I am working on a habitat model to predict changes in available steam habitat for acid- and thermally-sensitive aquatic species. The main goal is to combine stream temperature model results with existing acidity (ANC) results for the southern Appalachian Mountain region to evaluate the spatial extent of a habitat “squeeze” on these species. The relationship between air and water temperature will also be explored to generate future scenario results of habitat availability with changes in air temperature. I have been mostly working with non-spatial regression modeling techniques and I would like to explore the use of spatial statistical models to account for spatial autocorrelation among observations.
Here is my full study region:
Non-spatial regression model results for ANC and water temperature:
Here is a potential study area with relatively high ANC data density:
hi Todd,
I think it would be helpful for purposes of exploring this topic to try the following:
1) do a hot spot analysis of the ANC values. I’m assuming there is one ANC value per site.
2) export the ANC values to R and determine their nearest neighbor, then re-import to Arc and display (Charles has a tutorial on this for Wednesday).
3) do a hot spot analysis on temperature data. I’m assuming this temperature data can be summarized as single numbers for each site, such as maximum monthly temperature.
4) How do you expect ANC to be related to temperature? Can you write a study question?
5) What analyses have you conducted already? I don’t understand the map that you titled “non-spatial regression” model results. Are the blue streams intended to represent places where a species is limited by neither low ANC or high temperature?
6) Given your finer-scale data (Figure 3), what are you interested in learning? Do you want to test whether nearby values are more similar? Or do you want to link ANC values according to their position within a watershed and ask whether controlling for distance, ANC values in the same watershed are more similar than across watersheds (or vice-versa)?
Julia
Thanks for the comments Julia.
1) Sounds good
2) Ok
3) Sounds good. Yes, I have the water temperature data summarized as mean daily maximum for the month of July.
4) I don’t expect any relationship between ANC and temperature. I am asking: What is the extent to which habitat for acid- and thermally-sensitive species is currently “squeezed” by temperature in lower elevation reaches and low ANC in higher elevation reaches. I am also exploring air-water temperature relationships to answer: How will the current extent of available habitat for acid- and thermally-sensitive species be impacted under future air temperature scenarios?
5) I have used logistic and linear regression to predict the July mean daily maximum temperature AND the extent to which water temperature changes with air temperature throughout the study region. Yes, the blue strams on the “non-spatial regression” results map represent places where a species is limited by neither low ANC or high temperature.
6) I thought I would choose a smaller study region just to familiarize myself with the R and ArcGIS spatial stats tools. I certainly can use all of the available data, but didn’t want to get bogged down with processing times. I am hoping to explore using the R package called “SSN” for generating regional predictions of ANC and July mean max water temperature. I’d like to then compare these results to my non-spatial regression results.
Hey Todd,
In terms of spatial autocorrelation, I thought you might find this helpful, if you haven’t explored the local Moran’s I tool already: http://resources.esri.com/help/9.3/ArcGISDesktop/com/Gp_ToolRef/spatial_statistics_tools/cluster_and_outlier_analysis_colon_anselin_local_moran_s_i_spatial_statistics_.htm
I would run it with inverse distance + euclidean selected. Create a spatial weight matrixes file beforehand if you have a large dataset or multiple analyses on a dataset (instructions here: http://resources.esri.com/help/9.3/ArcGISDesktop/com/Gp_ToolRef/Spatial_Statistics_toolbox/modeling_spatial_relationships.htm)
I was curious how you isolated streams/portions of streams here from others? I am also working with stream data and really new at it!
Hi Jackie,
Thanks for the suggestions on the Moran’s I tool!
I’m not totally sure what you mean by “isolated streams/portins of streams here from others. Maybe check with me in class. However, I am working with stream data from the NHDPlus V2 dataset: http://www.horizon-systems.com/nhdplus/NHDPlusV2_home.php
This database is has all kinds of watershed attributes associated with each stream reach, if you need them, and has a bunch of other functionality built in. It might be worth exploring.