Category Archives: Seminar

Jordan’s bigdata talk in Math/Stats series

Really excel­lent talk by Pro­fes­sor Michael Jor­dan from the Berke­ley Sta­tis­tics depart­ment vis­it­ing us at Davis yes­ter­day. Haven’t seen a talk cover tech­ni­cal con­tent with remark­able clar­ity for the algo­rithms and insights involved, while also cov­er­ing such a breadth of mate­r­ial. My notes didn’t really keep up, but tran­scrib­ing some of my scrib­bling into elec­tronic

Alex Pfaff Seminar, meeting

Notes from Alex Pfaff talk at UTK (dur­ing my visit for PDG Con­trol meet­ing). aver­age impact vs dif­fer­en­tial impact — must com­pare against what would have hap­pened to the area if unpro­tected (at cur­rent time/short-timescale, or in future?) Prox­im­ity is not a good proxy of sim­i­lar­ity (i.e. bc we pro­tect steep slopes but farm the

Visuals for Communicating Uncertainty

Vis­it­ing the ques­tion of com­mu­ni­cat­ing uncer­tainty in Marissa’s lab meet­ing, focus­ing on this excel­lent Sci­ence arti­cle . Noam and Jamie are lead­ing the dis­cus­sion, and sug­gested we pick a few favorite exam­ples. None on my list really address the chal­lenge of com­mu­ni­cat­ing uncer­tainty directly, though the first one prob­a­bly comes clos­est. I think they do

Resilience Seminar course plan

Resilience Monte Carlo Sem­i­nar, 11am Tues­days. Val Eviner. Jan 17 Resilience of coral reefs/mangrove systems, Alexander Gaos Jan 24 Resilience of degraded vs. intact systems, Erica Case and Alex Web­ster Jan 31 Novel ecosystems, Ania Truszczyn­ski and Mark Noyes Feb 7 Land­scape configuration, Julia Moore Feb 14 Assisted Migration, Anna O’Brien Feb 21 Design­ing and man­ag­ing for resilience, Sarah McCul­lough and Car­o­line Wright Feb 28 Thresholds/ uncertainty, Matt Meis­ner

Steve Pacala — Storer lecture

Excel­lent visit from a for­mer advi­sor, Steve Pacala, at Davis for a Storer lec­ture.  Sem­i­nar atmos­pheric CO2 is roughly half CO2 pro­duc­tion, rest is being mit­i­gated by the nat­ural car­bon sink Car­bon sink is half ocean, half ter­res­trial A grim future: 450ppm not fea­si­ble 500ppm best case 550ppm is likely even if we are largely suc­cess­ful Even

Thursday

For­age Fish dynam­ics For­age fish con­sump­tion: who con­sumes, how impor­tant is it. e.g. Africa imports more quan­tity of fish, but of low value. net exports based on value. food secu­rity / under­nour­ished regions for­age fish sub­sti­tutable? Study in Asia: Pretty inelas­tic. income elas­tic­ity vs wealth elas­tic­ity (not a real dif­fer­ence. but pat­terns of con­sump­tion of a

Loo Botsford, CPB Seminar

Effects of Fish­ing on the Sen­si­tiv­ity of Fish­eries to Cli­mate Vari­abil­ity” Trun­ca­tion of age struc­ture. juve­na­tion (younger pop­u­la­tion), mater­nal effects (first-year breed­ers are poor), genetic selec­tion, abil­ity to track envi­ron­ment (vs smooth­ing — aver­ag­ing over longer times). Out­line: Empir­i­cal evi­dence for increased vari­abil­ity, model expla­na­tions Story begins with Hsieh, et al 2006.  Long time series,

EM Algorithm

Yaniv ran us through our sec­ond ses­sion on EM algo­rithms.  We imple­mented a sim­ple case described in this tuto­r­ial. Code doesn’t reflect the abstrac­tion of the algo­rithm into a proper Expec­ta­tion step and Max­i­miza­tion step.  We attempted this gen­er­al­iza­tion: Missed some­thing in fram­ing this cor­rectly, since the max­i­miza­tion step includes a func­tion that doesn’t depend

Matt Potts: Conserving Diversity in Tropical Landscapes

My notes from Matt’s sem­i­nar: The­ory and field work address­ing get­ting beyond binary deci­sion mak­ing.  Malaysia.  Trop­i­cal for­est becom­ing oil palm, rice, rub­ber, tea. Reserve Selec­tion vs Land­scape Reserve Design Opti­mal­ity of Spe­cial­ized vs Uni­form For­est Man­age­ment Uni­form (sus­tain­able for­est man­age­ment) vs very pro­tected areas, ignore what’s done out­side.  More inten­sive dis­tur­bance. An impor­tant eco­log­i­cal non­lin­ear­ity:

Algorithms Discussion Group: MCMC

Imple­mented a basic MCMC rou­tine in our lit­tle algo­rithms dis­cus­sion group today, works quite nicely once you remem­ber to use dif­fer­ences of log prob­a­bil­i­ties instead of ratios.  Code and results below.