I'll try to answer your specific questions below, but I'm skipping the first 5 for now (sorry to disappoint). By the way, thanks to everyone who has sent information off-list so far! From: [EMAIL PROTECTED] To: [EMAIL PROTECTED]; ECOLOG-L@LISTSERV.UMD.EDU Subject: SCIENCE Methods for making more sense than nonsense by sampling reality Re: [ECOLOG-L] Inferring weed distribution from herbarium records and GIS layers Date: Tue, 15 Jul 2008 13:02:25 -0700
For Dr. Quinn (and for comment by the Forum): 1. What are you trying to demonstrate or infer?This herbarium dataset was obtained as one small part of a larger project looking at the effects of peri-urban development and water quality on aquatic macrophytes in Australia. Because I found a series of GIS layers that classify land use and cover (almost) the whole of the continent and because I had this great continental-scale database of most of the aquatic weeds, I thought it would be interesting to see if there were strong correlations between land use and aquatic weed distribution on a national scale. I added another GIS layer, one that characterizes dominant vegetation and cover, because this appears to be the type of data that a lot of other published papers rely on to characterize land use (that is, I don't think there are a lot of widely-available GIS layers out there that specifically characterize polygons by land use, rather than by some proxy of land use like vegetation cover). 2. How did you determine the relevance of "100 km of a city," and to what factors was that considered relevant? I'd never heard the term "peri-urban" before coming to Australia, but it seems to be a buzz word here. However, it doesn't seem to be strictly defined anywhere that I know of and people use it to refer to any sort of transition zone between truly urban areas and completely undisturbed ones. Since my project is specifically related to peri-urban land use and I want to capture as many of those fuzzily-defined areas as possible, I somewhat arbitrarily chose 100 km radii. That could be easily changed to any other value. 3. How does time figure into your investigation? How are the 1580 records (ca 55 records per species, or how are the species distributed as a fraction of the total) distributed over time?Not sure about time. But, no, in the 1580 records, some species are more common than others. I can look at them singly, as "types" (emergent, submerged, free floating, etc), or as the whole lumped group. 4. How will the results of your investigation fit into the existing literature on the subject, and how will it advance it? As far as I know, there are very few papers that look at the distribution of freshwater aquatic species on a scale like this. There are studies that link aquatic weeds with land use, but this, again, would put that on a new scale. Also, it may be useful for land managers who could spend more time in the right areas to prevent or control invasions.5. Are there any ArcGIS layers that are not applicable or useful to your data set or investigation? Probably, but I don't think I'm adding more and more layers just because I can. As I have inferred, I tend to believe that "anecdote is the singular of data," so think there must be a "pony" in there somewhere if you keep looking long enough. Just what kind of pony, how big, and how fast--who knows? The important thing, it seems to me, is the quest itself. Even if you find that it is invalid to use herbarium records to interpret distribution, or that the interpretation possibilities are severely limited, you will still have made a significant contribution. Negative results are still results. Yes, I hope I can find a way to use this data to make that contribution. It seems like I have everything I need, but I still can't get my head around sampling from map data in a statistically sound sort of way. Some papers use logistic regression to explain presence/absence based on map-extractable environmental data, but I'm don't think my "absences" really count as true absences (some of you have brought up "presence-only" analyses off list). Anyway, still working on it. If anyone has any more thoughts now that I've explained a little more of my project, please send them my way. Thank you!Lauren I hope you will keep the Forum informed as your study progresses. WT ----- Original Message ----- From: "L Quinn" <[EMAIL PROTECTED]> To: <ECOLOG-L@LISTSERV.UMD.EDU> Sent: Monday, July 14, 2008 6:37 PM Subject: [ECOLOG-L] Inferring weed distribution from herbarium records and GIS layers Dear list, I am relatively new to ArcGIS and its ecological applications, so please excuse me if this question seems naive or wrongheaded... I have obtained all of the existing herbarium records for 29 aquatic weed species in Australia (approximately 1580 records total), as well as several GIS layers showing things like land use, dominant vegetation type, cover class, etc. Basically, I would like to be able to demonstrate that the spatial pattern I'm seeing is statistically valid, but I'm not sure how to do that. The principal spatial pattern I see from "selecting by" the various polygon features in my GIS layers is that the density of aquatic weed records is greater in "intensive" land use types (e.g. urban residential areas) than in other types. I derived density values by taking the total number of herbarium records (points) falling within those selected polygons and dividing by the total area (in km2) of the selected polygons. The problem is that this leaves me with only one density value for each land use type, which is, of course, not possibly to analyze statistically. How does one "replicate" when sampling from a map? I also went through the exercise of picking out each individual point (herbarium record, so each point is an individual of a particular species) and characterizing it in terms of the land use type, vegetation type, and cover class it sits in and whether or not it falls within 100 km of a city, but I am not really sure what I can do with that dataset. It is, at least, much bigger than the 5 density data points I have. If you can see an obvious solution to this or know of instructive texts or papers, please let me know. If you think there's nothing I can do with this dataset, I suppose that's good (but depressing) information too. If you have comments about the validity of using herbarium records to interpret distribution, I am somewhat aware of the issue already. Thank you. Lauren Quinn _________________________________________________________________ Making the world a better place one message at a time. http://www.imtalkathon.com/?source=EML_WLH_Talkathon_BetterPlace= _________________________________________________________________ Stay in touch when you're away with Windows Live Messenger. http://www.windowslive.com/messenger/overview.html?ocid=TXT_TAGLM_WL_messenger2_072008