires are possible in the future if hotter and drier conditions occur in the Daniel Bo
Gutierrez_E_Lab2
This document describes procedures for analyzing terrain and solar insolation for different vegetation types in the Santa Monica Mountains using GIS tools. The key findings are that chaparral exists at the highest elevations and steepest slopes, receiving the lowest solar insolation, while urban agriculture exists at the lowest elevations and receives the highest insolation. Annual grass has a relatively high slope given its low mean elevation. The analysis finds patterns between vegetation types and terrain/solar factors as expected.
Flood Risk Zonation for Karaj, Ezatollah GHANAVATI
6th International Disaster and Risk Conference IDRC 2016 Integrative Risk Management - Towards Resilient Cities. 28 August - 01 September 2016 in Davos, Switzerland
TREE BIOMASS ESTIMATION USING AIRBORNE LASER SCANNING
The document summarizes a study that used airborne laser scanning (ALS) to estimate tree biomass. It describes the methodology which involved: (1) preprocessing ALS data to classify ground and aboveground points and create a digital terrain model, (2) detecting individual trees, (3) calculating point cloud metrics to describe tree crowns, and (4) developing linear regression models to predict aboveground biomass using the metrics. The results showed the metrics had strong correlations with field-measured biomass, and the models improved biomass estimation accuracy with adjusted R2 values from 0.74-0.86.
What's hot (20)
FR4.L10.2: A MICROWAVE SCATTERING MODEL OF VEGETATED SURFACES BASED ON BOR/DD...
FR4.L10.2: A MICROWAVE SCATTERING MODEL OF VEGETATED SURFACES BASED ON BOR/DD...
Undergraduate Modeling Workshop - Southeastern US Rainfall Working Group Fina...
Undergraduate Modeling Workshop - Southeastern US Rainfall Working Group Fina...
Nichols, Meredith_Validation of Precipitation and Lightning Observations
Nichols, Meredith_Validation of Precipitation and Lightning Observations
Use of the German Weather Services KLAM Model to Investigate the Cold Air Dra...
Use of the German Weather Services KLAM Model to Investigate the Cold Air Dra...
Flood Risk Zonation for Karaj, Ezatollah GHANAVATI
Flood Risk Zonation for Karaj, Ezatollah GHANAVATI
TREE BIOMASS ESTIMATION USING AIRBORNE LASER SCANNING
TREE BIOMASS ESTIMATION USING AIRBORNE LASER SCANNING
Similar to SURF_Poster_REU2016Team3
Comparison of Spatial Interpolation Methods for Precipitation in Ningxia, China
Comparison of Spatial Interpolation Methods for Precipitation in Ningxia, ChinaInternational Journal of Science and Research (IJSR)
As basic data, the reliability of precipitation data makes a significant impact on many results of environmental applications. In order to obtain spatially distributed precipitation data, measured points are interpolated. There are many spatial interpolation schemes, but none of them can perform best in all cases. So criteria of precision evaluation are established. This study aims to find an optimal interpolation scheme for rainfall in Ningxia. The study area is located in northwest China. Meteorological stations distribute at a low density here. Six interpolation methods have been tested after exploring data. Cross-validation was used as the criterion to evaluate the accuracy of various methods. The best results were obtained by cokriging with elevation as the second variable, while the inverse distance weighting (IDW) preform worst. Three types of model in cokriging were compared, and Gaussian model is the best.
Coweeta ppt cd_ms
From all comments received by the LTER Network Office, this year's meeting in Estes Park was a complete success! The 2015 LTER All Scientists Meeting was held from August 30 through the evening of September 2. The Conference was organized around the theme: " From Long-Term Data to Understanding: Toward a Predictive Ecology". Almost 600 people attended the meeting. There were over 300 poster presentations and more than 75 formal and ad-hoc working group meetings. Drs. James Olds, Diana Wall, Knute Nadelhoffer, Ned Gardener and Christine O'Connell provided excellent plenary presentations to highlight the meeting. Chloe Wardropper (NTL) won 1st place in the student poster competition with Alexandra Conway (BNZ), Shinjini Goswami (HBR), Hafsah Nahrawi (GCE) and Bonnie McGill (KBS) winning runner-up awards.
Probalistic assessment of agriculture
This document summarizes a study that developed a probabilistic model for assessing agricultural droughts using graphical models. The study used hidden Markov models (HMMs) to relate unobserved drought states to observed crop water stress levels. HMMs were estimated using soil moisture and crop data at different spatial scales in Indiana. Results showed drought classifications varied by location and resolution. The probabilistic approach provided uncertainty estimates not available from other drought indices and identified additional drought events. Weekly analyses revealed complex temporal dependencies requiring advanced HMM frameworks.
Soil erosion
This document provides information about a study on the morphometry and soil loss estimation of the Naviluthirtha watershed in India. It includes the following key details:
- The objectives of the study are to conduct morphometric analysis, hypsometric analysis, and estimate soil loss using the Universal Soil Loss Equation.
- The study area is the Naviluthirtha watershed near Soudatti, Karnataka, India, which covers 2240 square km and contains three sub-watersheds.
- Methodology includes morphometric analysis of linear, aerial and relief aspects, hypsometric analysis to understand elevation distribution, and using the USLE model to estimate soil loss based on rainfall,
MCP_ES_2012_Jie
This document presents a hybrid Measure-Correlate-Predict (MCP) method for wind resource assessment that combines predictions from multiple nearby meteorological stations. The existing MCP methods only use data from one reference station and do not consider distance or elevation differences between stations. The new hybrid MCP method assigns weights to individual MCP predictions based on distance and elevation differences to the target site. It was evaluated using stations in North Dakota and showed improved accuracy over individual MCP methods based on error metrics and predicted wind farm power generation. The hybrid approach more accurately characterized the long-term wind distribution at the target site.
Application of Bayesian Regularized Neural Networks for Groundwater Level Mo...
Application of Bayesian Regularized Neural Networks for Groundwater Level Mo...Dr. Amir Mosavi, PhD., P.Eng.
This document presents a study that uses Bayesian Regularized Neural Networks (BRNN) to model groundwater levels in the Mahabad aquifer in Iran. The study area and data collection process are described. Five factors - precipitation, evaporation, temperature, streamflow, and previous month's groundwater level - are used as inputs to the BRNN model to estimate current groundwater levels. The results show the BRNN model performs excellently with low errors and high accuracy and determination values. Previous month's groundwater level and streamflow are found to be the most important predictors of current groundwater levels.Determining reservoir outflow using machine learning techniques.pptx
Determining reservoir outflow using machine learning techniques
Air quality challenges and business opportunities in China: Fusion of environ...
MMEA (The Measurement, Monitoring and Environmental Efficiency Assessment) research program final seminar presentation by Dr. Ari Karppinen, Finnish Meteorological Institute
Aquifer recharge from flash floods in the arid environment: A mass balance ap...
Estimation of the infiltration/natural recharge to groundwater from rainfall is an important issue in hydrology, particularly in arid regions. This paper proposes the application of The Natural Resources Conservation Service (NRCS) mass balance model to develop infiltration (F)–rainfall (P) relationship from flash flood events. Moreover, the NRCS method is compared with the rational and the Ф-index methods to investigate the discrepancies between these methods. The methods have been applied to five gauged basins and their 19 sub-basins (representative basins with detailed measurements) in the southwestern part of Saudi Arabia with 161 storms recorded in 4 years. The F–P relationships developed in this study based on NRCS method are: F = 39% P with R2 = 0.932 for the initial abstraction factor, λ = 0.2. However, F = 77% P with R2 = 0.986 for λ = 0.01. The model at λ = 0.01 is the best to fit the data, therefore, it is recommended to use the formula at λ = 0.01. The results show that the NRCS model is appropriate for the estimation of the F–P relationships in arid regions when compared with the rational and the Ф index methods. The latter overestimates the infiltration because they do not take λ into account. There is no significant difference between F–P relationships at different time scales. This helps the prediction of infiltration rates for aquifer recharge at ungauged basins from monthly and annual rainfall data with a single formula.
Mauro Sulis
Assessment of the catchment-scale energy and water balance using fully coupled simulations and observations
urpl969-group2-paper-03May06
The document summarizes research on land subsidence and elevation changes in Orleans Parish, Louisiana after Hurricane Katrina in 2005. It describes how the researchers used historical elevation data from control points between 1951-1991 to create a model predicting 50 years of subsidence through kriging interpolation. They generated an elevation surface map showing predicted subsidence across the parish. The surface was created in ArcGIS using a spherical semivariogram model, with assumptions like a constant subsidence rate, and could be improved by considering additional influence factors.
Determination of homogenous regions in the Tensift basin (Morocco).
The aim of this study is to determine homogenous region in the Tensift basin within which the hydrological behavior is similar. In order to do this we used two methods: The Principal components analysis on the monthly precipitation registered at the 23 rainfall stations. This resulted in setting apart 4 groups of stations. The second method is analysis of land use map, geological map, pedagogical map, vegetation map and slope map of the studied area. This method allowed us to delineate 4 homogenous areas. The two methods yielded complementary results and the superposition of groups and regions obtained allowed us to retain 4 homogenous regions corresponding to 3 groups of stations.
Monsoon Rainfall Forecast+.ppt
This document summarizes research on monsoon rainfall forecasting in India. It discusses:
1) The importance of monsoon prediction and approaches to long-term and short-term forecasting. Long-term prediction models use statistical correlations with ocean and atmospheric parameters, while short-term relies on numerical weather prediction models.
2) Factors used in the Indian Meteorological Department's long-term statistical forecasts in March/April and May/June, which include sea surface temperatures and pressures.
3) Evidence that short-term daily rainfall shows a scale-invariant power law distribution, making it difficult to predict precisely at a single location but easier when averaged over multiple locations.
4) The use of
APPLICATION OF KRIGING IN GROUND WATER STUDIES
The document discusses the application of kriging in groundwater studies. Kriging is a geostatistical technique used to interpolate the value of a random field between known data points. It provides the best linear unbiased prediction and honors the observed spatial structure of the data. Two case studies are summarized that demonstrate how kriging can be used to generate groundwater level contour maps and correlate declining water levels with land cover changes detected from satellite images. The studies show that kriging produces more accurate representations of spatial variability in groundwater compared to other methods.
Remote sensing in agriculture
In India, agriculture is one of the major application areas of the remote sensing technology. Various national level agricultural applications have been developed which showcases the use of remote sensing data provided by the sensors/satellites launched by the country’s space agency, Indian Space Research Organisation (ISRO)
Remote sensing in agriculture
In India, agriculture is one of the major application areas of the remote sensing technology. Various national level agricultural applications have been developed which showcases the use of remote sensing data provided by the sensors/satellites launched by the country’s space agency, Indian Space Research Organisation (ISRO)
CSU_Poster
This research summarizes measurements taken during the COPE-MED field campaign to better understand warm rain processes and entrainment effects on heavy precipitation. LWC probes were compared and generally agreed well, though the PVM overestimated LWC at higher concentrations and smaller diameters. A LWC survey with vertical statistics was conducted. Analysis of droplet spectra bimodality from low-precipitation cases found evidence of bimodality but secondary activation was unlikely the cause, with the bimodality mechanism remaining unclear. Future COPE analysis will utilize these LWC and bimodality findings to evaluate hypotheses regarding warm rain processes and entrainment impacts on heavy precipitation.
September 1 - 1116 - Tassia Brighenti and Phillip Gassman
This document summarizes a study that improved ecohydrological simulations in central Iowa watersheds by incorporating more accurate tile drainage and fertilizer application rate data into the SWAT model. The updated models showed better representation of streamflows, crop yields, and water balances compared to default models. Refining inputs like tile drainage maps and fertilizer rates improved model performance at over half of monitoring stations. The results highlight the importance of selecting accurate management data to realistically simulate baseline hydrologic and water quality conditions.
Thesis Presentation
1) The document analyzes weather window probabilities for offshore wind and wave conditions at three sites in Scotland that are relevant for marine renewable energy development.
2) It determines that wind speed follows a two-parameter Weibull distribution while significant wave height follows a three-parameter Gamma distribution.
3) The maximum likelihood estimation method produced probabilities closest to the empirical probabilities determined from 31 years of wind and wave data, providing evidence that the distributions are suitable for representing conditions at the sites.
Master's course defense presentation in Water Resource Management and GIS
1) The document summarizes a master's thesis that used the SWAT hydrological model within a GIS to simulate surface runoff in the Panjshir watershed in Afghanistan.
2) Key results included monthly and daily surface runoff predictions that matched observed discharge data with R2 values of 0.815 for calibration and 0.817 for validation.
3) The study also found that land use changes between 1993-2010 increased total water yield in the watershed, with average annual changes ranging from 1.2-4.5% between scenarios.
Similar to SURF_Poster_REU2016Team3 (20)
Comparison of Spatial Interpolation Methods for Precipitation in Ningxia, China
Comparison of Spatial Interpolation Methods for Precipitation in Ningxia, China
Application of Bayesian Regularized Neural Networks for Groundwater Level Mo...
Application of Bayesian Regularized Neural Networks for Groundwater Level Mo...
Determining reservoir outflow using machine learning techniques.pptx
Determining reservoir outflow using machine learning techniques.pptx
Air quality challenges and business opportunities in China: Fusion of environ...
Air quality challenges and business opportunities in China: Fusion of environ...
Aquifer recharge from flash floods in the arid environment: A mass balance ap...
Aquifer recharge from flash floods in the arid environment: A mass balance ap...
Determination of homogenous regions in the Tensift basin (Morocco).
Determination of homogenous regions in the Tensift basin (Morocco).
September 1 - 1116 - Tassia Brighenti and Phillip Gassman
September 1 - 1116 - Tassia Brighenti and Phillip Gassman
Master's course defense presentation in Water Resource Management and GIS
Master's course defense presentation in Water Resource Management and GIS
SURF_Poster_REU2016Team3
- 1. Dimensionality Reduction Using Sliced Inverse Regression In Modeling Large Climate Data UMBC REU Site: Interdisciplinary Program in High Performance Computing Ross Flieger-Allison1 , Lois Miller2 , Danielle Sykes3 , Pablo Valle4 RAs: Sai K. Popuri3 , Nadeesri Wijekoon3 Faculty: Nagaraj K. Neerchal3 Client: Amita Mehta5 1 Williams College 2 DePauw University 3 UMBC 4 Kean University 5 JCET • The Missouri River Basin (MRB) is a significant agricultural region that is not irrigated and thus dependent on rainfall • Precipitation predictions are histor- ically inaccurate due to the semi- continuous nature of observed precip- itation data Figure 1: The Missouri Rivier Basin Background • Precipitation data is semi-continuous, meaning many observations are equal to 0 and all other observations are positive and follow a continuous dis- tribution • Predictions are calculated using ob- served data and data from MIROC5, a Global Climate Model (GCM), across 57 years (1949-2005) from over 270,000 locations • Precipitation at any given location s depends on a large number of co- variates: current and past values of monthly precipitation, sea-level pres- sure, relative humidity, and max- imum/minimum temperatures at s and its neighboring locations Data • Sliced Inverse Regression (SIR) is a data-analytic tool that can be used to reduce the dimensionality of a large set of covariates • Nadaraya-Watson Estimator (NWE) is a simple non-paramentric smooth- ing regression method: ˆmx = n i=1 Kh(x−xi)yi n i=1 Kh(x−xi) • The NWE is adapted to account for the semi-continous nature of predic- tions Methodology • The following figures show a comparison between observed precipitation data, and their predictions Figure 2: Observed and predicted monthly precipitation, including the proportion of 0 values for both Figure 3: Mean Squared Error from positive predictions for months with positive rainfall (true positives) Results Table 1: Performance Study Results Processes Subregion∗ MRB 1 01:14:24 N/A 4 00:19:07 09:54:22 8 00:11:04 05:20:31 16 00:05:58 02:49:49 32 00:03:13 04:25:51 64 N/A 03:41:26 ∗ denotes a region from latitutude -101 to -97 and longitude 39.25 to 42.95 Figure 4: Observed speedup (Sp = (T1/Tp)) of parallelization Parallelization • We have successfully demonstrated that SIR and NWE methods can be implemented to work on a large dataset • Parallelization of SIR and NWE code greatly increases computational effi- ciency on the subregion, and also im- proves efficiency for the entire MRB region, up to 16 processes Conclusions [1] National Integrated Drought Information Systems. https://www.drought.gov/ drought/dews/missouri-river-basin/ about. [2] S. K. Popuri, K. P. Adragni, et al., “Spatio- Temporal Analysis of Daily Precipitation via a Sufficient Dimension Reduction.” (In Progress). [3] Full Technical Report: HPCF-2016-13 hpcf. umbc.edu > Publications REU Site: hpcreu.umbc.edu NSF, NSA, DOD, UMBC, HPCF, CIRC References & Acknowledgments