Responding to evolving threats using innovative tools, technologies and datasets - Kathy Willis

Professor Kathy Willis,

Biodiversity Institute, University of Oxford

Responding to evolving threats using innovative tools, technologies and

datasets

Evolving threats

Increasing demand on land

• Global population most likely to peak ~9B

2000 2050 2100

12B

8B

4B

Population projection (Lutz & Samir 2010)

20%60%

95%

• People will be richer and demand higher quality diet

China

India

Africa

1970 1980 1990 2000

Live

sto

ck c

on

sum

pti

on

Developed nations

Livestock consumption (FAO 2009)

Hwange National Park, Zimbabwe

Protected (12%)

Not protected (88%)

Biodiversity declines

Stokard 2010. Despite progress, biodiversity declines. Science. 329: 1272-1273.

Is all lost for biodiversity?

Convention of Biological Diversity targets (2011)

Target 5By 2020, the rate of loss of all natural habitats, including forests, is at least halved and where feasible brought close to zero, and degradation and fragmentation is significantly reduced.

Target 14By 2020, ecosystems that provide essential services, including services related to water, and contribute to health, livelihoods and well-being, are restored and safeguarded

Target 15By 2020, ecosystem resilience and the contribution of biodiversity to carbon stocks has been enhanced, through conservation and restoration

What innovative tools, technologies and datasets do we need to:

1. Identify and reduce loss of natural habitats?

2. Enhance ecosystem resilience?

3. Conserve ecosystems that provide essential services related to human well-being?

Talk outline

Case study:

Determining the ecological value of landscapes beyond protected areas

What tools are available to Identify and reduce loss of natural habitats?

Willis, K.J. et al., 2012, Biological Conservation, 147, 3-12

?

???

?

“ Where can we damage? ”

Points arising from workshops with Statoil

1. Need a tool that provides estimation of ecological value of land outside of protected areas

2. To produce landscape information at a spatial scale less than 500m;

3. Use existing available web-based databases;

4. Produce simplified displays – preferably maps;

5. Simple user input;

6. Able to assess any region in world;

Global vegetation cover at 300m pixel size resolution

(GLOBCOVER (Bicheron et al. 2009)

What is the finest spatial resolution (pixel size)?

What data are needed to provide an spatial distribution of ecological value on a landscape?

Need data on:

1. Key ecological properties of the landscape (e.g. biodiversity, threatened species)

2. Key features for supporting ecosystem functions (e.g. connectivity (migration routes, wetlands) habitat integrity, resilience)

3. Their spatial configuration on the landscape.

Biodiversity data

• For most regions in the world will rarely be enough detailed species data to obtain clear picture

• Necessary to model predictive diversity across landscape (generalised dissimilarity modelling)

• Can then use combination of point species occurrences + environmental variables to predict diversity (spatial heterogeneity) across landscape

Global Biodiversity (GBIF):

Data Portal (http://data.gbif.org) that provides access to more than330 million records of species occurrence worldwide

Biodiversity species occurrence data

GBIF network Data Coverage

Last updated: 2010

>330 million occurrence records from >8,500 datasets from

>360 publishers and spanning a wide range of geospatial,

temporal and taxonomic coverages being shared through

distributed network

Data sources for environmental variables

Beta-diversity for Canadian site measured using Generalised Dissimilarity modelling

Value provided for every 300m pixel

Threatened species data sources

• 2010 IUCN Red List of Threatened Species

• Assessments for ~56,000 species, of which about 28,000 have spatial data.

• Consider all categories in concession area except ‘least concerned’ and ‘extinct’

• More threatened species in pixel, higher its value

Threatened species distribution in Canadian concession area

Fragmentation data

• Spatial continuity of natural vegetation based on the size (ha) of each continuous patch

• Computer programme FRAGSTATS (McGarigal and Marks, 1995) defines individual patches and calculates patch size

• Apply FRAGSTATS to vegetation cover

• Greater the patch size, higher the ecological value

Fragmentation map Canadian concession areas

Global Register of Migratory Species

• Contains list of 2,880 migratory vertebrate species in digital format

• Also their threat status according to the International Red List 2000,

• Digital maps for 545 species

• Sum the number of migratory ranges occurring in each per pixel

www.groms.de

Connectivity (1) Migratory routes

Connectivity (2) – Migration processes

• Prioritize pixels that support migratory processes:

– Rivers, wetlands and lakes (at 300m resolution)

– Adjacent pixels to rivers (so as to allow migratory corridors)

Data source: HYDROSHEDS (USGS), Global lakes & wetlands database (WWF)

Water bodies and drainage networks for Canadian concession area

Global Lakes and Wetlands Database,

HYDROSHEDS; 30m pixel resolution

Resilience

– Areas of landscape that are particularly resistant to climate change/disturbance

– Areas of landscape that are able to recover from disturbance quicker than others

Resilience: measured through ability of vegetation to

maintain relatively high levels of productivity despite low levels of rainfall

Rainfall (mm) in driest month

Annualized NPP

Vegetation Type

Scoring Rule:

1, if highest quartile of productivity & lowest quartile of rainfall

0.5, if highest quartile of productivity & next lowest quartile of rainfall

0, otherwise

Assessed per vegetation type

Resilience, Canadian concession area

To summarise

Factors and data sources used in LEFT

Willis, K.J. et al., 2012, Biological Conservation, 147, 3-12

Final index

Final index: Local ecological footprint valuation

Species richness

Vulnerability

Connectivity

Fragmentation

Resilience

+

+

+

+

Automation

How accurate in comparison to field data?

Cusuco, Honduras

• Montane tropical moist forest• Surveyed 2004-2010• Extensive datasets e.g >50,000 records of terrestrial

vertebrates in database

Cusuco national park, Honduras

Can LEFT correctly identify which globally threatened terrestrial vertebrates are present in a study site?

Threatened birds

Threatened mammals

Threatened reptiles

Threatened amphibians

All threatened terrestrial vertebrates

Field data Web data

LEFT correct

LEFT omission error(detected byfieldwork, but missed by LEFT)

LEFT commission error(not detected by fieldwork, yet included in LEFT)

26 1753

1

0

1

4

2

1

19

10

6

0

1

Cusuco – normalised number of threatened speciesCan LEFT correctly identify which locations in a study site are most important for threatened species?

Difference mapWhite = agreement.

Red = LEFT predicts relatively more threatened species than field data (commission error)Blue = LEFT predicts relatively fewer threatened species than field data (omission error)

Cusuco, Aves Beta-diversity based on GBIF data

n = 405 (67 sites)

Cusuco – beta-diversity using GBIF

Beta-diversity calculated using species occurrence data (birds) in GBIF

Cusuco, Aves Beta-diversity based on field data

n = 3297 (116 sites)

Beta-diversity calculated using species occurrence data (birds) from field data

Cusuco – beta-diversity using field data

Summary

• Tool will work anywhere in the world at local-scale resolution (~ 300m pixel)

• Provides report, maps, files on all values used to calculated ecological value in ~10 minutes

• Preliminary studies to compare tool output with high resolution field data indicates that general ecological trends well represented

• Consistent and quick approach for obtaining most up-to-date biodiversity information

What innovative tools, technologies and datasets do we need to:

1. Identify and reduce loss of natural habitats?

2. Enhance ecosystem resilience?

3. Conserve ecosystems that provide essential services related to human well-being?

Talk outline

Target 15

“By 2020, ecosystem resilience and the contribution of biodiversity to carbon stocks has been enhanced, through conservation and restoration

”Resilience is the capacity of a system to absorb disturbance and still retain its basic function and structure” (Holling, 1973)

Alternative definition:

‘Resilience is speed of return to an equilibrium state following a perturbation from that state’ (Nystrom et al. 2000)

What is scientific information is needed to determine and plan for resilient landscapes?

1. How resilient is the landscape to environmental perturbations?

– e.g. climate change/land-use change

2. What is the spatial arrangement of resilient ecosystems across the landscape?

How resilient is the landscape to environmental disturbance?

Recovery rates of tropical forests to disturbance events

L. Cole, S. Bhagwat & K.J Willis, in prep

• Data from 40 individual fossil sedimentary pollen sequences• Contain records of vegetation dynamics spanning last 10,000 years• Document a total of 140 disturbance events across 3 continents

Disturbance source

Disturbance type Proxy

NATURAL Climate (C)

Precipitation (CP)Sea-level rise (CS)

Oxygen isotopes, fire (low levels, not linked to human presence), magnetic susceptibility, lithologyRainfall, monsoon strength variation, climate drying (CD)Sea level

Large infrequent (LI) Hurricane (LI-H), landslide (LI-L), fire (LI-F), volcano (volcanic ash) (LI-V)

HUMAN Burning (B) Micro- & macro-charcoal

Forest clearing (FC)

Temporary, predominantly resulting from shifting cultivation (SC), or more permanent, generally selective clearing, or not described (FC) signified by e.g. fruit trees, Poaceae, & disturbance indicators/secondary forest taxa, e.g. Arenga and Macaranga, or magnetic susceptibility

Agriculture (Ag) Agricultural indicators, e.g. fruit trees - Ficus, crops -Poaceae

Unclear U Disturbance indicators but type undefined

Classification of disturbance type

Metric Description Calculation

Recovery Rate (RR) Rate of forest recovery relative to degree of disturbance-induced percentage change

RR = ((Fmax - Fmin)/(Fpre - Fmin))*100/ Trec

Forest % decline (FD) Forest percentage decline relative to baseline forest cover percentage

Rel.D = ((Fpre - Fmin)/ Fpre)*100

Resilience (RS) Change in RR through time (RR1 represents oldest sample in study)

(RS) = RR2 – RR1

Calculation of resilience

How quickly have tropical forests recovered from disturbances in the past?

L. Cole, S. Bhagwat & K.J Willis, in prep

Does geographical location affect recovery rates?

Fastest recovery rates in Central America

Slowest recovery rates in S. America

Type of disturbance also indicated significant impact on recovery rates

Forest clearance through burning etc. resulted in slowest recorded recovery rates (and greatest variation)

L. Cole, S. Bhagwat & K.J Willis, in prep

• Using long-term datasets it is possible to start to determine relative recovery rates

• But this still doesn’t give a clear indication of which areas across a landscape are more resilient to climatic perturbations

• To do this we need to examine shorter-term/finer resolution datasets

Resilience: measured through ability of vegetation to

maintain relatively high levels of productivity despite low levels of rainfall

Rainfall (mm) in driest month

Annualized NPP

Vegetation Type

Scoring Rule:

1, if highest quartile of productivity & lowest quartile of rainfall

0.5, if highest quartile of productivity & next lowest quartile of rainfall

0, otherwise

Assessed per vegetation type

K.J. Willis et al., 2012 Biological Conservation, in press

• 12 year monthly time-slice of NDVI (MODIS) (144 layers in total)

• 5km resolution• Masked for sea-areas/

large terrestrial water bodies

• Red = high, green = low

Devising A Global Map of Ecological Resilience: Step 1- NDVI (photosynthetic ‘health’)

• Data are detrended for seasonality and transformed to Z-scores in each pixel.

• Provides an estimate of amount of variability away from the mean over the 10 years.

Red = high; Green = low

A.W.R. Seddon, P. Long and K.J. Willis in prep

• Variance of these Z scores provides a global map of the variance in productivity for each pixel

• Red = high variance, green = low variance

Devising A Global Map of Ecological Resilience: Step 1- NDVI

A.W.R. Seddon, P. Long and K.J. Willis in prep

Towards A Global Map of Ecological Resilience: Step 2- Temperature variance

• Converted to z scores to provide a global map of the variance in temperature for each pixel at 5 km resolution

• Red = high variance, green = low variance

•12 year monthly time-slices of mean monthly surface temperature (MOD-7 profiles)•5km resolution

Towards a Global Map of Ecological Resilience: Step 3

Sensitivity (γ) = Temporal Variance in Productivity

Temporal Variance in Climate

Resilience = 1/γ

(of NDVI (productivity) to climate variability over a 10 year period)

Global 12 year Resilience of NDVI to Climate Variability

• red = low and green = high

What innovative tools, technologies and datasets do we need to:

1. Identify and reduce loss of natural habitats?

2. Enhance and identify ecosystem resilience?

3. Conserve ecosystems that provide essential services related to human well-being?

Talk outline

Target 14

“By 2020, ecosystems that provide essential services, including services related to water, and contribute to health, livelihoods and well-being, are restored and safeguarded.”

R.S. de Groot et al. 2010 EcologicalComplexity 7 (2010) 260–272

What knowledge do we need?

Current landscape planning, management and decision making tools

InVEST(Integrated Valuation of Ecosystem Services and Tradeoffs)

ESValue

ARIES (ARtificial Intelligence for Ecosystem Services)

ARIES (ARtificial Intelligence for Ecosystem Services)

End-user needs to work with the ARIES team; developed for specific area; one site

output requires 200-300 hours of Senior GIS technician time

InVEST (Integrated Valuation of Ecosystem Services and Tradeoffs)

Time varies depending on the site and the technician’s expertise; one site output

requires 160-280 hours of Senior GIS technician time

ESValue

~ 200 hours for one site; requires GIS expertise, expert knowledge of ecological relationships plus data from stakeholders

EcoAIM (Ecological Asset Inventory and Management)

>25 hours; involves reviewing, downloading, converting and uploading data by stakeholder

Current Ecosystem Service Tools: (http://www.bsr.org/reports/BSR_ESTM_WG_Comp_ES_Tools_Synthesis3.pdf)

"a gap in biodiversity market infrastructure that persists is lack of landscape-scale ecological monitoring. While site-level ecological monitoring is not uncommon, the data is not easily available, much less complied in a comprehensive way".

Madsen, B., Caroll, N., Kandy, D., Bennett, G (2011) Update: State of Biodiversity Markets. Washington, DC: Forest Trends, 2011. http://www. ecosystemmarketplace.com/reports/2011_update_sbdm.

What data do we need to provide a tool to quickly and remotely determine ecosystem service provision?

landowner

What information is required to map pollination services?

Land cover GBIF species occurrence data

Environmental co-variables

DISTRIBUTIONS OF POLLINATORS

Crops

Pollination DEPENDENT

CROP

Availability of pollinators

Nesting habitat for P.

Pollination service delivery

Pollinator foraging distance

P.= pollinator

Final pollination service delivery

+

+

x

Distribution Model

Landscape featurese.g. nesting habitat

Landscape containing pollinators

Crop dependency

Foraging distance

Steps to follow

Preliminary pollination service delivery for Tenerife

Tenerife actual pollination service delivery

Tenerife tree cropsTenerife foraging

distanceTenerife nesting habitat

0.5 km

More service delivered

Less service delivered

More service delivered

Less service delivered

Important areas for pollination services for tree crops

Nogues, Long & Willis, in prep

• Large scientific biodiversity resource becoming available through databases, modelling and ecological knowledge

• Creation of tools to link this information together requires highly interdisciplinary research community

• … but must also have good knowledge of requirements of end-user

• The challenge is to bring together these tools, technologies and datasets but in a framework that is relevant to both science and stakeholder communities

• This requires pragmatism and a different approach to funding such work…

Responding to evolving threats using innovative tools, technologies and datasets