Tuesday, September 10, 2013

ANDS User Acceptance Testing of ePiSaT Version 1.0

The user acceptance testing was conducted once the ePiSaT Software was released to the public via github. Each tester was selected for their specialist expertise, thus representing a range of potential users. Firstly, Associate Professor Lindsay Hutley from Charles Darwin University was selected because he provided flux tower data to the project and he understands the ‘flux data’. Secondly, Dr Siddeswara Guru from Queensland University was selected because he understands data processing and software. The reviewers were provided the github repository URL, and the documentation and asked to test if the system was capable of perform as described.

The key outcome is that both reviewers concluded that the package performs as expected and commended it as a positive contribution to the community.  Following are the 2 reviewers complete responses;

Reviewer # 1

Dear Brad,

Thank you for allowing me to review your ePiSaT system.

I have followed simple instructions as provided and looked at your blog that hosts the system.

As novice user to this system, I was able, with no major problems to download and use the test flux data and run the these data through the R scripts.

The system / model appears to be robust and uses the latest gridded climate data and fPAR/GPP products for Australia. Functions as detailed are all standard versions as described in the literature.

I was able to generate 30 minute partitioned GPP and Re that appeared to me to be reasonable and in-line with previous efforts to partition flux data in savanna at a plot scale – your product we have the capability at biome and continental scales.

I could see some savanna specific improvements that could be made (e.g. see the LUE and GPP models of Kannihar et al) but this system is focused on partitioning fluxes but in my view it is relatively user-friendly, well documented and has provided credible outputs. Key variables outputted from the model (Amax, quantum efficiency, respiration etc) are in line with previous efforts to partition this data.

Congratulations on a very useful tool, there are numerous methods to partition flux data and this is a product pitched at integrating current OzFlux data and spatially extrapolate this over space and time, a hugely useful outcome. For a ‘flux ecologist’ like myself GPP and respiration data can now be compared with site values, compared to other models, integrated with maps of fire, soil types, climate envelopes to further unpack spatial and temporal patterns of carbon cycling in savanna.

Regards
Assoc Prof Lindsay Hutley
Principal Research Fellow
Research Institute for the Environment and Livelihoods
Charles Darwin University

http://riel.cdu.edu.au/people/profile/lindsay-hutley



Reviewer # 2


A selected quote from S.M.Guru;

"I was able to conduct the user acceptance testing on the ePiSaT product ... The plotting functionality is also impressive. Good luck with the software release and there is a huge potential for the product."


S.M.Guru




Dr. Siddeswara Guru

Data Synthesis and Integration Coordinator

Terrestrial Ecosystem Research Network (TERN) and Australian Centre for Ecological Analysis and Synthesis (ACEAS)



This project is supported by the Australian National Data Service (ANDS). ANDS is supported by the Australian Government through the National Collaborative Research Infrastructure Strategy Program and the Education Investment Fund (EIF) Super Science Initiative.

For more information visit the ANDS website ands.org.au and Research Data Australia services.ands.org.au.

Sunday, September 1, 2013

ePiSaT software version 1.0 released !

The ePiSaT software has been released. The code and sample OzFlux data can be downloaded from:

https://github.com/belgaroo/ePiSaT-ecosystem-Production-in-Space-and-Time

The package and code is self explanatory, enabling the user to partition flux data to generate an estimate of GPP. ePiSaT users wanting to replicate the project can download more OzFlux data directly:

http://www.ozflux.org.au/

Please note that there have been some subtle changes to previously-posted details on how and where the ePiSaT software will be released. This post details the first official release of the ePiSaT software.

Background


Ecosystem Production in Space and Time (ePiSaT) models Gross Primary Production (GPP) across the Australian continent from OzFlux flux tower data, gridded climate and satellite (MODIS) data. ePiSaT can be used to estimate ecosystem variables by partitioning OzFlux eddy covariance data to estimate GPP, light use efficiency (LUE), the maximum rate of carboxylation (Amax) etc. Once determined, the ePiSaT-modelled variables can be used in the well-founded and tested LUE approach to estimate GPP. Sample OzFlux data from Howard Springs is provided; ePiSaT requires the user to provide fAPAR, observed radiation, temperature and humidity data.


Australian National Data Service (ANDS)


This project is supported by the Australian National Data Service (ANDS). ANDS is supported by the Australian Government through the National Collaborative Research Infrastructure Strategy Program and the Education Investment Fund (EIF) Super Science Initiative. For more information visit the ANDS website ands.org.au and Research Data Australia services.ands.org.au


Installation Instructions

Once downloaded, you can install episat using the following terminal command.R CMD INSTALL episat_1.0.tar.gzAlternatively, from an R GUI go the package manager and follow the prompts to install a package from source.


Running the software in R

Once installed, from an R command prompt, run the following code:
library(episat) 

?episat # <-- for help and how to use the package.


About the license

This work is licensed under a Creative Commons Attribution 3.0 Attribution-NonCommercial- CC BY-NC 

http://creativecommons.org/licenses/by-nc/3.0/legalcode 

This license lets others remix, tweak, and build upon your work non-commercially, and although their new works must also acknowledge you and be non-commercial, they don't have to license their derivative works on the same terms.

REQUIRED ATTRIBUTION : DOI XXX.XXX.XXXX.XXX


AUTHORS: Bradley Evans, Colin Prentice, Tyler W. Davis, Xavier Gilbert 

ORGANISATION: Macquarie University, Sydney Australia

REFERENCE: Evans, B.J., Prentice, I.C., Davis, T.W., Gilbert, X., 2013, Ecosystem Production in Space and Time, http://episat-software.blogspot.com.au

EXAMPLE: You wish to use this code or data in your own work, a peer reviewed journal article, then you need to attribute the work by referencing published article listed above and/ or the DOI (i.e. for output data only).

Contact the author and maintainer if you have any questions.

MAINTAINER: bradley.evans << at >> mq.edu.au








Sunday, March 3, 2013

Multiple observation types reduce uncertainty in Australia’s terrestrial carbon and water cycles by Vanessa Haverd , M. R. Raupach, P R Briggs, J. Canadell , P. R.Isaac, C. Pickett Heaps, S. H Roxburgh, E. van Gorsel, R. Viscarra Rossel and Z. Wang

Information about the carbon cycle potentially constrains the water cycle, and vice versa. Benefitting from the public liberation of the data necessary to drive carbon and water cycle models, a recent study by the authors of this blog post succesfully demonstrated the utility of multiple observation sets - such as provided by the TERN Auscover and OzFlux facilities - to constrain carbon and water fluxes and stores in a land surface model, and a resulting determination of the Australian terrestrial carbon budget. It exemplifies the joint use of data from observational facilities and infrastructure (e-MAST).

In observations include streamflow from 416 gauged catchments, measurements of evapotranspiration (ET) and net ecosystem production (NEP) from 12 eddy-flux sites, litterfall data, and data on carbon pools. Coupled carbon and water cycles were simulated using a modified version of the CABLE land surface scheme in the BIOS2 modelling environment, a fine spatial resolution (0.05 degrees) offline environment built on capability developed for the Australian Water Availability Project (King et al., 2009; Raupach. et al., 2009). BIOS2 includes: (1) a modification of the CABLE land surface scheme (Wang et al., 2011) coupled with CASAcnp (a biogeochemical model) and SLI (Soil-Litter-Iso, a soil hydrology model including liquid and vapour water fluxes and the effects of litter).; (2) infrastructure for the treatment of inputs (gridded vegetation cover, meteorological data and parameters) and outputs for optimum efficiency; (3) a weather generator for downscaling of meteorological data; and (4) model-data fusion capability.

Results emerging from the multiply-constrained model are as: (1) on the Australian continent, a predominantly semi-arid region, over half (0.64±0.05) of the water loss through ET occurs through soil evaporation and bypasses plants entirely; (2) mean Australian NPP is 2200±400 TgC/y, making the NPP/precipitation ratio about the same for Australia as the global land average; (3) annually cyclic (“grassy”) vegetation and persistent (“woody”) vegetation respectively account for 0.56±0.14 and 0.43±0.14 of NPP across Australia; (4) the average interannual variability of Australia’s NEP (±180 TgC/y) is larger than Australia’s total anthropogenic greenhouse gas emissions in 2011 (149 TgCeq/y).

Gap filling techniques to use for OzFlux eddy covariance data by Natalia Restrepo-Coupe, Alfredo Huete and Kevin Davis

Our goal is to use the OzFlux network micrometeorological flux measurements of carbon dioxide exchange between different Australian ecosystems (biosphere) and the atmosphere, to better understand the physical meaning of satellite data generally associated to photosynthetic capacity. The Eddy Covariance Method (EC) used at all OzFlux sites, measures Net Ecosystem Exchange (NEE) on an hourly basis. NEE can be separated on its 2 main components, Ecosystem Respiration (Re) and assimilation (photosynthesis), the latter assumed to be equivalent to Gross Ecosystem Productivity (GEP). Therefore, GEP = NEE + Re. For detailed information about the EC method see Goulden et al. (1996).

Figure 1. Rectangular hyperbola fitted to 8-day worth of Gross Ecosystem Productivity (GEP) and Short Wave Incoming Radiation (SWin) data measured at Calperum-Chowilla flux tower data (OZflux). Photosynthetic Capacity (Pc), Light Use Efficiency (LUE) and GEP at saturation (GEPsat) are calculated, as shown. Special thanks to Prof D. Chittleborough, Prof W. Meyer, Dr. G. Whiteman and T. Luckbe

In order to harmonize and standardize the flux data we carry out the following process: (1) clean the data (e.g. remove data collected during rainfall) and remove existent outliers, (2) define a site specific turbulence threshold, as fluxes need to be corrected (removed) if measured during low turbulence periods, when the basic assumptions of the EC method are not satisfied, (3) NEE is partitioned in Re and GEP, were Re is derived from nighttime NEE (when there is no photosynthesis, GEP=0), and finally, (4) fit a rectangular hyperbola to the relationship between GEP and incoming solar radiation and using the fit to obtain different measures of ecosystem photosynthetic capacity (e.g. productivity at saturating light, GEP
sat) see Figure 1.

We compare different MODIS vegetation indices and reflectances to the flux derived data (Figure 2), to determine the product that best (similar seasonality and year-to-year variations) matches the insitu measures of ecosystem photosynthetic capacity, required by the proposed Australian terrestrial GPP near real-time monitoring system. For more information regarding MODIS vegetation indices, see Measuring Vegetation NDVI and EVI, http://earthobservatory.nasa.gov by Weier and Herring (2000)
 


Figure 2.Annual cycle (16-day mean) for Calperum-Chowilla flux station (a) tower measured Gross Ecosystem Productivity, GEP (black line) and Ecosystem Light Use Efficiency, LUE (red line). Special thanks to Prof D. Chittleborough, Prof W. Meyer, Dr. G. Whiteman and T. Luckbe. (b) MODIS Enhanced Vegetation Index, EVI (black line) and Normalized Vegetation Index, NDVI (red line)
http://earthobservatory.nasa.gov by Weier and Herring (2000)

Sunday, November 4, 2012

ePiSaT at The Google Earth Outreach launch, Nov 5 at Macquarie University

Members of the ePiSaT will be at the Google Earth Outreach launch to demonstrate, in a very Google Earth way, how ePiSaT is modelling Australia’s carbon cycle. Here is a video we produced to illustrate how our project is modelling Australia's terrestrial Gross Primary Production.







Friday, June 1, 2012

This projects deliverables

This project will deliver a hindcast and near real-time monitoring system for terrestrial Gross Primary Production by natural, semi-natural and agricultural ecosystems across the Australian continent. The new national dataset will be generated on a 0.05 ° grid and be available for download from a URL to be announced early 2013. Additionally, a set of tools, together with instructions on their use, will be provided for use under Creative Commons Attribution 3.0 Unported License from the projects google code repository. This is an example of what the modelled output might look like.






This work is licensed under a Creative Commons Attribution 3.0 Unported License.


This project is supported by the Australian National Data Service (ANDS). ANDS is supported by the Australian Government through the National Collaborative Research Infrastructure Strategy Program and the Education Investment Fund (EIF) Super Science Initiative.

For more information visit the ANDS website ands.org.au and Research Data Australia services.ands.org.au.

Tuesday, May 29, 2012

The features of our model and modelling system


The general equation of the model

Our Gross Primary Production (GPP) model design builds on the well founded and tested light use efficiency approach and incorporates recent developments in our understanding of the relationships between Carbon dioxide (CO2), Vapour Pressure Deficit, Soil Moisture and Photosynthesis. More specifically we recognise that;

→ VPD appears more predictable than previously thought.
→ SM appears to modify the VPD stomata relationship differently between plant functional types.
→ Soil dryness also reduces Vcmax and Jmax, compared to what would otherwise be predicted from stomatal conductance alone.

The general equation behind our modelling approach can be written as follows;

GPP = fAPAR . IPAR . E . f(T, CO2, VPD, SM)

Where fAPAR is the fraction of absorbed photosynthetically active radiation, IPAR is the incoming photosynthetically active radiation (400-700nm), E (epsilon) is an estimate of light use efficiency which is a function of Temperature (T), Carbon Dioxide (CO2) Vapour Pressure Deficit (VPD) and Soil Moisture (SM).


Key features of the modelling system

1.     Capable of reading in a range of datasets from the TERN and Bureau of Meteorology and re-grid them as required; a.k.a. data abstraction.
2.     Be usable stand-alone, embedded in a workflow or ported to a new model core; a.k.a. modular in design.
3.     Facilitate point-to-point calibration and validation against OzFlux tower data across a range of spatial and temporal scales.
4.     Output any metadata elements in a human readable format.
5.     Written in an open source language with cross platform interoperability.

Architecture





Technology

The computational requirements of this project are overshadowed by the need for usability and transparency- given our open access and source approach. As such, we identified the two most widely used open source platforms in our core user group: R-Project and Python. R was chosen by the development team given it is considered more widely used for benchmarking within our local user community, for example: PALS.

Data abstraction will be done using the Geospatial Data Abstraction Library (GDAL). In the R version, RGDAL bindings and the raster package will be used for low-level reading and writing of data and to facilitate the high-level model functions required for the model and modelling system functionality.



This work is licensed under a Creative Commons Attribution 3.0 Unported License.

This project is supported by the Australian National Data Service (ANDS). ANDS is supported by the Australian Government through the National Collaborative Research Infrastructure Strategy Program and the Education Investment Fund (EIF) Super Science Initiative.

For more information visit the ANDS website ands.org.au and Research Data Australia services.ands.org.au.