How to perform species distribution modeling using the software Maxent

this tutorial will show you how to perform species distribution modeling with the software MaxEnt to begin visit my website josh Banta comm and scroll down to the bottom of the tutorials tab to the section entitled species distribution modeling today's tutorial assumes that you've already done the tutorial where you decide which settings to use when running MaxEnt if you have not yet done that tutorial you can find it here on my website today we'll start by referring back to the species distribution modeling flowchart there are two ways to run MaxEnt one way is to filter occurrence points so that no two points are near each other and the other is to filter occurrence points and use a bias file that accounts for the spatial bias of the occurrence points I will be showing you today how to run MaxEnt with a bias file which is option two it's easy to take the information in today's tutorial and adapt it to running MaxEnt without about bias file simply do not include the bias file in the MaxEnt settings if that is the way that you want to do it instead go back to my website and down to the section how to perform species distribution modeling and click on the files needed for the tutorial click on the triangle within the download box in the upper right hand corner and select direct download one finished click file explorer navigate to your downloads folder and find the zip file that we just downloaded double click on it to see inside of the zip file take the MaxEnt folder within the zip file and drag it to your desktop then verify that the max and file has been uncompressed and placed on your desktop by navigating to your desktop and looking to see if it's there now return to my website notice that this tutorial requires the software MaxEnt you can download it at the link that I provide go down within the MaxEnt website prefer to download without providing this information and then click the button for download MaxEnt once the maxint software has been downloaded click on file explorer navigate to your downloads folder and find the zip file that we just downloaded it turns out it has the same name as the files needed for this tutorial and so Windows has added a 1 in parentheses that is the file that we need double click on it to look inside the zip directory next click ctrl n to open up another copy of file explorer and navigate to your desktop and into the MaxEnt folder take the file MaxEnt jar out of the zip file and drag it over into the MaxEnt directory on your desktop then close the zip file we are now going to open MaxEnt be sure that the copy of MaxEnt jar that you're opening is the copy that's within the MaxEnt folder on your desktop and not the copy within the zip file in your download directory double-click to open we need to change a bunch of MaxEnt settings before we can start first uncheck the button at the bottom left hand corner of your screen that says auto features next on the bottom right corner of your screen check the box for create response curves and check the box for dude jackknife to measure variable importance leave the box for make pictures of predictions checked set the output format to see log-log and set the output file type to dot ASC those are the defaults so keep them as they are we need to decide which feature types we will allow MaxEnt to use to do that we'll go back to file explorer and we'll open up a file called enm eval results csv this is the results from the previous tutorial about how to decide which MaxEnt settings to use when you're using your own data do not use the enm eval results dot CSV file provided to you with this tutorial instead use the enm eval results dot CSV file that you generated from running the enm evaluate function in the previous tutorial on your own data double collect the e/m enm eval results file to open it in Excel we need to look for the model settings that have a delta a ICC equal to zero in this case it's the model settings shown here l equals linear Q equals quadratic H equals hinge P equals product and T equals threshold so what this means is that we need to use the MaxEnt settings for linear and quadratic features only but not any of the other features again with your own real data you won't be looking at this file you'll be looking at the enm eval results file that you generated based on your own data and in your own data it's likely that the settings will be different and that you won't just be using linear and quadratic features that you'll be using some other features we will also be using a regularization multiplier of 0.5 again with your own data understand that the regularization multiplier that you select will very likely be different click on the coffee mug at the bottom of your screen to pull up MaxEnt again uncheck hinge features uncheck product features so that only linear and quadratic features are selected if it's your own data and the enm evaluate results indicated something different you would have different boxes checked here next set your output directory where the results will go click on browse navigate to your desktop and navigate to the MaxEnt folder on your desktops double click on the picture of the folder itself and not on the word Mac sense now click the icon that says create new folder underneath it when you hover over it as indicated on your screen change the name of this new folder to results double click on and then no don't double click your screen should look like mine here where the folder name indicated is results click on open next we need to change some more MaxEnt settings by clicking on the bar that says settings at the bottom under the basic tab check the box that says random seed under the Advanced tab check the box that says right plot data under the experimental tab check the box that says right background predictions be sure that all the other boxes are checked under the basic tab as you see here under the Advanced tab as you see here and under the experimental tab as you see here return to the basic tab we now need to decide on the replicated run type in the number of replicates we will refer back to this species distribution modeling flow chart from my website if there are fewer than 50 occurrence points we run MaxEnt with one-way if there are 50 or more occurrence points we run max n another way our occurrence points are in the file occ csv be sure that your occurrence points are set up exactly the same as in this file with species in the first column longitude in the second column latitude in the third column and the names of the columns the same is here all in lowercase for the species name pick any name for your species that you like just make the name short and don't put any spaces in it and you'll take this is only to help you remember what the name of the species is for when you ran this analysis you'll take whatever name you choose and you'll populate it throughout the spreadsheet all the way down to your last record of the occurrence point as you can we have more than 50 occurrence points therefore we were we are going to run MaxEnt using cross-validation with 10 model folds I'll click on the picture of the coffee in the coffee mug icon at the bottom to open up max n again I'll pick a click on the picture of the coffee mug and select the settings window to open up the settings window again I will keep the replicated run type as cross validate and I'll set the number of replicates to 10 corresponding to 10 model folds this tutorial builds on the previous one that you that you did where you chose the MaxEnt model settings in that previous tutorial I told you that we didn't have enough places on the landscape to use 10000 background points for the purposes of this tutorial we will use 5000 background points however in most cases you should use 10000 background points refer to the earth earlier tutorial on how to choose the right MaxEnt settings to decide whether to use 10000 background points or some smaller number for the purposes of this tutorial use 5000 usually with your own data I will caution you you are going to want to use 10000 background points but again see the previous tutorial to decide how many background points is appropriate for you change the regularization multiplier to what was indicated in the enm eval results CSV file again don't use the enm eval results dot csv file of the tutorial data if you're using your own real results for your own real results use the enm eval results dot CSV file that was generated on your own real data in our case we need to use a regular it for this tutorial we need to use a regularization multiplier of 0.5 the plier here under the basic tab 20.5 next under the advanced tab I need to add in the bias file click on the Browse button where you see bias file at the bottom of the options window navigate to the MaxEnt folder on your desktop and choose bias file ASC that's our bias file and click open we now have set all of the settings for MaxEnt and we can close the MaxEnt settings window now we will tell MaxEnt where our occurrence points are in the upper left hand corner click browse click OCC CSV which contains our occurrence points do not use OCC CSV from the tutorial if you're doing this on your own data you will want to use your own actual species occurrence points file instead I will click on open to open up OCC CSV we know that it imported properly because we see the name of the species showing up here next to a check box next we need to put our environmental layers into their own subfolder go back to the file explorer window and navigate to the MaxEnt folder on your desktop click the button new folder and rename this folder whatever you like in my case I'm calling it layers we're going to put all of our environmental layers into the layers folder our environmental layers our bio 1 bio 5 bio 6 bio 7 bio 8 bio 12 bio 16 bio 17 biotech 1 and bio cat tech 2 a short you can either drag those files one at a time into your layers folder or as a shortcut highlight the first of the files that you want to drag go to the bottom hit shift and then highlight the last of the files that you want to drag by clicking on the picture of one of the icons you can then drag all of those files together in batch into your layers subfolder if you have trouble with that method simply drag the files one at a time into the layers folder please note that I did not drag the file bias filed ASC into the layers subfolder because bias file that ASC is not an environmental layer click on the layers folder and verify that you have only the files that you see here within your layers subfolder and that it's within the MaxEnt folder on your desktop these environmental layers 501 through bio 17 are taken from a website where you can download that sort of environmental data called world climb and if you visit the website world climb org slash bio climb you can see translations of what the names of these environmental layers actually mean please note that I made up the environmental layers bio cattigan and bio cat tech 2 so you will not find them on the world climb website now we are ready to tell MaxEnt where our environmental layers are click on browse in the upper right hand corner navigate to the MaxEnt folder on your desktop and then select the layers folder so that it's selected as you see here on my screen click open Maxon has imported all of your environmental layers we need to change any environmental layers that we have as categorical if they're categorical layers in the case of the sample data by ok egg 1 and biotech 2 are two categorical environmental layers for more information on categorical environmental layers see the previous tutorial on choosing the right MaxEnt settings if you have a categorical variables in your own data you're going to need to change them just as I am doing here so we will change by ok egg 1 and by ok tag - - categorical by clicking on the little arrows and then selecting categorical from those drop down menus we are now ready to run Max n do so by clicking the Run bar in the bottom left corner running Matson can take quite a while even with the tutorial data here which is a smaller data set it takes about three to five minutes to run keep an eye on your screen when the progress bar stops running over and over again and the progress window is gone it means that Maxim has finished running once MaxEnt is finished we can go look at the results navigate back to the MaxEnt folder on your desktop and double-click on the results folder the file with your results in it will be the name of your species dot HTML for the purposes of this tutorial our species is Brad applause so the file that we're going to double click on is Brad a post out HTML in the case of your own data the name of this file will be whatever your species is as a caution do not open the files that have your species name underscore and a number dot HTML only open the one without an underscore in a number as you see at the top here double click on it and we can see many different results describing what these results mean is beyond the scope of this tutorial but briefly I will give you some of the main features notice that you have the test day you see indicated here notice that we have our species distribution model here on the left on the right this is not the species distribution model disregard the image on the right with the image on the left you can click on it to just see the image by itself you can click the magnifying lens to zoom in and you can right click on the image and save the image out as a picture you if you want your Mac sent results as an as a SC raster file that you can import into software like QGIS you would come to your Mac sent folder on your desktop click on the result subfolder and you would click on the file Brad APIs dot you Braddock was average ASC that's your Mac sent results written out as a raster that you can import into whatever GIS software that you like back to showing you the results the next thing that I want to highlight for you is this section here called jackknifed of test gain for Braddock was describing in depth how to interpret these results as outside of the domain of this tutorial but I'll briefly give you the main features do not look at the jackknife of a you see do not look at the jackknife of regularized training gain only look at the jackknife of test gain the red bar shows you the test gain of the full model that contains all of the environmental variables the dark blue bars show you the test gain of the environmental variable of the of the models that were fit using only the one environmental variable indicated in that row the light blue bars show you the test gain of the model when the variable indicated on that row was left out of the model and everything else was included looking at the jack of test gain for the different environmental variables or environmental layers in other words can show you how important the different environmental layers are to the species distribution model notice we have this section here called response curves it's likely that the most important response curves that you will need to see are the second set of response curves down here not the first set up here there's these response curves show you how habitat suitability is influenced by the environmental variable of interesting so for instance if we click on this response curve here under bio 1 the y-axis here is habitat suitability the x-axis is bio 1 which is that bio climatic variable from world climed org and in this case it's average annual temperature times 10 the point is is that as average annual temperature increases the abot the habitat suitability for bradda post also increases up to at least a certain point for categorical variables instead of a response curve we see bar graphs and what you want to look at are the dark red and we see that the habitat suitability is greatest for Brad opposed in a category type one as opposed to category type negative 1 or category type 3 for your own data be sure to know what these different category types of your categorical variable mean so one could be a particular soil type negative one could be a different soil type and three could be yet another soil type that are arbitrarily called type 1 type negative 1 and type 3 for instance so that's how you interpret your results next I want to show you where you can find this response curve data within your results you find that within the MaxEnt - folder on your desktop within the results subfolder within the plots sub subfolder and I have another tutorial which describes how you can take this response curve data out of here and plot it in whatever graphical software that you like but going into how you would plot this response curve data goes beyond the scope of this tutorial another thing I want to draw your attention to is a file in the results subfolder of the Macs and folder on your desktop that's called Mac sent results dot CSV this file contains all of the information about the the summary statistics that you could need understanding what these summary statistics meaning is outside the scope of this tutorial but understand that this is where you can find them and you are going to want to look at the results in the row called average so if we wanted to see the exact test gain of the full model we would look in this row labeled average we would see that it's point eight two six two we would see that the test AUC is point eight two six and so on and so forth if we wanted to see the if we wanted to see the test gain of the model that was built using only the the variable by oh five again we would look at this row for the average and we would see that that test gain is point 102 9 the is you can find all that information in this file Mac sent results dot CSV next I want to show you how to run MaxEnt when you have fewer than 50 occurrence points before we get started return to the MaxEnt folder on your desktop not the results subfolder make sure that you're in the MaxEnt folder itself click on new folder and we'll add a new folder called results to next reopen MaxEnt and we're going to change our output directory in the bottom right corner from results - results - and we're going to do that by navigating to the MaxEnt folder on our desktop and selecting results - as the output this is so that we won't override the results that we have from the other analysis with the new results this way we'll put the results into their own new folder now instead of using OCC CSV we're going to use this other set of occurence points called OCC small dot CSV we're going to pretend that our data set is smaller so come here to the upper left corner of your Mac sense screen and click browse change to OCC small dot CSV within the MaxEnt folder on your desktop and click on open everything will look the same as before except that everything will look the same as before except that now the file that's selected will be called OCC small dot CSV oops and make sure that you have Brad apos checked let's go and look at OCC small dot CSV and see how many occurrence points we have click on file explorer and double click on OCC small dot csv to open up you we have only twenty occurrence points if you go back to the niche model flow chart when we have fewer than 50 occurrence points we have to use something called the leave one out method when you click on maxint to open it again and I'm going to click on settings click on the basic tab here we're going to change the replicates to equal the number of data points in our sample I'll come back to the file explorer window and reopen that file again we have 20 occurrence records I know that you're seeing that it looks like 21 but notice that the first one is not an occurrence record those are just column headers so we actually have 20 occurrence records so for this method we will set the number of replicates to equal the number of occurrence records which is 20 and we will still keep the replicated run type as cross validate now we have all the settings we need we can close it and we can click the bar at the bottom left hand corner that says run do note that it will take MaxEnt longer to run using the leave one out method you go ahead and go ahead and click the button run for the progress window to close when it's done when finished navigate to the MaxEnt folder on your desktop within file explorer click on the folder results to open the file Brad a post at HTML as before now we'll see that we the AUC is point eight seven four and we can scroll down and see the MaxEnt model and the response curves the jackknife of test gained and if we look within File Explorer we also have that file MaxEnt results dot CSV and in plots although I I'm not going to go into it here we also have the response curve data that you could plot in your own graphical software and that's how you would run MaxEnt if you had fewer than 50 occurrence points that concludes this tutorial on running Maxo

Channel: Josh Banta

Views: 18,810

Keywords:

Id: f9vwrZf6ncU

Length: 30min 2sec (1802 seconds)

Published: Mon Apr 08 2019