wiki-archive/twiki/data/InvasiveSpecies/AdvisoryGroup.txt,v

head	1.5;
access;
symbols;
locks; strict;
comment	@# @;


1.5
date	2009.12.13.14.46.29;	author AnnieSimpson;	state Exp;
branches;
next	1.4;

1.4
date	2009.12.12.06.47.36;	author AnnieSimpson;	state Exp;
branches;
next	1.3;

1.3
date	2009.12.08.07.41.15;	author AnnieSimpson;	state Exp;
branches;
next	1.2;

1.2
date	2009.12.08.02.53.06;	author AnnieSimpson;	state Exp;
branches;
next	1.1;

1.1
date	2009.12.03.21.14.33;	author AnnieSimpson;	state Exp;
branches;
next	;


desc
@none
@


1.5
log
@none
@
text
@%META:TOPICINFO{author="AnnieSimpson" date="1260715589" format="1.1" reprev="1.5" version="1.5"}%
%META:TOPICPARENT{name="WebHome"}%
---+ Compilation of Suggestions from the Advisory Group to the GISIN GBIF proposal

---++++++ _Comments/suggestions are listed in no particular order. Newer Comments are added at the bottom. The first ones are anonymous because they are posted by the moderator. Signatures are encouraged for any additional comment._

   * I presume you are aware that there is an international pest mapping consortium that seems to travel around the world talking about best approaches to pest mapping?  I suspect they may have addressed some of the issues you propose but also may be quite critical of the methods mentioned for mapping distributions - especially GARP.  I wonder whether a clearer link between the work you propose and something like [[http://www.nappfast.org/][NAPPFAST]] would need to be highlighted since the aims seem similar and NAPPFAST, at least for pests, seems quite a way down this road already.  Roger Magarey would seem to be the best contact in this regard.

   * I am not a huge fan of CLIMEX but what we really want is simply a measure of relative risk to NZ rather than a prediction that we would stake our careers on.

   * Neither Maxent nor GARP are best suited to projecting species potential distributions into novel climate situations (new continents, future climates).  To that end, I’d prefer to use CLIMEX.  

   * For expediency, we could select a set of species for which we have existing models.  We have a good coverage of ag pests (insects, pathogens and diseases).  

   * We could try some comparative stuff – it’s very popular at present.

   * [[http://www.europe-aliens.org/][DAISIE]] has some good data for probably fewer than 100 species in terms of their sub-national species distributions and I am not sure how well recorded some of these species are outside of Europe.

   * The modellers I know suggest that it is better to ensure you get the most extreme (from an environmental point of view) distribution points rather than lots of data.

   * [In NZ we are forming] a federated network integrating invasive species management for the 24 local government bodies and central government agencies in NZ.  That should go forward as a substantial project to build the [GISIN] network in late 2010, 2011. And that will be integrated into our national research efforts around invasive species prediction and management.

   * CPHST is deeply involved in producing LUCID keys/ resources.  Terrence Walters from CPHST- Ft Collins is our point person.  He is building a tool called IDSource which will enable users to quickly find electronic keys to the taxa of interest.

   * SOMS (Self-organising maps) refers to a [[http://www.csiro.au/science/SOM-Analysis.html][project]]... that has ranked the ‘survivability’ of many insects, pathogens and diseases. For example some wood boring insects arrive in SWP material but fail to establish because of their biology. We should include links to the data products he mentions, and we should only bother modelling insects, pathogens and diseases with high ‘survivability’ ratings.

   * For SOMS info see pubs by Worner, and more recently Paini.  [We have] data products that could probably suit your purposes with minimal additional effort.

   * ...Pathway monitoring will sample the arrivals.  Getting a handle on which ones to REALLY watch out far is more problematic...  This is where SOM analysis should be able to help tease things apart.

   * ‘contextualising the invasion potential with assets at risk (crops, conservation areas etc)’ ...refers to the need to deliver outputs (data products) to potential users. [We should] think about delivering data products to regional decision-makers (e.g. regional Plant Pest Organisations) and national decision-makers.

   * [We need to be concerned with the] potential for data cancer, where erroneous occurrence data is propagated through different databases.  It is a problem that can be magnified through the distributed database model.  I’m just flagging the issue because when the data is included in a model, the source of the error can get buried, and the problem magnified.

   * ...a “killer app” could also generate mega mistakes in microseconds if left to its own devices.  Models can multiply or magnify any data mistakes.  I feel that it is critical that there are ample opportunities for manual intervention in the modeling process, and a means of 1) documenting duff data, 2) probing individual data records for quality, 3) automating feedback to data providers in a user-friendly manner (for both parties) and 4) capturing assumptions about the nature of records used for modeling.

   * It would be very useful to be able to distinguish between records of transient populations from established incipient populations.  Records of transient, or casual populations can be extremely misleading, heavily distorting a model of a  species potential distribution.  A good example is the USGS DB on _Eichornia crassipes._  It took me some time to gather from experts where the species was able to overwinter in the US. As another example, Queensland fruit fly has been recorded from Alice Springs, but it could only persist there under lab conditions or in a parallel universe.

   * [Invasiveness is] subjective and context specific at a fairly granular level.  One person’s GMO crop can be a neighbours nasty, herbicide resistant weed.  I’m not sure how to deal with this, but I’m just flagging it as an issue that we may need to be aware of when pursuing this aspect of the project.

   * I think that it's critical to differentiate between presence points where a given species is invasive vs. naturalized vs. native (along the lines of the definitions from [[http://www3.interscience.wiley.com/cgi-bin/fulltext/120191550/HTMLSTART][Richardson et al.]], 2000 in [[http://www.wiley.com/bw/journal.asp?ref=1366-9516&site=1][Diversity & Distributions]]).<2E> I've worked with cheatgrass a fair bit, and if you were to construct a model based on where it 'exists' using the USDA Plants database, you'd conclude that all of North America is at risk - but, the invasion in the western U.S. is far different from a few trace plants in the east.<2E> So, from a modeling standpoint, it becomes really important to define whether a species is actually invasive at a presence point.

   * when we finished collecting expert opinion data in the southeast, we compared the results to point distributions collected by the bugwood network and some other sources.<2E> <b>We found that point presences were skewed towards smaller infestations and underrepresented the major infestations.</b> That makes sense to some degree because most people out collecting points are focused on EDRR rather than treating areas already consumed by invasives, but it is problematic in terms of modeling because we may not be capturing the major infestations as well as we should.

   * [How about creating] a related table that is joined to a world countries shapefilie that has a list of pests with a country-specific risk score taken from the SOM analysis?  With the right query, users could select the country, and generate the sorted pest list for that country.  We could also set up the query to work the other way and map country-level risks for each pest. (03dec09)

   * Add David Richardson (South African member of ISSG) to Advisory Group? [[http://academic.sun.ac.za/cib/team/staff/dmrichardson/default.asp][Web page]] (03dec09)

   * [One modeling group offers] environmental datasets for current and future (decadal) climate.  [Would] run multiple GCMs (7-18) and provide averaged predicted future distributions, with full measures of uncertainty. [Timeframe] approx. 2 months to run this on 10 species, with multiple GCMs, provide results and write up short report.  Costs would be approx. 8,000 Euros, which will cover 2 months of time of a modeler, plus a week of supervision.  [Includes] all the hardware, software and data needed as co-financing...

   * If the GISIN system allows file upload, there will also be a need for a data format/content quality analysis tool(s). Being able to map a database to a data provider that can be harvested via web services usually (not always) implies a degree of good data management by the provider. When you allow 'file upload' (which is a good idea) you also open the doors to badly structured data and content. It would be appropriate to have a set of local data quality checking tools for file uploads. (09dec09)

   * PRATIQUE has already identified several pest species that they want to focus on [let's synergize with them].

   * If this project is a proof of concept [we should model]... a) a marine alga e.g. Codium fragile; b) freshwater weed e.g. Lagarosiphon major; c) a vine e.g. Asparagus scandens and d) a mammal e.g. Sciurus carolinensis.  [This] would test the modelling approaches quite well as well as assess data availability for different taxa/biomes.

   * It seems there must be many CLIMEX and Maxent models out there.  

   * Does Darren have a catalogue of all the taxa that have been successfully modelled using CLIMEX as well as the underlying data with which they were parameterised?

   * Tom Stohlgren's group has been running many US species through Maxent. (Comment from NZ)

   *  Regarding target species, we have done a lot of work on the CAPS target species (exotic pest threats to the US).  As others have mentioned, it would be good to include a diversity of taxa such as a weed, an arthropod, a mollusk, and a plant pathogen.  <b>Giant African Snail might be a good example. </b> (Fieselmann 08Dec09) 

   * We have developed pathway information (or can develop) for many invasive species.  We also can link with industry for data that may be hard to obtain.   (Fieselmann, 08Dec09)

   *  Roger and I think that a comparison of models would be beneficial, to determine their relative strengths and weaknesses. (Fieselmann, 08Dec09)

   * I've attached a file ClimateRefs of some citations to peer reviewed pubs that contain projections of distribution changes with climate change - note, these deal with plants only.<2E> I imagine we could contact the authors about hosting their spatial data on GISIN (both projections and distribution data if they're from non-standard sources) - you're certainly welcome to my data, they're all !ArcGIS grids so I can make them whatever format you wish. (Bradley 08Dec09)

   * If we end up running an ensemble of models, which I think is a great idea, I'd like to add <b>Mahalanobis Distance</b> to the mix - I've been using that for invasive plant work along with Maxent and I think it performs quite well.<2E> I can run that for the target species if it's not already available at NREL [Colorado State]. (Bradley 08Dec09)

   * RE Needs Assessment: Provider needs? Or user-needs? The difference probably needs an explanation. Are we assuming that GISIN already knows who the users are and what they need? (09dec09)

   * RE "we will identify data requirements for ecological niche models for a selection of invasive species, and then invite the appropriate data providers to share their data with GISIN.":  FYI we are doing the same thing for our national weeds database. The key issue is that _providers and modelers want point data, and polygons from GIS served up as WFS._ None of the GBIF-like data harvesting utilities do that, and I doubt the GISIN provider does that either. (09dec09)

   * RE our creating models of species' potential ranges under different climate scenarios: a member of this Advisory Group has stated that all models are overcome by non-climate variables and rapidly decay to having zero reliable information content. (09dec09)

   * RE our creating maps of projected distributions under different climate scenarios: Using what climate layers? I assume you mean the IPPC scenarios. These use a relatively coarse grid, ok for global views but not for national. What NZ wants, for example, is to use our own local high resolution future climate layers, and I assume others would want to do the same. (09dec09)

   * RE notification of species present in neighboring areas: How will you deal with issues of disagreement with the data. When I worked at CABI we had an on-going issues with national government agencies every time we published new issues of 'pest & pathogens' maps. Threats of legal action etc. (09dec09)

   * Factsheets [will be generated] for how many species? It would help me if there were some more estimated metrics &#8211; even if very approximate. (09dec09)

   * What infrastructure will be used to assemble and deliver factsheets ? Presumably it would be developed using an existing framework &#8211; ruby on rails, drupal etc. Has this been costed in? (09dec09)

   * [How will outreach be achieved?] Is there a component for an online discussion forum &#8211; which might include everything from idents to sharing experiences on management. Perhaps picking up some of the experience of groups like Pestnet, and platforms like iSpot (a Drupal application)? I see these social networking environments just important (perhaps more important!) than data sharing networks. (09dec09)

   * <b>Precautionary note:</b> In general I feel you are promising to deliver a lot for very little funding. I think GBIF needs to wake up to the fact that you don&#8217;t deliver 'Killer Apps for 50k euro! Personally I&#8217;d go for a strategy of under-promising and over-achieving. Non delivery or partial delivery could backfire badly on the future development of GISIN. (09dec09)


-------------



---++++++ _Suggested functionality for the GISIN project, by Michael Browne_

I envisage the key deliverable to be a user friendly &#8220;Invasive Species Decision Support&#8221;  home page with the following:

 

By-line:

*The GISIN Invasive Species Decision Support system provides information support for monitoring and surveillance.*

 

Explanation of what&#8217;s available:

Distribution maps give the present range of key invasive species, and ecological niche models describe their potential range under present conditions and future climate change scenarios. SOMs provide country-specific risk scores for target species. Pathway, impact and management information is provided to assist with early detection and rapid response. It is expected that this &#8216;proof of concept&#8217; with its 10 target species will attract further funding so that similar data products can be developed for most of the world&#8217;s worst invasive species.

Interface:

   * Links to ID tools

   * A link to the GISIN search page 

   * Search by species:   See distribution maps and ecological niche models for each of 10 target species. Invasion potential is contextualised with assets at risk (Red list species, conservation areas, crops etc)... Zoom in and navigate to your country, or to the country of your neighbour or your trading partner.

   * Search by country:   If you select &#8216;Mexico&#8217;, you will see a map of the present distribution of our 10 target species in Mexico (if applicable), and the results of our 10 ecological niche models for Mexico (if applicable), along with the SOM country-specific risk scores for Mexico and fact sheets for the target species (featuring pathway, impact and management information). This information will help prioritise monitoring and surveillance.  



----------------


---+++++ _Suggested request for input from Advisory Group, by Michael Browne_

At this point we need to ask our advisors (especially those who have models that may suit our needs, or who wish to create those models):
   1. Have you got any suggestions for specific target species (Darren suggested a mixture of environmental and biodiversity IAS, but didn't name names)?
   2. Which target species (that may suit our needs) do you have models for and are you ready 
      A. to provide those models (we don't have a map-server for interactive maps/models?) or,
      B. for GISIN to link to them (we would prefer to have the model direct on the GISIN Mash up page e.g. we may want to set the default view to make it user friendly)? 
   3. What additional data products would you provide? Could you provide these data products for all our target species? How much would all this cost?

Then we have to look at GISIN-mediated data to see if we have good data on their suggested target species. Next we have to decide to run some additional models if there isn't a good match or if we don't have enough species with good data. In any case we have to run at least 5 new models to demonstrate the value of integrating GISIN and GBIF data. The existing models don't do that, but they add variety/target species to the final data products. 

-----------------

---+++++ _Thoughts on SOMs and disclaimers_

   * The "country-specific risk scores" are generated by SOMs (not us). They add value to our other deliverables.

   * Yes, we need to put language around the scores, and the predicted distributions, saying that they are only models and should not be the only source used for decision making (i.e. no warranties implied).

   * We have to have a disclaimer and we have to present the entire decision support system in terms of &#8220;this system is designed to facility prevention by providing ready access to information that may help countries identify threats and prioritise monitoring and surveillance. It is not foolproof. Countries should supplement this information with local knowledge and their own efforts to learn more about the threats they face. 

 



-- Main.AnnieSimpson - 03 Dec 2009 started the page

-- Main.AnnieSimpson - 08 Dec 2009 Added Suggested Functionality by Michael Browne; Added Suggested request for input by Michael Browne

-- Main.AnnieSimpson - 13 Dec 2009 Added thoughts on disclaimers; tentative modeling quote; comments by Fieselmann, Bradley, and others.
@


1.4
log
@none
@
text
@d1 1
a1 1
%META:TOPICINFO{author="AnnieSimpson" date="1260600456" format="1.1" version="1.4"}%
d49 41
a89 1
   * 
d138 12
d154 3
a156 1
-- Main.AnnieSimpson - 08 Dec 2009 Added Suggested Functionality by Michael Browne; Added Suggested request for input by Michael Browne@


1.3
log
@none
@
text
@d1 1
a1 1
%META:TOPICINFO{author="AnnieSimpson" date="1260258075" format="1.1" reprev="1.3" version="1.3"}%
d45 7
d102 1
a102 1
-- Main.AnnieSimpson - 08 Dec 2009 Added Suggested Functionality by Michael Browne; Added Suggested request for input by Michael Browne
@


1.2
log
@none
@
text
@d1 1
a1 1
%META:TOPICINFO{author="AnnieSimpson" date="1260240786" format="1.1" reprev="1.2" version="1.2"}%
d76 17
d95 1
a95 1
-- Main.AnnieSimpson - 08 Dec 2009 Added Suggested Functionality by Michael Browne
@


1.1
log
@none
@
text
@d1 1
a1 1
%META:TOPICINFO{author="AnnieSimpson" date="1259874873" format="1.1" reprev="1.1" version="1.1"}%
d45 1
d47 32
a78 1
-- Main.AnnieSimpson - 03 Dec 2009
@