Modeling the hydrologically-relevant Modeling the hydrologically-relevant features of uncertainty of space-borne high features of uncertainty of space-borne high

resolution precipitation productsresolution precipitation products

LING TANGLING TANG Tennessee Technological UniversityTennessee Technological University

Department of Civil and Environmental Department of Civil and Environmental EngineeringEngineering

Preliminary Assessment for PhDPreliminary Assessment for PhD

OutlineOutline

• IntroductionIntroduction

• Research ObjectivesResearch Objectives

• Fundamental Research HypothesisFundamental Research Hypothesis

• Proposed Research Methodology Proposed Research Methodology

• Anticipated Broader Impact of the Planned ProjectAnticipated Broader Impact of the Planned Project

• My Research Background and PreparationMy Research Background and Preparation

IntroductionIntroduction

• The Global Precipitation The Global Precipitation

Measurement (GPM) MissionMeasurement (GPM) Mission - One of the next generation of - One of the next generation of

satellite-based Earth science satellite-based Earth science missions that will study global missions that will study global precipitation (rain,snow and ice).precipitation (rain,snow and ice).

- Currently scheduled for launch in - Currently scheduled for launch in 2013.2013.

(source from: gpm.gsfc.nasa.gov)

• Critical IssuesCritical Issues

- In-situ rainfall measuring networks are on the decline. In-situ rainfall measuring networks are on the decline. GPM will be able to demonstrate tangible benefits in GPM will be able to demonstrate tangible benefits in hydrologic application over ungauged regions.hydrologic application over ungauged regions.

- Knowledge of satellite rainfall estimation uncertainty is Knowledge of satellite rainfall estimation uncertainty is required – required – the Paradoxthe Paradox

- Scale mis-match between overland hydrologic processes Scale mis-match between overland hydrologic processes that evolve at smaller and more dynamic scales (i.e., < 1 that evolve at smaller and more dynamic scales (i.e., < 1 hour and < 1 km) and operational satellite precipitation hour and < 1 km) and operational satellite precipitation forcing datasets that are available at coarser scalesforcing datasets that are available at coarser scales

- Representing the error structure of satellite rainfall Representing the error structure of satellite rainfall as a function of scale against quality-controlled as a function of scale against quality-controlled ground validation (GV) datasets is both a critical ground validation (GV) datasets is both a critical research problem and an application need.research problem and an application need.

- Satellite rainfall will be most useful over ungauged Satellite rainfall will be most useful over ungauged (non-GV) regions – (non-GV) regions – so how can we generate so how can we generate uncertainty estimates for satellite data over those uncertainty estimates for satellite data over those regions ?regions ?

• The Unresolved ParadoxThe Unresolved Paradox

Research Objectives (Tentative) Research Objectives (Tentative)

• 1) 1) To mathematically model the To mathematically model the pertinent pertinent (hydrologically-relevant) spatio-temporal features (hydrologically-relevant) spatio-temporal features of satellite rainfall uncertainty as a function of of satellite rainfall uncertainty as a function of season and location over the US.season and location over the US.

• 2) To develop and test a geostatistical mapping 2) To develop and test a geostatistical mapping scheme for spatial interpolation of error information scheme for spatial interpolation of error information from small-domain GV regions to the neighboring from small-domain GV regions to the neighboring vast ungauged (non-GV) regions.vast ungauged (non-GV) regions.

MY FUNDAMENTAL RESEARCH MY FUNDAMENTAL RESEARCH HYPOTHESISHYPOTHESIS

• Hydrologists and other users, to varying degrees, Hydrologists and other users, to varying degrees, need to know the errors of the satellite rainfall need to know the errors of the satellite rainfall datasets across the range of time/space scales over datasets across the range of time/space scales over the whole domain of the data set. On the other hand, the whole domain of the data set. On the other hand, satellite rainfall datasets are most useful over the satellite rainfall datasets are most useful over the vast ungauged regions of the developing world vast ungauged regions of the developing world lacking in GV data. If the potential of GPM is to be lacking in GV data. If the potential of GPM is to be unleashed over the vast regions of the developing unleashed over the vast regions of the developing world, then a reconciliation of this paradox is needed world, then a reconciliation of this paradox is needed before the launch of GPM in 2013.before the launch of GPM in 2013.

ProposedProposed Research MethodologyResearch Methodology

• Study Region and DataStudy Region and Data

-Unites States as the Unites States as the primary study region primary study region further divided into further divided into six zones (Each zone six zones (Each zone is governed by is governed by distinct rainfall distinct rainfall climatology)climatology)

• Climate of United statesClimate of United states

------( Average annual precipitation from 1961-1990 )

• Study dataStudy data

- The primary source for The primary source for GV rainfall data will GV rainfall data will comprise the National comprise the National Weather Service Weather Service (NWS) WSR-88D radar (NWS) WSR-88D radar rainfall data. rainfall data.

- Satellite rainfall data Satellite rainfall data will comprise all major will comprise all major products of NASA’s products of NASA’s TMPA system (Huffman TMPA system (Huffman et al., 2007) – i.e., et al., 2007) – i.e., 3B40RT, 3B41RT, 3B40RT, 3B41RT, 3B42RT and 3B42V6.3B42RT and 3B42V6.

products

Spatial scale

(degree)

Temporal Resolution (hours)

Period(Years)

3B40RT 0.25 3 5

3B41RT 0.25 1 5

3B42RT 0.25 3 5

3B42V6 0.25 3 9

NWS(radar)

0.04 1 9

Methodology Flowchart

Satellite data Ground validation data

Task 1 Task 2

Error metrics and error classification

Spatial interpolation of error metrics from sparse GV sites

Error framework

Proposed Specific TasksProposed Specific Tasks (Tentative) (Tentative)

1) 1) TASK 1 (Years 1 and 2): Mathematical modeling TASK 1 (Years 1 and 2): Mathematical modeling of hydrologically relevant error metrics as a of hydrologically relevant error metrics as a function of regime, season and location.function of regime, season and location.

- - This task will identify the total number of error This task will identify the total number of error metrics that are meaningful for improving the metrics that are meaningful for improving the hydrologic potential of GPM and also interpretable hydrologic potential of GPM and also interpretable by both user and data producing communities.by both user and data producing communities.

- It will begin with the initial set of nine error - It will begin with the initial set of nine error metrics that my advisor has already devised for a metrics that my advisor has already devised for a Two-Dimensional Satellite Rainfall Error Model - Two-Dimensional Satellite Rainfall Error Model - SREM2D (see Hossain and Anagnostou, 2006; and SREM2D (see Hossain and Anagnostou, 2006; and Hossain and Huffman, 2007-Hossain and Huffman, 2007-in pressin press).).

Nine error metrics in SREM2D

NWS Radar dataCalibration

Simulated satellite rainfall

Simulation

Aggregated simulated satellite rainfall data

space/time downscaling

Aggregated actual satellite rainfall data

GV

YES?

comparisonNO?

Modify error framework

using various error

combinations

2) 2) TASK 2 (Years 2 and 3): Development and TASK 2 (Years 2 and 3): Development and assessment of a geostatistical-based mapping assessment of a geostatistical-based mapping scheme for spatial interpolation of error metrics scheme for spatial interpolation of error metrics from sparse GV sites. from sparse GV sites.

- - This task will seek answer to if This task will seek answer to if ““error is error is defined on the basis of GV, then how are defined on the basis of GV, then how are these metrics estimated for a satellite these metrics estimated for a satellite data product without the need for data product without the need for extensive GV data?extensive GV data?” ”

Results using NWS Radar reference data(TASK 1)

TRMM PR data (2A25)

Error metrics derived at finite PR overpass locations

Results

comparison

YES?

Surrogate reference

Extensive non-GV and non-PR (overpass) location

- - Currently iCurrently involved in the creation of a 9-year nvolved in the creation of a 9-year (1997-2006) mosaic of National Weather Service (1997-2006) mosaic of National Weather Service (NWS) ground radar rainfall data (WSR-88D) over (NWS) ground radar rainfall data (WSR-88D) over the coterminous United States. (the coterminous United States. (Hydro-NEXRADHydro-NEXRAD))

- I am performing statistical analysis of the rainfall - I am performing statistical analysis of the rainfall distribution to understand how these parameters distribution to understand how these parameters vary as a function of location. vary as a function of location. This part of my work This part of my work is expected to help refine the exact periphery of is expected to help refine the exact periphery of the five zones in manner analogous to Koppen the five zones in manner analogous to Koppen climate classification.climate classification.

• Initial work completedInitial work completed

Proposed Research PublicationsProposed Research Publications

• From From TASK 1TASK 1 –Two journal quality research papers. –Two journal quality research papers.

First one will be on the identification of distinct satellite First one will be on the identification of distinct satellite rainfall error classification regimes in a manner similar rainfall error classification regimes in a manner similar to Koppen Classification in J. Climate. Second will to Koppen Classification in J. Climate. Second will identify the ideal set of error metrics through identify the ideal set of error metrics through consistency analysis in J. Hydrometeorology or Water consistency analysis in J. Hydrometeorology or Water Resources Research.Resources Research.

• From From TASK 2TASK 2 – One research paper in J. – One research paper in J. Hydrometeorology on the effectivesness of spatial Hydrometeorology on the effectivesness of spatial mapping of error metrics over ungauged regimesmapping of error metrics over ungauged regimes

• Timetable for completion of TASKSTimetable for completion of TASKS

Completion Date

First year

(2008-09)

Second year

(2009-10)

Third year

(2010-11)

TASK#1TASK#1 XXXXXX XXXXXX

TASK#2TASK#2 XXXXXX XXXXXX

PublicationsPublications Paper 1Paper 1 Paper 2Paper 2 Paper 3Paper 3

Anticipated Broader Impact of the Anticipated Broader Impact of the Proposed ProjectProposed Project

- T- The long-term impact from my proposed he long-term impact from my proposed project is, therefore, expected to be in project is, therefore, expected to be in laying the foundation for understanding laying the foundation for understanding the level to which satellite rainfall the level to which satellite rainfall products can realistically advance societal products can realistically advance societal applications (such as flood detection) in applications (such as flood detection) in the developing world. the developing world.

- - My undergraduate background is in My undergraduate background is in hyperspectral remote sensing of hyperspectral remote sensing of geophysical parameters such as terrain geophysical parameters such as terrain features and topography.features and topography.

- - My masters research focused on My masters research focused on ‘artificial ‘artificial immune system’ technique for optimal immune system’ technique for optimal extraction of features and information extraction of features and information from remote sensing datasets.from remote sensing datasets.

My Research BackgroundMy Research Background

• Background courseworkBackground coursework

MastersMasters

(Wuhan (Wuhan University)University)

-Matrix Theory -Matrix Theory

-Numerical Analysis -Numerical Analysis

-Microwave Remote Sensing-Microwave Remote Sensing

-Error Processing and Reliability Theory-Error Processing and Reliability Theory

-Theory and Technology of GIS -Theory and Technology of GIS

-Image Processing and Analysis -Image Processing and Analysis

-Fundamentals of Intelligent System in Remote -Fundamentals of Intelligent System in Remote SensingSensing

PhDPhD

(Tennessee Tech. (Tennessee Tech. University)University)

Fall 2007Fall 2007

-CEE 6520 *Open-channel Hydraulics-CEE 6520 *Open-channel Hydraulics

-CEE 6610-CEE 6610 * *Applied Environmental ChemistryApplied Environmental Chemistry

-CEE 6440 Hydrometeorology-CEE 6440 Hydrometeorology

Spring 2008Spring 2008

-ECE 6900 Special Problems (Intelligent Systems)-ECE 6900 Special Problems (Intelligent Systems)

-CEE 7980 Directed Study (Satellite Rainfall -CEE 7980 Directed Study (Satellite Rainfall Uncertainty Analysis )Uncertainty Analysis )

Courses In future Courses In future

-CEE 6430-CEE 6430 *Probabilistic methods (Dr. Hossain)*Probabilistic methods (Dr. Hossain)

-Three more 7000 level courses-Three more 7000 level courses

(*----core course)(*----core course)

• Prior Research ExperiencePrior Research Experience

- Ant Colony Algorithm in Image Interpretation- Ant Colony Algorithm in Image Interpretation (China National Natural Science Fund Project(China National Natural Science Fund Project) ) By simulating ant colony behavior, this project aimed to By simulating ant colony behavior, this project aimed to establish a set of ant colony behavior optimal algorithms establish a set of ant colony behavior optimal algorithms and models, and using them to solve image interpretation and models, and using them to solve image interpretation problems, mainly on aerial and satellite images. problems, mainly on aerial and satellite images.

- Artificial Immune System in Image Segmentation - Artificial Immune System in Image Segmentation This research introduced the mechanism of biological This research introduced the mechanism of biological immune system into remote sensing image processing, and immune system into remote sensing image processing, and thus developed a refined optimal algorithm based on thus developed a refined optimal algorithm based on artificial immune system for satellite image processing such artificial immune system for satellite image processing such as segmentation, classification, and feature extraction. as segmentation, classification, and feature extraction.

• Prior PublicationsPrior Publications

1. Ling Tang, Zhaobao Zheng. A new approach based on 1. Ling Tang, Zhaobao Zheng. A new approach based on artificial immune system for texture segmentation on artificial immune system for texture segmentation on satellite imagery. The 15satellite imagery. The 15thth National Conference on Remote National Conference on Remote Sensing Technique. Guiyang, China. 2005Sensing Technique. Guiyang, China. 2005

2. Ling Tang, Zhaobao Zheng . An image segmentation 2. Ling Tang, Zhaobao Zheng . An image segmentation algorithm based on artificial immune system. Geomatics algorithm based on artificial immune system. Geomatics and Information Science of Wuhan University. Vol.32, and Information Science of Wuhan University. Vol.32, No.1, Jun. 2007, 68~70.No.1, Jun. 2007, 68~70.

3. Xin Yu, Zhaobao Zheng, Ling Tang. Aerial image texture 3. Xin Yu, Zhaobao Zheng, Ling Tang. Aerial image texture classification based on a new Bayes classifier. Geomatics classification based on a new Bayes classifier. Geomatics and Information Science of Wuhan University. Vol.31, and Information Science of Wuhan University. Vol.31, No.2, Feb.2006 108No.2, Feb.2006 108 ～～ 111.111.

• My Current Work on Hydrologic My Current Work on Hydrologic Remote Sensing at TTURemote Sensing at TTU

- - I have familiarized myself with some hydrologic I have familiarized myself with some hydrologic rainfall remote sensing products.rainfall remote sensing products.

- I have also familiarized myself the well-known - I have also familiarized myself the well-known Global Precipitation Climatology Project (GPCP)Global Precipitation Climatology Project (GPCP) and satellite rainfall remote sensing.and satellite rainfall remote sensing.

Six years (1997-2002) of GPCP daily rainfall Six years (1997-2002) of GPCP daily rainfall data (resolution of daily at 1data (resolution of daily at 100×1×100) was used to ) was used to analyze the spatial and temporal patterns of analyze the spatial and temporal patterns of the global precipitation over large river basins. the global precipitation over large river basins.

• My Recent Work on Hydrologic Remote SensingMy Recent Work on Hydrologic Remote Sensing

• Knowledge of Knowledge of Analysis and design toolsAnalysis and design tools

- C++ (programming)- C++ (programming)

- Matlab (programming)- Matlab (programming)

- ArcGIS- ArcGIS (mapping)(mapping)

- Cygwin (Unix Systems Simulator)- Cygwin (Unix Systems Simulator)

- ERDAS (Remote Sensing analysis)- ERDAS (Remote Sensing analysis)

- IDL (NASA language)- IDL (NASA language)

Thank Thank you !you !