SpatialDataScience/14_WallaceSDM.html at dev · learner-python-R/SpatialDataScience

History

647 lines (568 loc) · 27.4 KB

Raw

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

225

226

227

228

229

230

231

232

233

234

235

236

237

238

239

240

241

242

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

269

270

271

272

273

274

275

276

277

278

279

280

281

282

283

284

285

286

287

288

289

290

291

292

293

294

295

296

297

298

299

300

301

302

303

304

305

306

307

308

309

310

311

312

313

314

315

316

317

318

319

320

321

322

323

324

325

326

327

328

329

330

331

332

333

334

335

336

337

338

339

340

341

342

343

344

345

346

347

348

349

350

351

352

353

354

355

356

357

358

359

360

361

362

363

364

365

366

367

368

369

370

371

372

373

374

375

376

377

378

379

380

381

382

383

384

385

386

387

388

389

390

391

392

393

394

395

396

397

398

399

400

401

402

403

404

405

406

407

408

409

410

411

412

413

414

415

416

417

418

419

420

421

422

423

424

425

426

427

428

429

430

431

432

433

434

435

436

437

438

439

440

441

442

443

444

445

446

447

448

449

450

451

452

453

454

455

456

457

458

459

460

461

462

463

464

465

466

467

468

469

470

471

472

473

474

475

476

477

478

479

480

481

482

483

484

485

486

487

488

489

490

491

492

493

494

495

496

497

498

499

500

501

502

503

504

505

506

507

508

509

510

511

512

513

514

515

516

517

518

519

520

521

522

523

524

525

526

527

528

529

530

531

532

533

534

535

536

537

538

539

540

541

542

543

544

545

546

547

548

549

550

551

552

553

554

555

556

557

558

559

560

561

562

563

564

565

566

567

568

569

570

571

572

573

574

575

576

577

578

579

580

581

582

583

584

585

586

587

588

589

590

591

592

593

594

595

596

597

598

599

600

601

602

603

604

605

606

607

608

609

610

611

612

613

614

615

616

617

618

619

620

621

622

623

624

625

626

627

628

629

630

631

632

633

634

635

636

637

638

639

640

641

642

643

644

645

646

647

<!DOCTYPE html>

<head>

<title>Species Distribution Modelling with Wallace</title>

pre:not([class]) {

background-color: white;

}

</style>

if (window.hljs) {

hljs.configure({languages: []});

hljs.initHighlightingOnLoad();

if (document.readyState && document.readyState === "complete") {

window.setTimeout(function() { hljs.initHighlighting(); }, 0);

}

</script>

h1 {

font-size: 34px;

}

h1.title {

font-size: 38px;

}

h2 {

font-size: 30px;

}

h3 {

font-size: 24px;

}

h4 {

font-size: 18px;

}

h5 {

font-size: 16px;

}

h6 {

font-size: 12px;

}

.table th:not([align]) {

text-align: left;

}

</style>

</head>

<body>

.main-container {

max-width: 940px;

margin-left: auto;

margin-right: auto;

}

code {

color: inherit;

background-color: rgba(0, 0, 0, 0.04);

}

img {

max-width:100%;

height: auto;

}

.tabbed-pane {

padding-top: 12px;

}

.html-widget {

margin-bottom: 20px;

}

button.code-folding-btn:focus {

outline: none;

}

</style>

/* padding for bootstrap navbar */

body {

padding-top: 51px;

padding-bottom: 40px;

}

/* offset scroll position for anchor links (for fixed navbar) */

.section h1 {

padding-top: 56px;

margin-top: -56px;

}

.section h2 {

padding-top: 56px;

margin-top: -56px;

}

.section h3 {

padding-top: 56px;

margin-top: -56px;

}

.section h4 {

padding-top: 56px;

margin-top: -56px;

}

.section h5 {

padding-top: 56px;

margin-top: -56px;

}

.section h6 {

padding-top: 56px;

margin-top: -56px;

}

</style>

// manage active state of menu based on current page

$(document).ready(function () {

// active menu anchor

href = window.location.pathname

href = href.substr(href.lastIndexOf('/') + 1)

if (href === "")

href = "index.html";

var menuAnchor = $('a[href="' + href + '"]');

// mark it active

menuAnchor.parent().addClass('active');

// if it's got a parent navbar menu mark it active as well

menuAnchor.closest('li.dropdown').addClass('active');

});

</script>

$(document).ready(function () {

window.buildTabsets("TOC");

});

</script>

$(document).ready(function () {

// move toc-ignore selectors from section div to header

$('div.section.toc-ignore')

.removeClass('toc-ignore')

.children('h1,h2,h3,h4,h5').addClass('toc-ignore');

// establish options

var options = {

selectors: "h1,h2",

theme: "bootstrap3",

context: '.toc-content',

hashGenerator: function (text) {

return text.replace(/[.\\/?&!#<>]/g, '').replace(/\s/g, '_').toLowerCase();

ignoreSelector: ".toc-ignore",

scrollTo: 0

};

options.showAndHide = true;

options.smoothScroll = true;

// tocify

var toc = $("#TOC").tocify(options).data("toc-tocify");

});

</script>

#TOC {

margin: 25px 0px 20px 0px;

}

@media (max-width: 768px) {

#TOC {

position: relative;

width: 100%;

}

.toc-content {

padding-left: 30px;

padding-right: 40px;

}

div.main-container {

max-width: 1200px;

}

div.tocify {

width: 20%;

max-width: 260px;

max-height: 85%;

}

@media (min-width: 768px) and (max-width: 991px) {

div.tocify {

width: 25%;

}

@media (max-width: 767px) {

div.tocify {

width: 100%;

max-width: none;

}

.tocify ul, .tocify li {

line-height: 20px;

}

.tocify-subheader .tocify-item {

font-size: 0.90em;

padding-left: 25px;

text-indent: 0;

}

.tocify .list-group-item {

border-radius: 0px;

}

</style>

</div>

</button>

<a class="navbar-brand" href="index.html">GEO 503: Spatial Data Science</a>

</div>

<li>

</li>

<li>

<a href="Syllabus.html">Syllabus</a>

</li>

<li>

<a href="Schedule.html">Schedule</a>

</li>

Content

</a>

<li>

<a href="CourseContent.html">About Course Content</a>

</li>

<li class="dropdown-header">Module 1: Introduction to R</li>

<li>

<a href="00_CourseIntroductionFrame.html">00 Course Introduction</a>

</li>

<li>

<a href="01_Rintro.html">01 First Steps</a>

</li>

<li>

<a href="02_graphics.html">02 Graphics</a>

</li>

<li>

<a href="03_DataWrangling.html">03 Data Wrangling</a>

</li>

<li>

<a href="03b_DataWrangling.html">03 Data Wrangling 2</a>

</li>

<li class="dropdown-header">Module 2: Spatial Analyses</li>

<li>

<a href="04_Spatial_with_sf.html">04 Spatial Data with sf</a>

</li>

<li>

<a href="05_Raster.html">05 Spatial Raster Data</a>

</li>

<li class="dropdown-header">Module 3: Advanced Programming</li>

<li>

<a href="06_CreatingWorkflows.html">06 Creating Workflows</a>

</li>

<li>

<a href="07_Reproducibile.html">07 Reproducible Research</a>

</li>

<li>

<a href="08_WeatherData.html">08 Weather Data Processing</a>

</li>

<li>

<a href="09_RemoteSensing_appeears.html">09 Satellite Data Processing</a>

</li>

<li>

<a href="11_ParallelProcessing.html">11 Parallel Processing</a>

</li>

<li>

<a href="12_DynamicVisualization.html">12 Dynamic Visualization</a>

</li>

<li>

<a href="13_SDM_Exercise.html">13 Species Distribution Modeling</a>

</li>

</ul>

</li>

Assignments

</a>

<li>

</li>

<li>

<a href="DataCamp.html">DataCamp</a>

</li>

<li>

<a href="PackageIntro.html">Package Introduction</a>

</li>

<li>

<a href="Project.html">Final Project</a>

</li>

</ul>

</li>

Resources

</a>

<li>

<a href="Provenance.html">Provenance</a>

</li>

<li>

<a href="Projects_2017.html">2017 Final Projects</a>

</li>

<li>

<a href="Resources.html">Resources</a>

</li>

<li>

<a href="GitSSHNotes.html">Setting up Github</a>

</li>

</ul>

</li>

</ul>

<li>

</a>

</li>

</ul>

</div>

</div>

</div>

<h1 class="title toc-ignore">Species Distribution Modelling with Wallace</h1>

</div>

<h1>Setup</h1>

<p>For <code>wallace</code> to work, <strong>you need the latest version of R</strong> (or at least later than version 3.2.4). Download for <a href="https://cran.r-project.org/bin/windows/base/">Windows</a> or <a href="https://cran.r-project.org/bin/macosx/">Mac</a>.</p>

<p>Load necessary libraries.</p>

<pre class="r"><code>if (!require('devtools')) install.packages('devtools')

devtools::install_github('rstudio/leaflet')

install.packages('knitr',dep=T) # you need the latest version, even if you already have it

install.packages('wallace',dep=T)

library(wallace)

run_wallace() # this will open the Wallace GUI in a web browser</code></pre>

<p>The <code>wallace</code> GUI will open in your web browser and the R command line will be occupied (you only get a prompt back by pushing ‘escape’). If you find this annoying and want to use your typical R command line, open a terminal window, type <code>R</code> to start R, and then run the lines above in that window. This will tie up your terminal window but not your usual R command line (e.g. RStudio, or the R GUI). <strong>You need to avoid exiting your browser window or closing the R window that initiated <code>wallace</code> or you’ll have to start over! Luckily that’s pretty fast…</strong> Using the terminal looks like this:</p>

<p><br> Typing <code>run_wallace()</code> will give you the following in your web browser:</p>

</div>

<h1>Get Occurrence Data</h1>

<p>Start by getting about 300 records of <strong>Acer rubrum</strong> (red maple) from GBIF. Throughout, I’ll use a red arrow in the images below to indicate which buttons I’m referring to.</p>

<p><br> While you’re at it, download the data for later use (bottom left).</p>

<p>Notice that there are tabs along the top, and you can view the sources of the occurrence data. Later you can choose to ditch some if it looks suspect.</p>

<p><br> Each <strong>Module</strong> (the tabs labeled 1-8 at the top of the screen) comes with guidance and references by select the tabs at the right.</p>

</div>

<h1>Prep Occurrences</h1>

<p>Now let’s clean up the data. If we want to model <em>A. Rubrum</em> in the US, we can toss that odd point in Europe. Click the point to see it’s info and then enter the ID at the left to remove it.</p>

<p><br> The samples may exhibit spatial autocorrelation, which is best to account for in the model or remove before modeling. For example, there might be a bunch of samples near cities because these are mostly from iNaturalist (citizen science) and citizen often live near cities. So lets spatially thin the points and make sure they’re all at least 10km from one another. It takes a sec. That left me with 163 points for modeling (yours may be different).</p>

<p>Download these points for later reference.</p>

</div>

<h1>Get environmental data</h1>

<p>Now we need some covariates to describe occurrence patterns. Worldclim is global climate data base that is very popular to both use and complain about. It seems pretty good in regions with lots of weather stations, but has issues, especially with precipitation-related things. Lesson: statistical models have problems if you don’t have data. So its perfectly good for coarse resolution work, and was a decade ahead of competitors that are only emerging now. We’ll add those to Wallace eventually.</p>

<p>Choose the <strong>10 arcmin</strong> data and press download. The first time you use <code>wallace</code> these data are slowly downloaded; after that you don’t have to wait. <strong>Don’t select finer resolution</strong> or you’ll be downloading while the rest of us are modeling.</p>

<p><img src="14_assets/Wallace3a.png" /> Notice that the log window notes that one point was discarded because environmental data were not available.</p>

</div>

<h1>Prep environmental data</h1>

<p>Now we need to choose the extent of the modeling domain. This jargon means that we have to define a sensible region to fit the model. Contrary to many publications, species ranges are not typically best modeling on domains defined by squares or political boundaries. Press some buttons on this screen to explore the options, but end up with something like what I show below. We’ll explore the consequences shortly. Press the <strong>Mask</strong> button to trim the environmental layers to this polygon and download the result.</p>

</div>

<h1>Partition occurrences</h1>

<p>In order to check whether you’ve built a decent model, you need some data to validate it. One solution is to partition your data into subsets (here 4) and build a models while witholding 1 subset at a time. Here we have 4 subsets, so we build 4 models, allowing us to get 4 independent measures of model performance. This is called k-fold cross-validation, and here k=4. There’s a whole literature on how to best make these subset; one option is to just do it randomly. A better option is to spatially stratify so that your model is forces to predict to regions that weren’t used for fitting. If it predicts well, you know you’ve got the general patterns right and have avoided overfitting to noise in the data. Below, I show some options for spatial stratification. Notice the 4 folds are now shown as 4 colors. Again, download the data.</p>

<p><br> Take a moment to scroll through the log window at the top of the screen and review all the steps you’ve taken so far.</p>

</div>

<h1>Model</h1>

<p>Finally, we’re going to make use of that results tab in the middle of the screen. Let’s build a Maxent model; this is a machine learning method that fits wiggly functions to patterns in the data. Its great for exploring complex patterns. If you construct it with a particular set of decisions it becomes very similar to a simple GLM. So it can conver a wide spectrum of complexity. If you want more details, ask Cory; he’s written a lot of papers on this so he may talk for while….</p>

<p>Below I’ve chosen some modeling options:</p>

<ul>

<li>Select the Maxent button</li>

<li>Select LQH features. These are the shapes that can be fit to the data:

<ul>

<li>L = Linear, e.g. temp + precip</li>

<li>Q = Quadratic, e.g. temp^2 + precip^2</li>

<li>P = Product, e.g. interaction terms of the form temp*precip</li>

<li>H = Hinge, e.g. piecewise linear functions. Taking all possible pairs of these between data points, you can build a very flexible function, similar to a GAM (generalized additive model).</li>

<li>T = Threshold, e.g. step functions between each pair of data points</li>

</ul></li>

</ul>

<ul>

<li>Select regularization multipliers between 1-3

<ul>

<li>regularization is a way to reduce model complexity. Higher values = smoother, less complex models. Its kind of like using AIC during model fitting to toss out certain predictors. Just ask for more details.</li>

</ul></li>

<li>RM Step Value = 1

<ul>

<li>how large of step to take between values in the slide bar above.</li>

</ul></li>

</ul>

<p>This takes about 2 min; you’re building 3 model types and 4 folds for each, and you’re using a load of features…</p>

<p><br> Notice that the first time I ran this, I got an error, which you may get too. You need to put the maxent software (maxent.jar) in the directory where <code>wallace</code> will look for it. This is a bug we’re working on. As the log window indicates, download the file and put it in the appropriate directory. Then click <strong>Run Models</strong> again. The result is number of statistics to compare the different models you just built. There should be 3: one for each of the regularization multipliers 1,2,3.</p>

</div>

<h1>Visualize</h1>

<p>Now let’s see which model performs best. Statistics based on the data witheld from fitting (<strong>test data</strong>) are the most reliable for determining the model’s generality and ability to transfer to other locations. The model with regularization mulitiple = 2 appears to be the best model based on avg.test.auc=0.804. Moving to the Visualize tab (module 7), you can plot the performance statistics across models. Explore the options a little.</p>

<p>To evaluate whether the model makes biological sense, we can look at <strong>response curves</strong> that define how each predictor (x-axis) relates to suitability (y-axis). The modal shape seems reasonable; there are places where the temp range is both too wide and too narrow for A. rubrum. The little jagged pieces intuitively seem like overfitting; why would a species, over an evolutionary time span, have an abrupt dip in response to temp range (as seen around 430). (Note that the units are degrees C x 100; worldclim serves the files this way to compress them.)</p>

<p>The <strong>bioclim</strong> predictors are a series of summaries of temp and precip that have been determined to have some biological significance. They’re all listed <a href="http://www.worldclim.org/bioclim">here</a>. bio7, seen below, is the annual temperature range.</p>

<p>We can also map the predictions. At first glance it looks like a decent model because the presence points correspond to regions of higher suitability.</p>

</div>

<h1>Project</h1>

<p>Next we can evaluate the models ability to project first to new locations (extending the domain) and new times (2070). First, extend the domain by drawing a polygon like the one below. Double-clicking on each of the 4 corners of this one draws the polygon. Next, press <code>Select</code> and then <code>Project</code> to build the new map.</p>

<p>Below are the predictions to a much larger region. Looks like northern Canada should get ready for a red maple invasion… So it appears we’ve fit a response that doesn’t generalize well to other regions. One way to determine how that happened is to look back at the response curves to determine which predictors are contributing to high suitability in northern Canada. This is best done with MESS plots in the maxent software package, and we haven’t included those in <code>wallace</code> just yet. So we’ll explore another way to avoid this poor prediction below.</p>

<p>As another example, of <code>wallace</code>’s features, we can also project to future climate scenarios. Select the options as below and forecast the future range of A. rubrum under an extreme climate change scenario. <br></p>

</div>

<h1>Extracting the code</h1>

<p>A major advantage of <code>wallace</code> compared to other GUI-based software is that you can extract all the code used to run the analysis. This allows you to recall what you did, share it, or modify it. The code is best extracted in <strong>R markdown</strong> format, which is a convienient format for combining R and text (and actually forms the basis of this website). Other formats are also available; for example Microsoft Word output mught be useful if you live in the ’90s.</p>

<p>To download the code, select Rmd and click Download. You may need to go to your R window and allow R to set up a cache to proceed. Extraction takes a minute; currently it has to rerun all the analyses we just did. We’ll fix this in an upcoming release to avoid the redundancy.</p>

<p>Now, you should have an .Rmd file that contains your complete analysis. Sometimes, if you make a bunch of mistakes while playing with the GUI, you might get an error when extracting the .Rmd. If that happens, you can download mine <a href="https://cmerow.github.io/RDataScience/14_assets/Outputs/wallace-session-2017-01-26.Rmd">here</a></p>

<p>Rmd files combine regular text with <strong>code chunks</strong>, shown by the red arrow below. Modules from <code>wallace</code> are indicated as headers denoted by <strong>###</strong>. For a quick reference to Rmd syntax, see <a href="https://www.rstudio.com/wp-content/uploads/2015/02/rmarkdown-cheatsheet.pdf">here</a></p>

<p>You might want to open a new R window and try running some of this code. Just note that if you close your <code>wallace</code> session you’ll lose your progress in the web browser (but your Rmd will be unaffected).</p>

<p>If you use RStudio, you can open this Rmd an click to <strong>knit</strong> to compile your workflow into a sharable html document. Just in case you encountered an error, you can see mine <a href="https://cmerow.github.io/RDataScience/14_assets/Outputs/wallace-session-2017-01-26.html">here</a></p>

<p>Note that you can change anything you like in this code to build upon your workflow. Future versions of <code>wallace</code> will enable you to upload such modified Rmds to <code>wallace</code> to fill in all the options you chose and pick up where you left off in a previous analysis in the GUI.</p>

<p>At the moment we don’t have anything built into <code>wallace</code> for post-processing, so you can use R directly to build from the code created above.</p>

</div>

<h1>Improving the model</h1>

<p>Let’s revisit that crappy prediction into northern Canada. This issue derived from a poor choice of modeling domain and an overfit model. Try rerunning the analysis by extending the domain to include many locations where the species does not occur (see below) and using a simpler model that includes only linear and quadtratic features.</p>

<p>Here are the improved predictions that avoid prediction in northern Canada.</p>

</div>

(function(i,s,o,g,r,a,m){i['GoogleAnalyticsObject']=r;i[r]=i[r]||function(){

(i[r].q=i[r].q||[]).push(arguments)},i[r].l=1*new Date();a=s.createElement(o),

m=s.getElementsByTagName(o)[0];a.async=1;a.src=g;m.parentNode.insertBefore(a,m)

})(window,document,'script','https://www.google-analytics.com/analytics.js','ga');

ga('create', 'UA-2684666-2', 'auto');

ga('send', 'pageview');

</script>

<!-- disqus

/* * * CONFIGURATION VARIABLES: EDIT BEFORE PASTING INTO YOUR WEBPAGE * * */

var disqus_shortname = 'rdataanalysis'; // required: replace example with your forum shortname

var disqus_identifier = "{{ page.url }}";

/* * * DON'T EDIT BELOW THIS LINE * * */

(function() {

var dsq = document.createElement('script'); dsq.type = 'text/javascript'; dsq.async = true;

dsq.src = '//' + disqus_shortname + '.disqus.com/embed.js';

(document.getElementsByTagName('head')[0] || document.getElementsByTagName('body')[0]).appendChild(dsq );

})();

</script>

comments powered by Disqus.</a>

</noscript>

-->

<br/>

</div>

</div>

</footer>

</div>

// add bootstrap table styles to pandoc tables

function bootstrapStylePandocTables() {

$('tr.header').parent('thead').parent('table').addClass('table table-condensed');

}

$(document).ready(function () {

bootstrapStylePandocTables();

});

</script>

</body>

</html>

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

FilesExpand file tree

14_WallaceSDM.html

Latest commit

History

14_WallaceSDM.html

File metadata and controls