Modelling Economics of Land Degradation

1. Modelling Economics of Land Degradation
Utkur Djanibekov
Email: u.djanibekov@ilr.uni-bonn.de
December 4, 2014
Center for Development Research (ZEF)

2. Aim of the presentation
• Mathematical programming models for the issues of land
degradation
• Get a feel of applying a model in land degradation problem
• Learn about some aspects of including a MP model (scale of
analysis, payments for ecosystem services, stochastic
simulation …)
• Provide policy recommendations based on model output
2

3. Why we need to model the land degradation?
• The problems of land degradation result in food insecurity and
poverty
• Finding solutions to this problem requires identifying effective
entry points for farmers, governments, and civil society
organizations, and understanding the potential impacts and
tradeoffs that are likely to arise from alternative interventions
• Mathematical programming models can assist us to adapt to a
changing ecological and technological conditions, and policies
3

4. Previous studies (1)
Reference Objective of study Model description
Barbier and Bergeron
(2001)
To simulate the effect of
population pressure,
technological improvement,
and policy decisions on forest
and soil erosion management in
the microwatershed area of
Honduras
Recursive LP method that maximizes the utility
of the community (ranchers and small farmers).
Model characteristics:
- Crop (maize, vegetables), pasture and
livestock production (cattle, oxen, mules)
and their interaction;
- Soil erosion: yields decrease as a function of
quantity of soil eroded (modeled by EPIC);
- Perennial crops (3 tree types);
- Biophysical model EPIC: used to model the
effect of land practices on soil quality and
effect of changed soil quality on yields;
- Product allocation (constrain in market sale
of vegetables)
4

5. Previous studies (2)
Reference Objective of study Model description
5
Holden, Shiferaw,
Pender (2005)
To analyze the development
path, impacts of alternative
policies on poverty reduction,
increased food security, and
promotion of sustainable land
use in Ethiopia.
Dynamic NLP method that maximizes the utility
of farm.
Model characteristics:
- Crop (6 crops) and livestock production (6
livestock) and their interaction;
- Linear soil erosion and nutrient depletion
functions based on land use type;
- Risk averse behavior of farm and variability
in rainfall;
- Credit, labor, land, crop, animal, input and
output markets;

6. • Analysis of afforestation on marginal lands as an option to
tackle the problems of land degradation
• Multi-scale analysis:
• Field level
• Farm level
• Bimodal agricultural system level
Previous studies (3): Example of Uzbekistan

7. Source: MAWR (2001, 2010); Dubovyk et al. (2013)
• Irrigated agriculture: 35% of GDP
• Marginal croplands: 23% of arable area
• Main crop on marginal land: cotton
• Farms possess: ~90% of arable land
• Semi-subsistence households possess:
~10% of arable land
Example of Uzbekistan
7

8. • Russian olive (Elaeagnus angustifolia), Euphrates Poplar (Populus euphratica),
Siberian elm (Ulmus pumila) showed a high potential on marginal croplands
• Require less irrigation than crops due to reliance on groundwater
• Multiple products: fuelwood, fruits, leaves as fodder, carbon revenues through
Clean Development Mechanism (4.76 USD tCO2
-1)
Afforestation of marginal croplands (1)
Photos: Khamzina et al. (2012)
8
March 2004 May 2006 August 2007
Source: Lamers et al. (2008); Khamzina et al. (2006; 2009)

9. Afforestation of marginal croplands (2)
• Costs and benefits of trees are not known
• Variability in yields, prices, irrigation water uncertainty in incomes
• Farmers follow the state cotton procurement policy: 50% of farmland,
cotton output, state purchase price low flexibility in land use
• Household incomes depend on employment at farm spillover effects
0
100
200
300
400
500
2001 2002 2003 2004 2005 2006 2007 2008 2009
Wheat Rice Maize Vegetables
Years
Changeincropprices,%
9

10. Research questions and methods
What are the costs and benefits of
afforestation on marginal croplands?
What are the costs and benefits of
afforestation under uncertainty?
What are the impacts of afforestation on
rural livelihoods?
Costs and benefits at
field level (1 ha)
Bimodal system
(Farm and household)
What policies are needed to facilitate the adoption of more sustainable
land use on marginal lands – afforestation?
Costs and benefits at
field and farm level
(1)
(2)
(3)
(4)
AnalysisResearch questions
10

11. Method (1): Costs and benefits
Data Description
Surveys 160 farms, 400 households, market
Tree growth parameters Russian olive, Euphrates poplar, Siberian elm from
experimental cite over 7 yrs
Irrigation-yield response
function
Cotton, wheat, rice, maize and vegetables on
marginal, average, good and highly productive soils
• Net present value (NPV): difference between the discounted value of cash
inflows and outflows of land uses. Comparison of NPV of land uses
• Scale: 1 ha level
11

12. Results (1): Costs and benefits of trees
Source: Djanibekov et al (2012) For Policy Econ
• Russian olive has the highest NPV due to annual fruit harvest
• Low management costs
-1,500
-500
500
1,500
2,500
3,500
4,500
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
Costs(-)andbenefits(+),USDha-1
Leaves Fruits Fuelwood Carbon Labor costs Other costs
Russian olive Poplar Elm
Years
12

13. Results (1): NPV of crops and trees over 7 years
Source: Djanibekov et al. (2012) For Policy Econ
• Afforestation is more profitable on marginal lands than crops, except rice
Crops and trees
Net present value
Irrigation water
requirement
USD ha-1
1,000 m3
ha-1
Wheat -74 5.38
Cotton -330 5.98
Rice 8,369 26.59
Maize 1,800 5.3
Vegetables 2,405 8.6
Russian olive 5,794 1.6
Poplar 1,894 1.6
Elm 724 1.6
13

14. Method (2): Uncertainty in revenues
Methods Description
(1) Stochastic dominance Comparison of varying NPV of land uses (1 ha level, 7 yrs)
(2) Expected utility Maximization of farm weighted average NPV/utility of
possible outcomes (100 ha, 7 yrs)
Farm size: 23 ha marginal, 56 ha average, 20 ha good, and
1 ha highly productive lands
Scenarios:
1) Business-as-usual (BAU): cotton on 50% of farmland,
cotton output, state cotton price
2) Afforestation: cotton output, state cotton price
Monte Carlo simulation: yields, prices, irrigation
availability
14

15. Results (2): Varying NPV of land uses
Source: Djanibekov and Khamzina (in press) Environ Resource Econ
• Due to variability additional financial incentives are needed to initiate
afforestation
0
0.2
0.4
0.6
0.8
1
-3 6 15 24
Cotton Wheat Rice Maize
Vegetables Russian olive Poplar Elm
Cumulativeprobability
Net present value of land uses, 1,000 USD ha-1
0
0.2
0.4
0.6
0.8
1
-3 6 15 24
Rice Russian olive
Cumulativeprobability
Net present value of land uses, 1,000 USD ha-1
Lowest NPV of
trees
Highest NPV
of crops
15

16. Results (2): Payment for carbon under uncertainty
• Variability in returns may require high carbon prices (110 times higher than
the current value) to initiate afforestation
Russian olive
0
100
200
300
400
500
600
-3 6 15 24
MinimumNPV of trees Maximum NPV of trees
Net present value of crops, 1,000 USD ha-1
Carbonprice,
1,000USDtCO2
-1
Source: Djanibekov and Khamzina (in press) Environ Resource Econ
Highest NPV of
crops
Carbon price to
make NPV of
tree=crops
16

17. Results (2): Farm land use
• Even without carbon revenues, trees are planted on marginal lands
• Afforestation: farm incomes are higher by 29 % than under the BAU
Source: Djanibekov and Khamzina (in press) Environ Resource Econ
0
25
50
75
100
125
150
BAU 0 4.76
Elm
Poplar
Russian olive
Vegetables
Maize
Rice
Wheat
Cotton
Landusepatternoffarm,ha
Afforestation with
USDtCO2
-1
Business-as-
usual
Afforestation with
carbon payment,
USD tCO2-1
17

18. • The financial attractiveness of trees is higher in the
deterministic analysis than in the stochastic analysis
• Using the expected utility approach (whole-farm level), the
modification of cotton policy leads to afforestation and
increases farm incomes
But…
• Since rural interlinkages exist, impact on rural livelihoods might
be different
Discussion (1 & 2)
18

19. • Semi-subsistent households are dependent on employment at farms
Source: Djanibekov et al. (in press). In Lerman and Csaki (Eds)
Rural interdependencies
Household characteristics
Household members 7
Household members employed at farm 3
Household members employed at non-agricultural activities 2
Share of income from marketing livestock and crops from own plot, % 24
Share of income from cash, crops as payments in kind from farm employment, % 13
Share of income from land rented, given as payment in kind, sharecropping, % 19
Share of income from non-agricultural activities, % 44
19

20. Rural interdependencies: Framework
Source: Djanibekov et al. (2013) Land Use Policy; Djanibekov et al. (2013) J Rural Stud
• Economies of farms and households are interlinked through wage-labor
relations spillover effects of farm afforestation
Legends:
Activities
Constraints
Farm-household
interdependencies
Constraining effects
Products
Crops
Food
consumption
Domestic
energy use
Livestock
fodder
Land
Income
Labor
State policy
Afforestation
Crops
Products
Land
Income
Labor
Farm (100 ha) Households(0.2 ha)
Wage-labor
relationship
Irrigation water
Products
Land
Cash
Labor
20

21. Method (3): Bimodal model
Stochastic dynamic bimodal model
Settings Description
Time-step 28 years with 3 tree plantation rotations (each 7 years)
Scale Farm and households level
Farm size 23 ha marginal, 56 ha average, 20 ha good, and 1 ha highly
productive lands
Household size 0.2 ha average productive lands
Scenarios 1) Business-as-usual (BAU): cotton on 50% of farmland,
cotton output, state cotton price
2) Afforestation: cotton output, state cotton price
Main assumptions Storage of goods, weight carriage, various contracts
21

22. Afforestation
Business-as-usual
50
100
150
200
1 4 7 10 13 16 19 22 25 28
Result (3): Rural livelihoods
Incomeofhousehold,
USDcapita-1
60
90
120
150
1 4 7 10 13 16 19 22 25 28
Profitoffarm,
USDcapita-1
500
800
1,100
1,400
1,700
2,000
1 10 19 28
BAU Afforestation
Years
500
800
1,100
1,400
1,700
2,000
1 10 19 28
BAU Afforestation
Years
50
60
70
80
90
1 4 7 10 13 16 19 22 25 28
500
800
1,100
1,400
1,700
2,000
1 10 19 28
Years
Households'laboremployed
atfarm,1,000workinghours
10 13 16 19 22 25 28
Afforestation
Years
1 4 7 10 13 16 19 22 25 28
BAU Afforestation
YearsScenarios
Clear cut
of trees
22

23. Summary
Analysis Results
NPV of land uses
(1 ha scale)
Afforestation is more profitable than crops on marginal
lands
Varying NPV of land uses
(1 ha scale)
Variability in revenues requires additional financial
incentives to initiate afforestation (such as increase in
carbon payments via CDM)
Farm level model
(100 ha)
Flexibility in land use activities of farmers leads to
afforestation (additional financial incentives are not
needed)
Bimodal system model
(farm and household level)
Land use policies aimed for farmers indirectly affect
other groups of rural population
23

24. The model of the ELD project
24

25. Model description
Objective function Maximize incomes of heterogeneous agricultural
producers (smallholder, crop and livestock farms)
Farm activities and
land uses
6 annual and 3 perennial crops, pasture, forest, 3 types of
livestock. Detailed product info: main and by-products
Technology Multiple input piecewise production functions for crops
Dynamics Fully dynamic recursive
Stochasticity For yields, prices, irrigation and rainfall
Time resolution Annually
IT GAMS (General Algebraic Modelling System), CONOPT 3
solver, Excel for data import and export
Typical size With 20 years and all crops, 140.000 variables, 1.6 Mio
non-zeros
Model characteristics (1)
25

26. • We are developing:
- A standardized set of equations and variables, which can
be applied to different farms with different attributes
- A set of various land degradation indicators that estimate
land degradation at farm level
- Normative approach as empirical observations (partly) not
available
• That model is used to:
- To simulate how farmers adjust their production decisions
when facing land degradation
- To derive farm specific land degradation tackling options
26
Model characteristics (2)

27. • Cropland degradation:
- The crop allocation and input use decisions of individual farm
affect soil nutrient
- The more intensive is production the higher is yield in short-
term, but decreases in long-term
- Soil nutrients can be increased through using conservation
agricultural practices (residue retention) and increased
application of fertilizers and manure
27
Land degradation (1)

28. • Pasture degradation:
- Overgrazing of pastures reduce grass production next year
- Conversion of pasture to cropland
- The rainfall also affect grass output from pasture
• Forest degradation:
- Conversion of forest to cropland
28
Land degradation (2)

29. • To integrate data on agricultural production information
for some countries
• To integrate into the model the biophysical parameters
(data) on land degradation
• Validate the model
29
Further steps

30. Thanks for your attention
Email: u.djanibekov@ilr.uni-bonn.de