Is this reptile's movement restricted to the riverbed, and is that why you are forcing your random points to be within the riverbed? As for choosing a random point on the ground to conduct your vegetation registration....your GPS is already generating that randomness for you. I suggest that you simply walk to the point where the GPS indicates you should be (i.e, when the "distance to waypoint = 0" for the first time). This GPS location is the actual (true) location + random error (

My suggestion is, when doing habitat selection studies, do get an overview of the habitats available to your species, and then analyse if the animals use the different habitats more than would be expected from the availability of the habitat. A sort of chi-square-thinking, use versus availability. To define the area which will be defined to be available to your animals, you need to know something about their range, how far do they walk, are there any part of the area they cannot access etc. For instance, a buffer along the river, as you suggest. After defining the available area, I would classify the area into different habitats or vegetation types. Then you  need to know what it is about the environment that is important to your animals, e.g. presence of water, rocks for hiding, something for eating (sorry, I do not know your species at all :o). Then you can either create random points or make a stratified randomised sampling design, randomised in order to make sure all you vegetation types are visited. Good luck. Regards, Trine

Is this reptile's movement restricted to the riverbed, and is that why you are forcing your random points to be within the riverbed? As for choosing a random point on the ground to conduct your vegetation registration....your GPS is already generating that randomness for you. I suggest that you simply walk to the point where the GPS indicates you should be (i.e, when the "distance to waypoint = 0" for the first time). This GPS location is the actual (true) location + random error (

In my answer, I assumed that you need to select points to visit, and then use the GPS to get there. If your animals are actually wearing a GPS colar, then it is a different mattet, then your positions are collected for you, and no sampling design is needed. Trine 

Hi Ronny,

Let's consider that we don't know a damn thing about your study area and topic. Already you have some boundary conditions for random sampling, due to your GPS accuracy. You state that your portable GPS has a positional accuracy of about 5 m. OK. Then the maximal relevant resolution of a remote sensing image of your study area is about 10 m, considering that you can be 5 m off from a preferred position in all directions.

Now what about using a Landsat ETM + multi-spectral image with a resolution of 30 m? Or several if there is seasonality in your study area? You can classify your vegetation using this type of imagery, by a supervised classification technique, since you undoubtedly already have the field data available of your vegetation types (in function of seasonality, temperature, precipitation or both of the two last variables cited).

Next step is that you can perform field sampling in function of vegetation or even better  at random. For random sampling I would overlay a grid of 30 x 30 m over your study area. You can determine the coordinates of the mid-points of the grid quite easily, since they correspond with the center points of your image pixels. Mind that your image has to be in a surface conform projection, to go ahead like this.

In a next step figger out a multi-temporal sampling scheme to enable the whole image pixel grid to be sampled in the field. Since you have your vegetation types classified, you just have to look for the presence of you reptile(s) in 30 x30 m grid cells corresponding with your image pixels, using your GPS or even better fixed points which you have indicated in the field once (and for all) corresponding with your image pixel coordinates, reducing your positional error significantly along a multi-temporal sampling time axis.

Once you have your desired number of samples per pixel, preferably about more than a 100 positive observations of your reptile species at the pixels where observation frequency is maximal and over all seasonal periods, you can go back to your PC or laptop and start to do the interesting work,... multi-temporal geo-statistics.

You can make frequency distributions per pixel and in function of time of the reptile species you have observed in function of time, I guess preferably with an hourly resolution. I figger that for reptiles the radiation and energy budget in your area of study is quite important. You would be able to visualize this with your frequency distributions in function of time. Vegetation in this respect will play a significant role, because it does have a strong effect on the energy balance of a site even at the scale of the size of a reptile, which likes to heat up in the morning to become active, and to cool down, when radiation is too intense when the sun has reached its zenith and hence a reptile will look for a spot being shadowed (by vegetation or rocks). The Beatles would sing, 'That's all I need!". So do the Reptiles ;-)

Of course the reptiles need prey too, unless they are herbivores, then they need vegetation not only as source of shadow, but as source of food as well. I guess, you can learn a lot about the behaviour of the reptiles this way.

Keep me posted if you like the scheme,

Frank

Thanks for reply. 

To Trine: The data of the reptile observations are already collected. I only need to measure the habitat availability. I also agree that the number of random points should ensure that I cover all vegetation types. 

To Jason: I agree with you that when I use random generated points I could simply walk to the point where the GPS indicates I should be (i.e, when the "distance to waypoint = 0" for the first time).

To Frank: Thank you for the suggested method. I will see if Landsat ETM + multi-spectral image are available for the study area. If so, I will try this approach in addition to the more simple habitat selection model (i.e. reptile observation vs. random point, binomial, as function of vegetation). 

Ronny, before advising, I would have to know some about the scale around which you assessed vegetation for your reptile points because whatever random points you choose, they should have the same scale. I can say, though, that both the reptile points and the random points should be large enough for you to feel comfortable about the vegetation assessment, that is, as your point size increase, somewhere you asymptote on further improving the vegetation description for that point, and that depends on the local small-diversity of vegetation at the site you are working.  Further, determining what is truly available to your reptile species depends on the scale at which they use habitat over the analysis time period of interest.  We did a similar study on a ranid frog species in a marsh and collected vegetation data at the scale of 1 m around each frog, that is vegetation was assessed in a 0.5 m radius circle.  However, since we wanted to assess temporal fluid parameters of microhabitat, such as temperature, we made measurements of frog versus random points within 15 minutes of one another to avoid a diel bias (which is probably not an issue for the parameters you are measuring). As a consequence, though, we decided that the time scale associated with the footprint that might available to a frog for selection of the random point (for each frog point) as a distance that was within the movement distance a frog might make within a summer day (which we as assessed as 6 meters bases on telemetry data).  In your case, I assume that you are working in GIS of a low aerial photograph that was taken within a time interval that can be reasonably expected to represent the time at which you collected your reptile points; the more disparate in time the aerial photograph you use in GIS is from when you collected the reptile points on the ground, the more error is introduced about the reality of vegetation in your reptile and random points. We then made sure that the footprint of the random points never overlapped the footprint of the frog points (or in your case, the reptile points).  In our case, we assessed vegetation in a 0.5 m radius circle around the frog, so random points could be chose with the centers at distances of 1.0 to 5.5 meters from individual frogs at randomly chose distance and direction. This enabled the footprint of random points to fall anywhere within 6 m excluding the 0.5 m circle around the frog.  Hence, if you are trying to do something similar with use versus availability, you need to know the diel scale of movement of your reptile, which will set the limits of your available footprint. The resolution at which we chose a random direction was 1 degree and the resolution at which we chose a random distance was 1 centimeter, but those choices were dependent on our scale, you would have to consider random selection of distance and direction with a resolution appropriate to your scale.

Hi, Ron,

I share Jasons dislike for "close-your eyes-and-toss-a-coin" and it is also true, that your method at present has a strong bias towards near-river habitat points. 

I suggest, that you first find out, how far the reptiles go away from the river. If you have no better information, you could use the median distance of your reptiles from the river and multiply it by 2. Then you could generate random points along the river course and make transects originating from these points perpendicular to the river into the surrounding land - as far as your estimated corridor of reptile habitat goes.

Frank also made an important point: vegetation may only be a source of shade and insects for the reptiles, so plant morphology may be a more important factor than plant taxon. Are birds preying on your reptiles? If yes, you may want to include a parameter like "length of covered path within one 1m² " in your survey.

best wishes

Irene

I hope you enjoy the references attached.

jwalter