Crowd-sourced knowledge generation for the validation of global - - PowerPoint PPT Presentation

Jonas Eberle 25th March 2015 1

Crowd-sourced knowledge generation for the validation of global vegetation change analyses —

A feedback tool to foster tests and evaluations of scientific algorithms

Jonas Eberle, Dr. Christian Hüttich, Prof. Christiane Schmullius Friedrich-Schiller-University Jena, Germany Institute for Geography, Department for Earth Observation www.eo.uni-jena.de

Breakout Session 1.3: Social Revolution: Crowdsourcing movement and Earth Observation

Jonas Eberle 25th March 2015 2

Vegetation time-series change analyses

• Data

– MODIS Vegetation Index (NDVI, EVI)

• 250m spatial resolution
• 16-daily product
• 14 years of data!
• Analyses

– Trend calculation – Break-point detection – Phenological parameters

Jonas Eberle 25th March 2015 3

Vegetation trend calculation

Greenbrown (R-package)

Forkel et al. (2013). “Trend Change Detection in NDVI Time Series: Effects of Inter-Annual Variability and Methodology.” Remote Sensing 5 (5): 2113–44.

• Example:

– Use case: Anklam, Germany – Former moor area is being recultivated – Since 2008 controlled water logging

• Web Service

– OGC Web Processing Service – Inputs: data id + optional analysis parameters

Jonas Eberle 25th March 2015 4

Vegetation break-point detection

bfast (R-package)

Verbesselt et al. (2010). “Detecting Trend and Seasonal Changes in Satellite Image Time Series.” Remote Sensing

• f Environment 114 (1). Elsevier B.V.: 106–15.
• Web Service

– Accessible via OGC Web Processing Service – Inputs: data id + optional analysis parameters

• Example:

– Use case: Bavarian Forest – Bark beetle attack starting in the 90s à forest dieback – Now: greening on the ground

Center Lat.: 53.8328 Center Long.: 13.8309

Jonas Eberle 25th March 2015 5

Derivation of phenological parameters

TIMESAT

Jönsson & Eklundh (2004). “TIMESAT—a Program for Analyzing Time-Series of Satellite Sensor Data.” Computers & Geosciences 30 (8): 833–45.

• Example

– Use case: Anklam, Germany – Large Integral (mean)

• Web Service

– Accessible via OGC Web Processing Service – Inputs: data id + optional analysis parameters

Jonas Eberle 25th March 2015 6

Crowd-sourced validation of analysis results

• How can we evaluate algorithms in different regions for

different aspects of change?

• How can users easily explore algorithms without having any

knowledge in data processing? Ø Automated data access linked with automated execution of analysis tools based on that data Ø Solutions:

– Providing web services for data access and analysis – Developing easy to use clients

Automated ¡access, ¡analysis, ¡and ¡monitoring ¡

• f ¡global ¡vegeta6on ¡6me-­‑series ¡data ¡

Earth ¡Observa,on ¡Monitor ¡

Datasets: ¡ ¡

• MODIS ¡16-­‑Daily ¡Vegeta2on ¡

Index ¡(NDVI, ¡EVI) ¡

¡

Data ¡access: ¡

• Pixel ¡or ¡Polygon-­‑based ¡

extrac2on ¡service ¡

¡

Analyses: ¡ ¡

• Trend ¡calcula2ons ¡
• Breakpoint ¡detec2on ¡
• Phenological ¡parameters ¡

¡

Applica6ons: ¡

• Web ¡Portal ¡
• Mobile ¡App ¡(mobileEOM) ¡

for ¡iOS ¡and ¡Android ¡

www.earth-­‑observa6on-­‑monitor.net ¡

Jonas Eberle 25th March 2015 8

Crowd-sourced validation of analysis results

Execute scientific algorithms with individual parameters around the world without the need to process any data

Jonas Eberle 25th March 2015 9

Crowd-sourced validation of analysis results

• Objectives

– Help to improve & validate scientific algorithms – Give feedback / response to author of an algorithm – Easy to use!

• Easy to use

– User just needs to write a short description – Parameters used are automatically inserted into the text field

Ø Online analysis feedback tool

Jonas Eberle 25th March 2015 10

Analysis feedback tool

• What do we want to achieve?

– Interaction and exchange between users and authors of an algorithm – Better understanding of how useful the algorithm is for different regions / study areas – What are the needs of users using the algorithm? – Bringing scientific algorithms to operational services

• What is needed?

ü Data access ü Algorithm as web service

• Feedback possibility (still to do)

Jonas Eberle 25th March 2015 11

Analysis feedback tool

• Implementation

– Within the Earth Observation Monitor – Author of an algorithm needs to be involved – First algorithms to test feedback tool

• Greenbrown trend calculations
• bfast break-point detection

– Results (will it be used?)

• Further ideas

– Discussing platform to communicate with other users – Add other datasets for validation of analysis results

Jonas Eberle 25th March 2015 12

Satellite based Wetland Observation System (SWOS)

• New 3-year EU H2020 project starting in June 2015
• Call: Making Earth Observation and Monitoring Data usable for

ecosystem modelling and services

• Objective: Development of an operational and standardized

monitoring system for wetlands based on multi-sensor earth

• bservation data.
• Contributions to GEOSS are included in the project!
• Crowd-sourcing with mobile devices

– Validate wetland delineation on mobile device – Create new in situ data relevant for wetland processing & analysis

Jonas Eberle 25th March 2015 13

Conclusions

• Automated data access and web-based analysis tools have to

be linked closely.

• Analysis feedback tool still has to be implemented (next step)
• More interaction between users and author of an algorithm
• Algorithms can be tested everywhere around the world

Ø Crowd-sourcing leads to new possibilities in

– Testing algorithms – Evaluating algorithms – Exchanging experiences

Ø We just need the right tools for users!

Jonas Eberle 25th March 2015 14

Thank you for your attention! Questions?

Contact information: Jonas Eberle Friedrich-Schiller-University Institute for Geography Department Earth Observation Loebdergraben 32 07743 Jena, Germany phone: +49 3641 94 88 89 email: jonas.eberle@uni-jena.de Acknowledgement: Friedrich-Schiller-University Jena and EU FP7 EuRuCAS project (No. 295068) for financing work and travel.