TEMPOROSPATIAL ANALYSISOF AGRICULTURAL SYSTEMSAT REGIONAL WATERSHED LEVEL

30 YEARS OF DATA TO CHARACTERIZE THE MEUSE & MOSELLE WATERSHEDS (FRA)

Session 5: FARM SYSTEM DESIGN

•• Temporospat ia l ana lys i s o f agr i cu l tura l systems • RIZZO et al. •• ESA Session 5 • August 26th 2014 •• bit.ly/ESA_2014

Aim & background

Davide

Guillaume

Marc

SAD-ASTER Mirecourt

1

RIZZOGODFROY

BENOÎT

Water protection issues urge agronomy to deal with large scale impacts on environmental resources

Characterizing agricultural systems

major land use sequences & related fertilization practices

map of potential pressure on water quality

Agricultural dynamics in 23 primary watersheds

(north-eastern France~24 000 Km2)

Rizzo, Godfroy, Benoît 2013 DynAMM’Eau project report

•• Temporospat ia l ana lys i s o f agr i cu l tura l systems • RIZZO et al. •• ESA Session 5 • August 26th 2014 •• bit.ly/ESA_2014

A landscape agronomy perspective…2

…to address interactions between farming practices and natural resource (e.g., water)

management through spatially-explicit modeling at the landscape level

wheatbarleyrapeseedmaizegrassland

% of land cover (on total study area) land use changes?

time and spatial regional scalesrelevant for watershed managers

30%

50%

23%

47%

MATERIAL & METHODS

•• Temporospat ia l ana lys i s o f agr i cu l tura l systems • RIZZO et al. •• ESA Session 5 • August 26th 2014 •• bit.ly/ESA_2014

Method #1: a time dominant approach3

Space dominant

sequence of (land cover)

images

Software for LUCC analysis generally implement unsupervised clustering on spatial

entities based on space-dominant attributes

Data mining approaches involving clustering of sequences allowed knowledge extraction to get a landscape description in terms of land-use patterns

Time dominant

image of (land cover)

sequences

logics of human driven activities

Mari et al. 2006 Temporal and spatial data mining

•• Temporospat ia l ana lys i s o f agr i cu l tura l systems • RIZZO et al. •• ESA Session 5 • August 26th 2014 •• bit.ly/ESA_2014

4

Method #2: clustering land cover return time

81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 08 10

01 02 03 15

possible return time every© x x x xx x x x ©© x x x ©x © x x xx x x © x© x x © xx © x x ©x x © x x© x © x ©x © x © x

presence

5 years wheatbarley & other cerealsrapeseedmaizeother cropsset-asidegrasslands (3 types)artificial & undet.semi-natural

04 05 06 07 08 09 10 11 12 13 14

29 possible combinations over 15 5-year periods

of major land covers

23 583 points

TerUti 1981-1990

23 586 points

TerUti 1992-2003

11 588 points

TerUti-Lucas 2006-10

absence

4 years

3 years

2 years

5-year sliding

windows

TerUti is a yearlyland use stratified

sampling on a regular grid

Hierarchical clustering on Principal Components

FactoMiner, Lê et al., 2008

cf. Slak & Lee, 2003

data-mining

multivariate analysis

•• Temporospat ia l ana lys i s o f agr i cu l tura l systems • RIZZO et al. •• ESA Session 5 • August 26th 2014 •• bit.ly/ESA_2014

5

Method #3: evaluation of nitrogen pressure

plot balance per year

Using the survey « Pratiques culturales » (PK)

Balance Azotée Spatialisée des Systèmes de Culture de l’Exploitation

BASCULE Benoît, 1992

INPUT OUTPUT

fertilization(mineral + organic)* yield x N content*

* Integrated with literature

Plot level information on: - fertilization & yield (crop of the year)- preceding crop

TerUti subsample • • 1994, 2001 & 2006

plot BASCULEaverage (median) of plot

balance over multiple years

area BASCULESum of the (positive) plot

BASCULE over the target area *

NB negative values do not subtract

*farm, farm system, watershed, cluster, etc.

1

2

3

time

space

RESULTS

•• Temporospat ia l ana lys i s o f agr i cu l tura l systems • RIZZO et al. •• ESA Session 5 • August 26th 2014 •• bit.ly/ESA_2014

Results #1: major land use dynamics6

barley

seq.

4 crop

seq.

crops & urban

semi-natural

rapeseed& wheat

mixedfarming

crops & urban

semi-natural

80’s 90’s 00’s

1981-1990 1992-2003 2006-2010

6 clusters4 regions2 periods

•• Temporospat ia l ana lys i s o f agr i cu l tura l systems • RIZZO et al. •• ESA Session 5 • August 26th 2014 •• bit.ly/ESA_2014

axe 1 (56% var.)

no crops

Results #1bis: cluster dynamic per watershed

6

max crops

2n

dax

is (

31

% v

ar.)

no pastures

perm. pastures

1st axis (56% var.)

•• Temporospat ia l ana lys i s o f agr i cu l tura l systems • RIZZO et al. •• ESA Session 5 • August 26th 2014 •• bit.ly/ESA_2014

Results #2: evaluation of nitrogen pressure7

BASCULEglobal average

other land uses

Cluster BASCULE example of 2006 data

30 ] 38 [ 51

30 ] 39 [ 53

39 ] 50 [ 64

44 ] 63 [ 100

Kg N / Ha / year

min ] median [ max

threshold for ground water pollution (≈ 50 mg NO3- / liter)

CONCLUSIONS

•• Temporospat ia l ana lys i s o f agr i cu l tura l systems • RIZZO et al. •• ESA Session 5 • August 26th 2014 •• bit.ly/ESA_2014

Conclusions8

synthetic representation of long time series over a wide area

spatial explicit evaluation that can be coupled with vulnerability maps

scalable on available input data(e.g., within the watershed)

(mitigating) effects of contextual land uses(e.g., forest)

inclusion of other practices(e.g., catch crops, pesticides)

influence of specific climate conditionsand other potential practice drivers

using custom group of pixels to better understand the

land system architecture

Leverages… …and potentials

to match action spaces of relevant managers and

coordinating landscape level decision-making

Turner II et al., 2013 Land system architecture

Future applications Reinterpreting watershed hierarchy (Strahler)

Rizzo et al., 2013 Farming systems designing landscapes

Lazrak et al., in prep. for SAGEO congress

Thank you for the attention

ridavide@gmail.com

bit.ly/ESA_2014

@pievarino

REF

Davide RIZZO

•• Temporospat ia l ana lys i s o f agr i cu l tura l systems • RIZZO et al. •• ESA Session 5 • August 26th 2014 •• bit.ly/ESA_2014

DATA| TerUti (1981-2004)

.015

1 2 3 4 5

13 14 15 16 17

4 700 mailles pour la couverture du

territoire…

la maille

la photo aerienne

… 8 positions de photographies

par maille …

… 36 points à enquêterpar photographie

6

18

7 8 9 10 11 12

19 20 21 22 23 24

25 26 27 28 29 30

31 32 33 34 35 36

553 250 points 1981-1990555 903 points 1992-2003154 501 points 2004

about 0.5 million of points

•• Temporospat ia l ana lys i s o f agr i cu l tura l systems • RIZZO et al. •• ESA Session 5 • August 26th 2014 •• bit.ly/ESA_2014

DATA| TerUti-Lucas (2006-...)

.016

Sous-echantillon

18 km (Lucas) 1

6 km 1+2

6 km doublé 1+2+3

maître 3 km 1+2+3+4

11 12 13 14 15

21 22 23 24 25

31 32 33 34 35

41 42 43 44 45

51 52 53 54 55

18 Km

300 m

•• Temporospat ia l ana lys i s o f agr i cu l tura l systems • RIZZO et al. •• ESA Session 5 • August 26th 2014 •• bit.ly/ESA_2014

Analyses des successions de cultures

22

cluster 6CB(O) ‘90

cluster 3 polycult.élev.

cluster 4cultures ’90 (urbain)

sN -Blé -

Jach -

PPP -autres -

P.Art -

P.Tem -Maïs -Colza -

Orge -Art -

Maïs tête de rotation

sN -Blé -

Jach -

PPP -autres -

P.Art -

P.Tem -Maïs -Colza -

Orge -Art -

sN -Blé -

Jach -

PPP -autres -

P.Art -

P.Tem -Maïs -Colza -

Orge -Art -

sN -Blé -

Jach -

PPP -autres -

P.Art -

P.Tem -Maïs -Colza -

Orge -Art -

sN -Blé -

Jach -

PPP -autres -

P.Art -

P.Tem -Maïs -Colza -

Orge -Art -

sN -Blé -

Jach -

PPP -autres -

P.Art -

P.Tem -Maïs -Colza -

Orge -Art -

CBO, maïs-blé CBO, MBCB

19

92

-19

96

19

99

-20

03