Global Precipitation Measurement GPM

1
Global Precipitation Measurement (GPM)

• GV Data Exchange Protocol

Mathew Schwaller GPM Formulation Project Ground
Validation Manager mathew.r.schwaller_at_nasa.gov
2
Objectives for this Presentation

• Examine the assumptions and issues surrounding a
  data exchange protocol for international global
  precipitation measurement (GPM) ground validation
• Identify many problems associated with defining
  and maintaining a standard data exchange protocol
• Propose an applications-based approach to
  interoperability among international GPM ground
  validation sites

3
Some Assumptions

• Assumptions
• Each GV site will have its own unique set of
  instrumentation optimized for its local users
• Not possible or desirable to standardize site
  instrumentation
• Each GV site will have its own data system for
  archive and distribution
• These data systems will be optimized for local
  users
• All GV sites make voluntary contributions of
  local data and information, with the goal of
  validating Global Precipitation Measurement
  instrumentation, algorithms and products
• There is no central funding authority (definition
  of voluntary)
• There is no central authority to manage voluntary
  contributions
• Issues
• There will be a relatively small number of
  applications that will utilize the data from an
  international consortium of GV sites
• At present, there are no applications that fall
  into this category
• Is it possible for a band of volunteers to adopt
  a common data exchange format leading to global
  validation products?

4
Graphic Illustration of the Assumptions
middleware
data exchange protocol
data system
GV site-1
GV site-n
GV site-2
instrumentation

• Each site will have its own unique
  instrumentation
• Each site will have its own unique data system
• Only a few shared/global applications are
  anticipated
• Is it possible to adopt common data model, data
  exchange attributes, formats, protocols, and
  middleware for generating global validation
  products?

5
The Scaling Problem

• Coordination costs rise exponentially as partners
  are added
• Any n partners can form up to (2n-1) possible
  subsets
• Specifying GV site attributes, and coordinating
  shared attributes among varying sites quickly
  becomes a logistical nightmare!
• The more successful you are (the larger the
  number of GV participants) the time and spent
  on coordination rises exponentially
• If unchecked, data coordination costs could eat
  into the GV budget for instrumentation,
  measurement and analysis

6
The Problem of Site Bias

• Each site will have its own set of instruments,
  measurement protocols, calibration procedures,
  analysis methods and data products
• There is no way to coordinate or dictate
  instrumentation and operations commonality among
  all GV sites
• It should be possible, but may be very difficult,
  to agree on a common data product there may be
  many interpretations and representations of
  instantaneous rainfall for example
• Even if common data products are found (e.g.,
  reflectivity), it will be difficult to agree on a
  common method for measuring and reporting the
  errors associated with the product
• We need to assume that the measurement error of
  any data product will vary from one GV site to
  another, and that there will be (unknown) bias
  from site to site

7
The Problem of Incentive

• There are a number of costs that each GV site
  must consider before entering into an
  international GV consortium
• Data conversion costs product content and format
  will likely vary from site to site
• Data reliability costs product measurement error
  and bias vary at each site, and they may not be
  well characterized
• Utilization costs no global applications at
  present that can use the data even if it were
  available
• There is certainly value in generating a globally
  consistent data set for precipitation validation,
  but
• The value of participating in a GV consortium
  must be greater than the cost

8
An Ideal Exchange Protocol

• Maximizes incentives for voluntary participation
• Provides some value to each participant
• Minimizes cost for participation
• maximizes re-use of existing resources at local
  sites
• Minimizes coordination costs
• Compensates for or otherwise quantifies
  within-site error and site-to-site bias

9
Typical 3-Tier Architecture

• In many data system architectures, applications
  are separated from data sources by middleware
• Middleware strengths
• Common interface for applications programmers to
  access data objects and services
• Resolution of location/access information about
  data objects and services
• May provide workflow services for complex tasks
• Weaknesses
• Middleware needs to be designed, developed,
  tested, maintained
• Longer term, the project could still falter if
  the International Virtual Observatory middleware
  standards make it too expensive for institutions
  to prepare their survey data

applications
middleware
data sources
10
Example from space sciences the International
Virtual Observatory

• Applications

Data Sources
11
Recommendation

• At this stage focus on the applications, let the
  protocol follow the applications
• Define applications/algorithms for validating
  global precipitation that are interesting and
  useful
• Implement the application(s) at cooperating
  partner sites in the international GV consortium
• Initially, this will require developing a custom
  interface for each site
• As the number of global GV applications grows
  (beyond 2 or 3), work on a common data exchange
  protocol should be reconsidered

Replicated application
application-A
application-A
application-A
Custom interface for each data source
data source-1
data source-n
data source-2
12
Example GV Application
Algorithm resamples coincident ground and
space-based radar observations (Bolen and
Chandresekar, 2000 Liao et al., 2001)
TRMM PR
GPM DPR
ground radar

• Matching ground and space-based (PR, DPR) data
  for statistical validation

13
Prototype in Development

• Resampling prototype provides statistical
  comparison of ground and space radar reflectivity
• Good agreement high in the storm (where PR/DPR
  attenuation effects are minimal) indicates good
  relative agreement between PR/DPR and the GV
  radars
• Good agreement near the surface indicates that
  the PR/DPR attenuation-correction algorithms are
  working well
• Can be extended to comparison of precipitation
  rate, DSD and other variables
• Prototype will focus on TRMM PR comparison with
  NOAA NSR-88D radars
• Starting with one or two NSR-88D sites
• Will evaluate the possibility of scaling the
  prototype to all 158 WRS-88D radars
• Will evaluate the possibility of scaling to other
  S-band radars from other US and international
  sites

14
Conclusions

• Focus on global applications that can exploit GV
  data from an international consortium of
  providers
• These applications will help define the
  requirements for a GV data protocol
• Once the applications requirements are understood
  a number of possible frameworks for data exchange
  may be considered, for example
• CEOS/GEOS/CEOP focus on catalog services
• Open Geographic Information System (OGIS) focus
  on web services
• National/International Virtual Observatory (and
  others) focus on Grid storage and computing
• Whatever the framework, the GV data exchange
  protocol must address practical issues of
  scalability, site bias, and incentives for
  participation

15
(No Transcript)