Managing the Temporal Geography of Futures Michael Flaxman, MIT.

24
Managing the Temporal Geography of Futures Michael Flaxman, MIT

Transcript of Managing the Temporal Geography of Futures Michael Flaxman, MIT.

Page 1: Managing the Temporal Geography of Futures Michael Flaxman, MIT.

Managing the Temporal Geography of Futures

Michael Flaxman, MIT

Page 2: Managing the Temporal Geography of Futures Michael Flaxman, MIT.

Motivation

Page 3: Managing the Temporal Geography of Futures Michael Flaxman, MIT.

Overview

Alternative Futures Methodology

Current Scenario & Impact Model Data Management

Four Problems

Two (Partial) Solutions

Page 4: Managing the Temporal Geography of Futures Michael Flaxman, MIT.

Problem 1: Too Many Futures!

Page 5: Managing the Temporal Geography of Futures Michael Flaxman, MIT.
Page 6: Managing the Temporal Geography of Futures Michael Flaxman, MIT.

Problem 2:Logical Dependencies Are Important

Correct Interpretation often depends on understanding underlying assumptions

Large Update Problems

– Scenarios dependencies propagate

– If dependencies are not tracked, danger of false attribution

Page 7: Managing the Temporal Geography of Futures Michael Flaxman, MIT.

Problem 2:Logical Dependencies Example

– Scenario 1 (S1) Impact of S1 on Hydrology Impact of Hydrology under S1 on Species Habitat

– Scenario 2 (S2) Impact of S2 on Hydrology Impact of Hydrology under S2 on Species Habitat

If Scenarios Change…

– Dependencies propagate i.e.above must recompute hydrology twice & habitat twice

– If dependencies are not tracked, danger of false attribution i.e. Species Habitat map not correctly updated to S1v17

Page 8: Managing the Temporal Geography of Futures Michael Flaxman, MIT.

Problem 3:Sharing

Creating a single isolated system to manage spatiotemporal data is hard

Creating a networked system is much harder still!– Must track dependencies *between* systems– Must deal with broken connections, latencies, and

time changes

Page 9: Managing the Temporal Geography of Futures Michael Flaxman, MIT.

Problem 3:Sharing Example

Parties Involved– Hydrologist at USGS (Ft. Lauderdale) – Land Use Modeler at UFL (Gainsville)– Habitat Specialists at FWS (Vero Beach & 2 Refuges)– Vegetation Specialist at Everglades National Park

Action– Land Use Modeler receives updated demographic estimate, updates land

cover model

One scenario change requires sequential notification to 5 distributed parties + manager

Page 10: Managing the Temporal Geography of Futures Michael Flaxman, MIT.

Problem 4:Metaphors & Mechanisms Are Weak

Available object “metaphors” and mechanisms– Files (on disk)

Bundled by space, discrete for each time ‘slice’ NetCDF – multidimensional file format

– Supported by climate models, some GIS– Not well supported outside of science

– Layers (in GIS) User re-orderable with occlusion (for better and worse) Independent visibility toggle

– Hierarchical Folders (both) Allows development of hierarchical file or layer representations of time

Other Metaphors & Concepts– Time Line”

Understandable interface, but not sharable implementation– Dependency Diagram

Again, well understood, but each implementation separate

Page 11: Managing the Temporal Geography of Futures Michael Flaxman, MIT.

MIT Prototypes

ScenarioCMS– A content management system for spatial scenarios organized as

“time slices”– Provides “ScenarioXML” language to document assumptions and

dependencies– Status: working prototype (Telluride), Phase 2 (BajaEcoInfo)

EverView2– Extension of ScenarioCMS for the Everglades– Visualizes & manages assumptions, choices and dependencies– Organizes “stories” within Scenarios

Stories are complex sub-scenarios with temporal sequencing – Status: early schematic

Page 12: Managing the Temporal Geography of Futures Michael Flaxman, MIT.
Page 13: Managing the Temporal Geography of Futures Michael Flaxman, MIT.

ScenarioCMS: Scenarios & Constraints

Page 14: Managing the Temporal Geography of Futures Michael Flaxman, MIT.

ScenarioXML

Vendor-neutral, software-neutral Organizes scenarios logically

– Like HTML, separates presentation from data– Metadata for scenario (machine & human

readable)

Page 15: Managing the Temporal Geography of Futures Michael Flaxman, MIT.

ScenarioCMS: Simple Time Slider

Page 16: Managing the Temporal Geography of Futures Michael Flaxman, MIT.

ScenarioCMS: Dynamic Legend

Page 17: Managing the Temporal Geography of Futures Michael Flaxman, MIT.

Florida EvergladesDis-integrated Management Systems

Water Manager’s View: Pipes Only

Refuge Manager’s View: Habitat & Species Observation Only

Page 18: Managing the Temporal Geography of Futures Michael Flaxman, MIT.
Page 19: Managing the Temporal Geography of Futures Michael Flaxman, MIT.

Scenario 1Conditions:Wet SeasonHurricane IV approachingLoxahatchee NWR and Miami-Dade at High Flood-Risk

Management Options:

A.

Miami-Dade Impact:Flood-risk reduced

Loxahatchee NWR/Everglades Impact:Flood-risk reduced

Caloosahatchee Estuary and St. Lucie Estuary Impacts:Water quality decreasesLow O2 levelsFish KillInundate Sea Grass

Urban Flood-Risk LevelsL M H

Conservation Areas Flood-Risk Levels L M H

Lake Okechobee Water Level 14.5’ 16.5’ 17.5’

release

Lock

timeline

Flood-risk high

Release to C-44 and C-43

Water reaches first locks, Port Mayaca, Moore Haven

Water reaches St. Lucie Lock

Water reaches Franklin Lock

Urban flood-risk reduced; Estuarine Impacts

A.

Release water toC-43 and C-44 St. Lucie Canal; Caloosahatchee Canal

B.

Release water toL-8 and L-10 STA 1-W and 1-E; WCA 1 (Lox. NWR)

Decision Impacts:

Preview Preview

Page 20: Managing the Temporal Geography of Futures Michael Flaxman, MIT.

Scenario 1Conditions:Wet SeasonHurricane IV approachingLoxahatchee NWR and Miami-Dade at High Flood-Risk

Management Options:

Urban Flood-Risk LevelsL M H

Conservation Areas Flood-Risk Levels L M H

Lake Okechobee Water Level 14.5’ 16.5’ 17.5’

release

Lock

Release to L-8, L-10

A.

Release water toC-43 and C-44 St. Lucie Canal; Caloosahatchee Canal

B.

Release water toL-8 and L-10 STA 1-W and 1-E; WCA 1 (Lox. NWR)

Decision Impacts:

Preview Preview

B.

Miami-Dade Impact:Flood-risk reduced

Loxahatchee NWR/Everglades Impact:Apple snail population andwaterfowl nesting inundated

High Flood-Risk

Water reaches STA 1-W and 1-E

Water released to WCA 1, (Lox)

Water reaches WCA 2, WCA 3

Water reaches Everglades

Apple Snail pop. disturbed

Page 21: Managing the Temporal Geography of Futures Michael Flaxman, MIT.

Scenario 2Conditions:Drought SeasonSFWMD Phase IV Drought “Critical”Loxahatchee NWR and Everglades National Park need water

SFWMD Drought Protocol: Water Restrictions Stages I II III IV

Flow RateNone Minimum Adequate

Severe Drought

timeline

Release water to L-8, L-10

Water reaches STA 1-W and 1-E

Water released to WCA 1, (Lox)

Water reaches WCA 2, WCA 3

Water reaches Everglades

Management Options:

A.

Miami-Dade Impact:Water restrictions remain Phase IV

Loxahatchee NWR/Everglades Impact:Minimum flows received, still dry

A.

Release min. flows to L-8, L-10, STA 1-W and 1-E, WCA 1 Loxahatchee NWR; Everglades

B.

Release water toL-15 and L-18 Miami-Dade and West Palm Beach

Decision Impacts:

Preview Preview

Page 22: Managing the Temporal Geography of Futures Michael Flaxman, MIT.

Scenario 2Conditions:Drought SeasonSFWMD Phase IV Drought “Critical”Loxahatchee NWR and Everglades National Park need water

SFWMD Drought Protocol: Water Restrictions Stages I II III IV

Flow RateNone Minimum Adequate

Drought severe

Release water to L-15, L-18

Water reaches Miami-Dade County line

Water restrictions reduced Phase III

Management Options:

A.

Release min. flows to L-8, L-10, STA 1-W and 1-E, WCA 1 Loxahatchee NWR; Everglades

B.

Release water toL-15 and L-18 Miami-Dade and West Palm Beach

Decision Impacts:

Preview Preview

B.

Miami-Dade Impact:Water restrictions reduced to Phase III

Loxahatchee NWR/Everglades Impact:Apple snail population failsMandatory minimum flows not met

Min. flows to Everglades not met

Page 23: Managing the Temporal Geography of Futures Michael Flaxman, MIT.

Conclusions

Spatiotemporal Scenario Management Needed– Typical scenario study generates 100+ layers– Logical dependencies important to preserve

Sharing is Nice– Single-application solutions inadequate– Many raw data ‘standards’ to pick from– Higher-level aggregations desirable

Page 24: Managing the Temporal Geography of Futures Michael Flaxman, MIT.

Future Work

Telluride Prototype– Go live this summer– Kept simple

Time slices only Interface exposes dependencies as hierarchies Back-end ScenarioXML drives interface

Everview2– To be developed next academic year