Agriculture, Ecosystems and Environment 75 (1999) 55–73

A framework for integrated biophysical and economic land use analysis

at different scales
Bas A.M. Bouman a,b,∗ , Hans G.P. Jansen a , Rob A. Schipper c , Andre Nieuwenhuyse a ,
Huib Hengsdijk a , Johan Bouma d
aWageningen Agricultural University, REPOSA, Apartado 224, 7210 Guápiles, Costa Rica
bDLO-Research Institute for Agrobiology and Soil Fertility, Wageningen, The Netherlands
c Wageningen Agricultural University, Department of Development Economics, Wageningen, The Netherlands
d Wageningen Agricultural University, Department of Environmental Sciences, Laboratory of Soil Science and Geology, Wageningen, The
Netherlands
Received 1 September 1998; received in revised form 1 February 1999; accepted 18 March 1999

Abstract
There is a general need for quantitative tools that can be used to support policy makers in regional rural development. Here, a
framework for (sub-) regional land use analysis is presented that quantifies biophysical and economic sustainability trade-offs.
The framework, called sustainable options for land use (SOLUS), was developed over a 10-year period of investigation in
the Northern Atlantic Zone of Costa Rica and encompasses scale levels that range from field to region. SOLUS consists of
technical coefficient generators to quantify inputs and outputs of production systems, a linear programming model that selects
production systems by optimizing regional economic surplus, and a geographic information system. Biophysical and economic
disciplines are integrated and various types of knowledge, ranging from empirical expert judgement to deterministic process
models are synthesized in a systems-analytical manner. Economic sustainability indicators include economic surplus and labor
employment, and biophysical ones include soil N, P and K balances, biocide use and its environmental impact, greenhouse
gas emission and nitrogen leaching loss and volatilization. Land use scenarios can be implemented by varying properties
of production inputs (e.g., prices), imposing sustainability restrictions in the optimization, and incorporating alternative
production systems based on different technologies. Examples of application of SOLUS in the Northern Atlantic Zone of
Costa Rica show that introduction of alternative technologies may result in situations that satisfy both economic as well
as biophysical sustainability. On the other hand, negative trade-offs were found among different dimensions of biophysical
sustainability themselves. ©1999 Elsevier Science B.V. All rights reserved.
Keywords: Land use analysis; Sustainability; Policy decision support

1. Introduction

Policy makers and stakeholders (i.e., individuals,

∗ Corresponding author. Present address: Soil and Water Sci- communities or governments that have a traditional,
ences Division, International Rice Research Institute, P.O. Box
3127, 1271 Makati City, Philippines; Tel.: +63-2-845-0563; fax:
current or future right to co-decide on the use of land
+63-2-891-1292 (FAO, 1995)) concerned with regional (rural) devel-
E-mail address: b.bouman@cgiar.org (B.A.M. Bouman) opment increasingly face the need for instruments that

0167-8809/99/$ – see front matter ©1999 Elsevier Science B.V. All rights reserved.
PII: S 0 1 6 7 - 8 8 0 9 ( 9 9 ) 0 0 0 5 9 - 6
56 B.A.M. Bouman et al. / Agriculture, Ecosystems and Environment 75 (1999) 55–73

can improve transparency in the policy debate and and land use allocation (Veldkamp and Fresco, 1996).
that enhance understanding of opportunities and limi- Over the years, a general framework has evolved for
tations to development. To this end, a number of land land use studies at (sub) regional level using tools
use models have been developed in the last decade and methods of systems-analysis, called sustainable
based on methods of systems analysis (De Wit et al., options for land use (SOLUS). The framework en-
1988; Rabbinge and van Latesteijn, 1992; Kruseman compasses scale levels that range from field to farm,
et al., 1995; Kuyvenhoven et al., 1995; Penning de sub-region and region. This paper presents the SO-
Vries et al., 1995). A common characteristic of such LUS framework with examples for three applications
models is that they aim at the quantification of bio- in the Northern AZ of Costa Rica: a farmers’ settle-
physical and economic sustainability trade-offs to sup- ment (Neguev), an administrative canton (Guácimo)
port policy decision making with respect to agricul- and the entire Northern AZ. Special attention is given
tural land use. Quite often, however, such models are to the integration of biophysical disciplines with eco-
strongly biased to either the biophysical sciences (Pen- nomics; to issues relating to scale in the design; and
ning de Vries et al., 1995) or economics (Kruseman to building and application of the framework. The ul-
et al., 1995) which has been both defended (Van Lat- timate clients of the framework are regional to na-
esteijn, 1995) as well as criticized (Tims, 1995). The tional stakeholders in rural development, such as gov-
successful integration of the biophysical sciences and ernments, policy makers and interest groups (such
economics still constitutes a major challenge to re- as producer groups or nature conservation groups).
searchers in land use studies (Crissman et al., 1997). Farmers are not considered individually as stakehold-
Another issue that recently received much attention ers, nor are they individually addressed in SOLUS;
is that of spatial scale (Fresco, 1995; Jansen, 1995; specific models developed at the level of individual
Bouma, 1997; Dumanski et al., 1998). Spatial scale is- farms within the REPOSA program have been reported
sues relate to aggregation and up- and down-scaling of by Kruseman et al. (1995) and Kuyvenhoven et al.
information (model input data) and of biophysical and (1995).
economic relationships. Many of the mentioned land
use models operate at a particular scale level, vary-
ing from farm level (Kruseman et al., 1995) to global 2. Site description
level (Penning de Vries et al., 1995), and integration
of scales is yet another research challenge. The Northern AZ is in the Caribbean lowlands of
In 1986, the Wageningen Agricultural University Costa Rica and covers the northern half of the province
initiated a joint program with the Centre for Research of Limón, roughly between 10◦ 000 –11◦ 000 latitude and
and Education in Tropical Agriculture (CATIE) and 83◦ 000 –84◦ 000 longitude (Fig. 1). The humid tropical
the Ministry of Agriculture and Livestock of Costa climate is characterized by a mean daily temperature
Rica on sustainable land use in the Northern Atlantic of 26◦ C, a mean annual rainfall of 3500–5500 mm,
Zone (AZ) of Costa Rica, of lately known as research and average relative humidity of 85–90%. Four ma-
program on sustainability in agriculture (REPOSA). jor soil groups were distinguished, based on crite-
Its aims were to develop quantitative methodologies ria of fertility and drainage: (i) young alluvial, well
for analysis and evaluation of alternative scenarios for drained volcanic soils of relatively high fertility (In-
profitable and sustainable land use at the farm and ceptisols and Andisols; USDA Soil Taxonomy classi-
(sub) regional level, that can serve to support pol- fication), classified as soil fertile well drained (SFW),
icy design and decision making. Among the main (ii) old, well drained soils of relatively low fertil-
challenges were integration of biophysical and eco- ity developed on fluvio-laharic sediments (oxisols and
nomic disciplines, and spatial scale issues. A number inceptisols), classified as soil infertile well drained
of approaches were elaborated at various scale levels, (SIW), (iii) young, poorly drained alluvial soils of rel-
ranging from field-level decision support (Stoorvogel, atively high fertility (entisols and inceptisols), clas-
1998), to farm-level policy models (Kruseman et al., sified as soil fertile poorly drained (SFP), and (iv)
1995), regional biophysical exploration of land use soils not suited for agriculture including peat soils
(Bessembinder, 1997) and national trend extrapolation (histosols), shallow infertile mountain soils on sed-
 B.A.M. Bouman et al. / Agriculture, Ecosystems and Environment 75 (1999) 55–73 57

Fig. 1. Location of the Northern Atlantic Zone, canton Guácimo and the Neguev settlement in Costa Rica.

imentary rock (inceptisols) and soils developed on the area is quite irregular (Fig. 2 (b)), and gives rise to
volcanic ash under extreme humid conditions (hy- differentiated product transport costs throughout the
drudands) (Nieuwenhuyse, 1996). The distribution of area (see Fig. 2 (c) and Section 4.3 below).
these main soil groups is given in Fig. 2 (a). Of the Guácimo is a county of 58,000 ha within the North-
total area of some 447,000 ha, 334,000 ha is suitable ern AZ, and the Neguev is an agrarian settlement of
for agriculture of which 55,000 ha is protected area for 5340 ha partly within Guácimo county.
nature conservation. Land use is dominated by natural
forest (48%), cattle ranching (39%) and banana (Musa
AAA) plantations (10%). The 3% remaining area is 3. Methodology development
cultivated with various crops such as plantain (Musa
AAB), palm heart (Bactris gasipaes), root and tuber The SOLUS framework to analyze and evaluate
crops (e.g., cassava, Manihot esculenta Crant), maize land use scenarios comprises three main components:
(Zea mays L.), pineapple (Ananas comosus L.), beans technical coefficient generators to quantify inputs and
(Phaseolus vulgaris L.) and timber plantations (e.g., outputs of production systems; a linear programming
melina, Gmelina arborea). The road infrastructure in model; and a geographic information system (GIS)
58 B.A.M. Bouman et al. / Agriculture, Ecosystems and Environment 75 (1999) 55–73

Fig. 2. Distribution of major soil types (a), road infrastructure (b), and identified sub-regions (c) on the basis of product transport costs in
the Northern AZ. Maps were created in GIS, and the data served as input for SOLUS in the Northern AZ implementation.
 B.A.M. Bouman et al. / Agriculture, Ecosystems and Environment 75 (1999) 55–73 59

Fig. 3. Schematic illustration of the land use exploration framework SOLUS. Boxes are models/tools; ovals are data; blank names are
activities; drawn lines are flow of data; dotted lines are flow of information. In the Neguev and Guácimo applications of SOLUS, the
technical coefficient generator was MODUS, and in the Northern AZ application two technical coefficient generators were used: PASTOR
and LUCTOR.

(Fig. 3). A linear programming model selects, for an a land utilization type (FAO, 1993)). For cattle ranch-
area under consideration, production systems from a ing, three sub-systems are defined: pastures, herds
large number of alternatives by optimizing a so-called and feed supplements. The herds produce marketable
objective function given certain optimization con- output and are characterized by certain feed require-
straints. The objective function in a linear program- ments which must be met by a combination of feed
ming model is the function Z = f(X) in the optimiza- from pastures and from feed supplements.
tion problem ‘maximize (or minimize) Z’, subject to The objective function of linear programming mod-
constraints on X. In applications in land use analy- els in the SOLUS framework typically maximizes eco-
sis, X represents a vector of production systems with nomic surplus, though optimization of other goal func-
corresponding technical coefficients, and Z represents tions is possible as well, e.g., the minimization of
the variable to be optimized, e.g., economic surplus certain environmental effect indicators, or the maxi-
or use of biocides. The selection of production sys- mization of employment in agriculture (Jansen et al.,
tems is based on their quantified inputs and outputs, 1997). Economic surplus is calculated as the sum of
which are called technical coefficients in linear pro- the output value (i.e., quantities of crop and livestock
gramming terminology. Three categories of technical products times their corresponding farm-gate prices)
coefficients are distinguished: (i) economic, i.e., costs minus production costs, consisting of labor costs and
of production and labor use, (ii) physical production, material costs (i.e., costs of fertilizers and biocides,
(iii) biophysical sustainability indicators (e.g., soil annualized costs of capital items such as machinery,
nutrient status, emission of harmful substances to the corals etc.). Optimization constraints are restrictions
environment, such as biocides or greenhouse gases). under which the objective function of an optimization
Biocides comprise all kinds of agrochemicals used to problem is maximized (or minimized). In SOLUS,
combat weeds (herbicides), pests and diseases (e.g., two types of constraints are distinguished: resource
nematicides, fungicides, insecticides). Land based constraints to agricultural production (e.g., available
production systems include annual and perennial food land and labor), and normative constraints (e.g., up-
crops, cattle ranching, and forestry, ranging from tree per limits on sustainability indicators or lower limits
plantations to the extraction of timber from natural on production levels). Running a linear programming
forest. Technically, these systems are defined as land model subsequently under different constraints results
use system (i.e., the combination of a land unit with in quantified trade-offs among economic and biophys-
60 B.A.M. Bouman et al. / Agriculture, Ecosystems and Environment 75 (1999) 55–73

ical sustainability indicators. Different linear program- systems are quantified per land unit. Where possible,
ming models were developed for the various applica- scale effects are taken into consideration in the cal-
tions of SOLUS to the Northern AZ (see Section 4). culations of technical coefficients by using produc-
The technical coefficients of production systems tion functions that are homogeneous in the second
that feed a linear programming model are calculated order. Up-scaling is done in the linear programming
by so-called technical coefficient generators. Tech- model where land units are aggregated into farms,
nical coefficient generators are expert systems that sub-regions or whole regions.
combine process-based knowledge with empirical data The general temporal unit of analysis is one year.
and expert knowledge to calculate inputs and outputs Occasionally, it might be necessary to define labor use
of production systems (Hengsdijk et al., 1998). For for shorter time periods, e.g., to take account of the
different production systems, a large number of op- effects of labor peaks due to coinciding management
tions can be generated by varying (targeted) output operations of different production systems. Similarly,
levels and the level and/or type of technology em- in regions with pronounced growing seasons, it might
ployed, including possibilities for substitution between be decided to define technical coefficients on a sea-
inputs. As with the linear programming models, dif- sonal basis. The use of different temporal scales in
ferent technical coefficient generators were developed technical coefficients presents no problem as long as
for the various applications of SOLUS (as described they are properly used in the constraint and balance
in Section 4). equations in the linear programming model. Typical
In SOLUS, two types of input data exist: non- time horizons of land use studies with SOLUS vary
geographically referenced and geographically ref- from short term (0–5 years) to long term (≈50 years).
erenced data. Most non-geographic data are used
directly in the calculation of technical coefficients,
and include such elements as crop, pasture and herd
characteristics, land and soil properties, a discount 4. Implementation of SOLUS
rate and so-called attribute data of production in-
puts. Examples of attribute data are prices, labor use, 4.1. The Neguev settlement
depreciation time and, in the case of biocides, con-
centrations of active ingredients, toxicity parameters The implementation of SOLUS to the Neguev set-
and half-life time. Examples of non-geographic data tlement was known by the Spanish acronym USTED
used in the linear programming model are prices and (Uso Sostenible de Tierras En el Desarrollo) (Schip-
quantities of products, price elasticities of demand per et al., 1995; Stoorvogel et al., 1995). Main data
and supply for outputs and labor. Geographic data input and characteristics for the Neguev implementa-
characterize the region under study and quantify the tion are given in Tables 1 and 2. The study focused
spatial distribution of resource endowments such as on the exploration of cropping options for small and
land and labor. Information on characteristics of land medium-scale farmers (i.e., below 50 ha in size).
units and weather is used in the generation of tech- Large-scale banana plantations and cattle ranching
nical coefficients, while resource endowments figure were excluded from the analysis, although pasture
in the linear programming model as boundary con- was included by directly valuing forage production.
straints in scenario analysis. GIS is used for archiving The technical coefficient generator that was devel-
and manipulating geo-referenced input data and for oped for this study is called MODUS (modules for
presenting spatial output results. The data flow be- data management in USTED; Stoorvogel et al., 1995).
tween the GIS, the technical coefficient generators MODUS is mainly a descriptive book-keeping system
and the linear programming model is highly auto- where all inputs and outputs of production systems are
mated due to well specified data exchange protocols specified by the user. For example, for a certain land
(Stoorvogel, 1995; Bouman et al., 1998b). use system, yield level and all operations should be
The basic spatial unit of SOLUS is the land unit, specified by the user. The specification of operations
which is defined as a unique combination of soil, land- includes a full description of timing and use of pro-
scape and weather. All crop and pasture production duction factors (labor hours, machines, tools, fertil-
 B.A.M. Bouman et al. / Agriculture, Ecosystems and Environment 75 (1999) 55–73 61

Table 1
Characteristics of the implementation of SOLUS to the Neguev settlement (taken from Alfaro et al., 1994), Guácimo county (taken from
Jansen et al., 1997) and the NAZ region (taken from Bouman et al., 1999)
Characteristics Neguev Guácimo Northern AZ
Production systems 5 crops: cassava, maize, palm 5 crops: cassava, maize, palm 8 crops: banana, bean, cassava, grain
heart, pineapple, plantain heart, pineapple, plantain maize, fresh cobs maize, palm heart,
pineapple, plantain
2 forestry (1 logged, natural 2 forestry (1 logged, natural 3 forestry (1 logged forest, 2 plantations
forest, 1 plantation Gmelina forest, 1 plantation Gmelina (Gmelina arborea and Tectona grandis))
arborea) arborea)
1 natural pasture (mixture of 1 natural pasture (mixture of 5 pastures: natural grass pasture (mixture
various spp.) various spp.) various spp.), grass–legume mixture (B.
brizantha–A. pintoi), Estrella (Cynodon
nlemfuensis), Brachiaria (Brachiaria
brizantha), Tanner (Brachiaria radicans);
2 beef cattle systems: breeding, fatteninga
5 feed supplements: rejected bananas, sugar
cane molasses, two types of chicken-dung
based concentrates, P mineral salt.
Production system 120 land use systems 169 land use systems 3288 land use systems (1532 crops and
options and forests; 1756 pastures)
8 animal production systems
5 feed supplements
Farm types 5 9 none
Sub-regions none none 12
Markets not included not included demand and supply relationships for
product prices and labor
a In the breeding system, calves are bred and subsequently sold at a certain age or live weight; no animals are bought externally. In the
fattening system, young animals are bought, fattened for a period of time, and then sold; no animals are bred internally.

izer, biocides, etc.) for all activities carried out during in the production system. The soil nutrient balances
the cropping cycle (land preparation, sowing, weed- are calculated using the NUTMON model (Stoorvo-
ing, harvesting, etc.) (Jansen and Schipper, 1995). For gel, 1993), which is a straightforward book-keeping
the Neguev application, this information was mainly of all inputs (such as atmospheric deposition, fixa-
derived from field experiments and farm surveys, sup- tion by micro-organisms, manure, urine, fertilizer)
plemented by data obtained from crop growth models and all outputs (such as losses by erosion, leaching,
for alternative cropping systems not (yet) encountered volatilization and denitrification/nitrification). Input
in the area (such as the near potential production data for NUTMON were obtained from a wide range
of maize). The biophysical sustainability indicators of literature data, own field experiments and mea-
are not entered by the user, but are calculated by surements in the area, and expert opinions (see e.g.,
MODUS as a function of production inputs and soil Bouman and Nieuwenhuyse (1999) for a detailed pre-
properties. The indicators were derived from relevant sentation and discussion of data and their sources).
sustainability issues in the area (see Jansen et al., The reasoning behind the use of soil nutrient balances
1995, for a detailed review of sustainability issues in is that with negative balances, soils are being mined
Neguev), and included soil N, P and K balances and and productivity is expected to decline over time.
an index that quantifies the environmental hazard of Three land units were identified exclusively based
biocides by taking account of the amount of toxic on soil type differentiation (landscape aspects and
ingredients used, their toxicity level and their half life weather being uniform over soil types). A total of
time, called biocide index (Jansen et al., 1995). The 120 land use systems were quantified by varying the
biocide index is calculated from all biocide inputs level of technology and production (as specified by
62 B.A.M. Bouman et al. / Agriculture, Ecosystems and Environment 75 (1999) 55–73

Table 2
Input data inputs for the application of SOLUS to the Neguev settlement (taken from Alfaro et al., 1994), Guácimo county (taken from
Jansen et al., 1997) and the NAZ region (taken from Bouman et al., 1999)
Data Neguev/Guacimo NAZ
Biophysical crop, forest and pasture characteristics crop, forest and pasture characteristics (e.g., yield
(e.g., yield potential, nutrient concentrations) potential, nutrient concentrations) herd characteritics
(e.g., meat production , feed requirements)
management characteristics of crops, forests and pastures management characteristics of crops, forests and pastures
weather data management characteristics of herds
land characteristics weather data
soil properties land characteristics ( e.g., slope, stoniness, drainage class)
attribute data soil properties
attribute data
GIS soils soils
administrative boundaries roads
farm boundariesa district boundaries
labor force distribution
protected areas
Economic input/output prices input prices
farm size distribution base prices and quantities of products
farm assets: land (soil) and labor base quantity and wage of agricultural labor
total labor availability national price elasticities of demand
base labor wage regional price elasticities of product supply
discount rate share of regional product supply
price elasticities of product supply of other regions
national price elasticity of labor supply
national price elasticity of labor demand
share of regional labor in national labor supply
product market outlets
transport costs of products
labor mobility costs
total labor size
discount rate
a Neguev only.

the description of operations and the associated yield jectives of the farm household did not result into an
level in MODUS) for five crops, two forestry systems improved classification (Schipper et al., 1995). The
and one pasture on the three land units (Table 1). five farm types are specified in the linear program-
Decisions regarding land use are usually taken by ming model by their resource endowments, which are
the farmer. Therefore, in USTED, the farm level served summed to get the total resources of the whole Neguev
as intermediate level between the level of land unit settlement. Interaction among the farm types was taken
and the Neguev as a whole. Farms were classified into into consideration by restricting the amount of hired
five types based on resource endowments: size, avail- labor on each farm type by the availability of off-farm
ability of land units and land–labor ratio (Table 3). labor from other farm types. Employment opportuni-
Information on farm size and land units was derived ties outside the Neguev consisted of work in banana
from overlaying a map of boundaries of all 307 farms plantations which, at this scale level, could safely be
that were present in the Neguev at the time of the assumed unlimited. Next to these fixed resource con-
study with a detailed 1 : 20,000 soil map; land–labor straints, optional normative constraints could be im-
ratios were based on an assumed labor force of two posed on the biophysical sustainability parameters at
per household derived from field inquiries. Attempts the level of land unit, farm type or the whole Neguev.
to further refine the farm typology by including ob- Transport costs were found to be the same throughout
 B.A.M. Bouman et al. / Agriculture, Ecosystems and Environment 75 (1999) 55–73 63

Table 3
Farm types with their resource endowments in the Neguev application of SOLUS (Schipper et al., 1995)
Farm type Number of farms Average area (ha) Land units (%) Labor (person)

SFWa SFPb SIWc

1 33 15.7 12 60 28 2
2 4 32.1 10 12 78 2
3 46 13.5 52 7 41 2
4 35 14.1 91 3 6 2
5 189 13.1 6 6 88 2
Total 307 13.8 23 13 64 2
a SFW: soil fertile well drained.
b SFP: soil fertile poorly drained.
c SIW: soil infertile well drained.

the Neguev. Since the Neguev is relatively small, pro- To prepare SOLUS for up-scaling towards the level
duction could safely be assumed not to affect prices, of the entire Northern AZ, sensitivity analyses were
and all input and output prices were fixed. Examples of performed with the Guácimo model, focusing on the
applications of the SOLUS/USTED framework to the issue of farm aggregation (Jansen and Stoorvogel,
Neguev are given by Alfaro et al. (1994); Jansen et al. 1998). It was found that farm type classification did
(1995); Schipper et al. (1995) and Schipper (1996). significantly affect results at the farm level, but hardly
at the level of the entire county. That is, even though
4.2. Guácimo county at the farm level, the model was sensitive to the di-
agnostic criteria of size and land unit distribution,
For the county of Guácimo, a similar set-up as for differences in resource endowments between farm
the Neguev was implemented, with only slight modi- types did not significantly influence model outcomes
fications (Tables 1 and 2). The number of technology at the level of the county. Also, even though the
alternatives was extended so that 169 land use systems model was found to be sensitive to spatial variation
were generated, even though the Guácimo model still in product transport costs, such variation in Guácimo
excluded the cattle ranching sector and large-scale ba- was too small to significantly affect model results. In
nana plantations. The diagnostic criteria size and land addition, it was shown that the assumption of a fixed
unit distribution resulted in the differentiation of nine wage rate throughout the year (i.e., ignoring labor
farm types (Table 4). Because of the large number of market competition between farm types) implies a
farms in Guácimo (a total of 1816), however, the deter- significant source of bias.
mination of size and land resources was less absolute
than in the Neguev model: though maps with exact
farm boundaries were available for some settlements 4.3. Northern AZ
within the county, for the majority of the farms, size
was estimated from aerial photographs based on cor- A number of considerations led to redesign of the
relation between parcel size and total farm size (Van linear programming models and the technical coeffi-
de Steeg, 1997). The constant average labor availabil- cient generator used in SOLUS so far, for application
ity per farm type of two man-equivalents as found in to the entire Northern AZ:
the Neguev was assumed to apply to Guácimo as well. 1. Cattle ranching and banana production are the
Input and output prices were fixed, again based on the most important land use activities and conse-
presumption that Guácimo is too small to influence quently needed to be taken into account.
product prices and that transport costs do not result in 2. The intermediate farm level could no longer be
significant spatial variation in farm-gate prices. Exam- maintained because it was impossible to derive
ples of applications of the SOLUS/USTED framework information on the diagnostic criteria of size and
to Guácimo include Jansen et al. (1997) and Stoorvo- endowments of land and labor with sufficient ac-
gel et al. (1995). curacy to arrive at an appropriate farm typology.
64 B.A.M. Bouman et al. / Agriculture, Ecosystems and Environment 75 (1999) 55–73

Table 4
Farm types with their resource endowments in the Guácimo application of SOLUS (Jansen et al., 1997)
Farm type No. of farms Average area (ha) Land units (%) Labor (person)

SFWa SFPb SIWc

1 639 12 79 19 2 2
2 150 12 8 84 8 2
3 274 12 0 8 92 2
4 165 22 80 16 4 2
5 248 22 7 87 6 2
6 286 22 0 7 93 2
7 4 50 40 30 30 2
8 34 50 7 72 21 2
9 16 50 0 4 96 2
Total 1816 17 31 34 35 2
a SFW: soil fertile well drained.
b SFP: soil fertile poorly drained.
c SIW: soil infertile well drained.

3. Transport costs are not uniform in the area, but de- labor to realize a particular production level are used
pend on distance to market outlets and road qual- than necessary from a technical point of view. The
ity. soil nutrient balance is imposed to be in equilibrium,
4. Product prices can no longer be assumed fixed. so that the calculated amount of fertilizer balances the
For some crops, production in the Northern AZ is deficit in other nutrient inputs minus outputs. Thus, the
a significant proportion of total domestic or world generated alternative land use systems are sustainable
market supply, so that it can be expected to influ- from the soil nutrient balance point of view, and can be
ence prices (Hazell and Norton, 1986). practiced theoretically for an infinite number of years
5. Unlike in the Neguev and Guácimo applications, without changing input–output combinations. Next to
the Northern AZ is too large an area to justify the the target production levels, the technology used to re-
assumption that farmers can obtain any quantity alize them is specified as well. The technology used
of labor at a fixed wage rate. affects yield, costs, labor requirements and sustainabil-
Instead of MODUS, new technical coefficient gen- ity indicators. The calculation of inputs is based – as
erators were developed: one for crops, called LUC- far as possible – on well-established system-analytical
TOR (land use crop technical coefficient generator; knowledge of physical, chemical, physiological and
Hengsdijk et al., 1998), and one for cattle ranching ecological processes involved. When such knowledge
systems, called PASTOR (pasture and animal system is incomplete or absent, calculations are based on ex-
technical coefficient generator; Bouman et al., 1998a). pert knowledge, literature data and empirical data. For
Both LUCTOR and PASTOR calculate technical co- example, in PASTOR, while feed requirements of cat-
efficients of actual production systems following the tle are calculated using equations presented by the Na-
descriptive MODUS approach. In addition, alternative tional Resource Council (NRC, 1996), herd structure
production systems are generated using the so-called (i.e., distribution of number, age and weight of ani-
target-oriented approach (Van Ittersum and Rabbinge, mals in the herd) is computed using a model presented
1997): target production levels are predefined and sub- by Hengsdijk et al. (1996), soil nutrient balances are
sequently the amount of required inputs are calculated. calculated using the NUTMON model (Stoorvogel,
For example, for crops and pastures, target produc- 1993), the effect of stocking rate on pasture produc-
tion levels may vary from maximum attainable, via tion is estimated by livestock experts, fertilizer recov-
close-to-actual situations to very low levels, resulting eries are obtained from published field experiments,
in computed high and low external input levels (e.g., and operations such as animal health care are derived
fertilizers, crop protection agents), for the first and the from field survey data (Van Loon, 1997). As part of
last case respectively. No more nutrients, biocides or quality control, all input data, subroutines and equa-
 B.A.M. Bouman et al. / Agriculture, Ecosystems and Environment 75 (1999) 55–73 65

tions, as well as simulated output of both PASTOR zonation. For each sub-region, land unit and labor en-
and LUCTOR were carefully checked against field ob- dowments were calculated by map overlaying in GIS
servations and were reviewed by a number of outside (Table 5). Agricultural labor availability was derived
experts. Detailed information on PASTOR and LUC- from combining administrative boundaries with pop-
TOR is given by Bouman et al. (1998a) and Hengsdijk ulation and labor market data (Schipper et al., 1998).
et al. (1998), respectively. For some products, the production originating from
In the implementation for the Northern AZ, LUC- the AZ is a large part of domestic supply, viz., bananas
TOR quantifies technical coefficients of various op- (95%), plantain (94%), palm heart (50%) and meat
tions for eight crops and three forestry systems, and (13%), whereas the AZ production is even a consider-
PASTOR for five pastures, two cattle production sys- able part of the world supply for bananas (21%), plain-
tems, and five feed supplements (Table 1). Beside soil tain (21%) and palm heart (11%) (Roebeling et al.,
type, other characteristics on which the differentiation 1999). Therefore, it can be assumed that price forma-
in land units was based include slope and stoniness, tion in the markets for each of these products is sig-
being diagnostic criteria for mechanization feasibility. nificantly influenced by the supply from the North-
The four biophysical sustainability parameters already ern AZ. A method presented by Hazell and Norton
included in the SOLUS applications for the Neguev (1986) was used to endogenize prices in the linear
and Guácimo were augmented with N leaching loss programming model. The procedure consists of lin-
(a possible water pollutant), the amount of N lost by earizing downward-sloping demand curves around ob-
(de)nitrification (as a proxy for emissions of the green- served price and quantity values, using estimated price
house gasses NO and N2 O; Keller et al., 1993), the elasticities of demand, the share of the supply from the
amount of N lost via volatilization (NH3 being a po- region in total supply (domestic or world market), and
tential contributor to acid rain), and the quantities of price elasticities of supply of other suppliers (Schip-
active ingredients in biocides applied. In a later stage, per et al., 1998). In order to adequately model the la-
the emission/sequestration of the greenhouse gas CO2 bor market in the Northern AZ, including competition
was included as well (Plant and Bouman, 1999). Es- for labor between sub-regions, it was assumed that the
pecially the potential for water pollution and the emis- available labor pool in each sub-region can be em-
sion/sequestration of greenhouse gasses are sustain- ployed at a fixed wage. Labor can move freely between
ability issues that various stakeholders in the Northern sub-regions within the Northern AZ but at a so-called
AZ are getting concerned with (Jansen et al., 1995; labor mobility cost calculated on the basis of bus fares
SEPSA, 1997). For example, based on their function between sub-regions. Labor from the non-agricultural
of carbon sequestration, Costa Rican landowners re- sector and from outside the Northern AZ can be at-
ceive (1997 situation) an average of $31 ha−1 per year tracted at increasing cost, i.e., an upward-sloping sup-
from the government for maintaining a forest cover on ply curve for labor is assumed based on a calculated
their land. elasticity of national labor supply (Schipper et al.,
Instead of using the farm as the intermediate level 1998).
between the level of land unit and the entire region, The linear program model for the entire Northern
the Northern AZ was divided into sub-regions to ac- AZ is called REALM (regional economic and agricul-
count for spatial variation in transport costs. Based tural land use model). REALM selects, per sub-region
on presence and quality of roads (Fig. 2 (b)) and ge- and per land unit, the optimal combination of cropping
ographical distances between farm gates and market and livestock options by maximizing regional eco-
outlets, transportation cost models were developed for nomic surplus. For products with endogenous prices,
different classes of products (Hoekstra, 1996). These REALM selects the optimal price-quantity combina-
models were used to stratify the Northern AZ into 12 tions for all products taken together, thereby maximiz-
sub-regions using procedures available in GIS (Fig. ing the sum of producer and consumer surplus (Hazell
2 (c)). Climatic characteristics and major topograph- and Norton, 1986). Optional, normative constraints
ical features are distributed rather homogeneously can be imposed on biophysical sustainability parame-
throughout the Northern AZ, doing away with the ters at the level of land unit, sub-region or the whole
need to include these biophysical characteristics in the Northern AZ.
66 B.A.M. Bouman et al. / Agriculture, Ecosystems and Environment 75 (1999) 55–73

Table 5
Sub-regions with their resource endowments in the Northern AZ application of SOLUS (Bouman et al., 1999)
Sub-region Land units (ha) Labor (person)

SFWa SFPb SIWc

1 63 437 19 711 26 365 13 942
2 9666 14 516 7263 4409
3 1493 1642 812 363
4 276 818 726 183
5 6521 15 991 10 384 3230
6 11 047 41 838 9257 11 485
7 2662 4004 3432 1817
8 563 3552 141 600
9 667 0 950 250
10 4553 13 504 565 1781
11 391 107 33 643
12 1748 0 265 1346
Total 103 024 115 683 60 193 40 049
a SFW: soil fertile well drained.
b SFP: soil fertile poorly drained.
c SIW: soil infertile well drained.

4.4. Scenario development Third, different land use options can be offered
to the linear programming model. For example, in
Technically, land use scenarios can be implemented a long-term explorative study, Bessembinder (1997)
in SOLUS in three ways. offered only alternative land use options that were
First, certain input parameters, such as prices, can sustainable in the sense of having a zero soil nutri-
be changed in the technical coefficient generators. In ent balance (thus maintaining soil productivity over
various SOLUS applications, typical policy-related is- time), and that included hypothetical technologies
sues studied in this manner included the optimal type that may become available in the distant future based
and level of biocide tax needed to induce a desired on the concept of best technical means (Van Itter-
reduction in biocide use (a biophysical sustainabil- sum and Rabbinge, 1997). In shorter term studies,
ity objective) while meeting certain minimum income offered systems may include descriptions of current
and/or employment requirements (economic sustain- systems and calculated alternatives based on techno-
ability objectives); and the role of the price of fertil- logical possibilities that are currently feasible, even
izer in preventing soil degradation in terms of nutri- though they may not (yet) be widely practiced in the
ent losses (Jansen et al., 1997; Schipper et al., 1995, area. Also, completely new land use systems may
1998). be incorporated that are based on prototyping from
Second, normative constraints can be imposed in experimental or on-farm field research. For exam-
the linear programming model with respect to sustain- ple, Bouman and Nieuwenhuyse (1999) explored
ability parameters. In this kind of analysis, typically the possibilities of restoration of degrading pastures
first a so-called base run is evaluated which involves in the Northern AZ through the introduction of a
maximization of economic surplus without any con- grass-legume prototype developed from field station
straints on sustainability parameters. Next, constraints and farmers’ fields research. In a similar manner,
that tighten some of the biophysical sustainability cri- possibilities of precision farming for bananas that
teria are introduced; e.g., a reduction in the use of are currently being prototyped in the Northern AZ
biocides or greenhouse gas emissions, or a limit on (Stoorvogel, 1998) may be explored, and effects
allowed soil nutrient mining (Bouman et al., 1999; on regional economic and sustainability parameters
Schipper et al., 1995). quantified.
 B.A.M. Bouman et al. / Agriculture, Ecosystems and Environment 75 (1999) 55–73 67

5. Results of scenarios with SOLUS for the 5.1. Results for entire Northern AZ
Northern AZ
The results of the above five scenarios are presented
The Costa Rican government has called for the in Table 6. In terms of land use, the actual scenario
execution of research that explicitly analyzes, for simulates current land use in the Northern AZ rather
a range of policy options, the trade-offs between closely. Negative soil N and K balances point towards
socio-economic and environmental goals (SEPSA, soil mining implying, all other things equal, declining
1997). In this context, the capabilities of the SOLUS soil productivity over time. In this sense, most actual
methodology are demonstrated through the evaluation production systems can be considered unsustainable.
of five scenarios related to projected technological The base scenario and the three sustainability scenar-
developments and to relevant policy and sustainability ios indicate that the introduction of new technologies
issues: may result in positive, rather than the traditionally en-
1. Actual. The linear programming model was solved countered negative trade-offs between economic and
by optimizing regional economic surplus, taking biophysical sustainability objectives (i.e., lead to situ-
account of key biophysical boundary constraints ations that will simultaneously satisfy both economic
(i.e., distribution and quantity of land resources and biophysical sustainability). Including alternative
and labor availability) but without any additional land use options in the base scenario (which use in-
normative constraints. Only production options puts more efficiently due to supposed efficiency gains
that are currently prevailing in the Northern AZ in agricultural production; Hengsdijk et al., 1998)
(i.e., that are representative for current farming increases regional economic surplus by 33% while
practices in the region) were offered to the linear simultaneously reducing some of the biophysical
programming model. sustainability parameters as compared to the actual
2. Base. As above, but now alternative production scenario. However, the base run also illustrates a
systems, based on technologies that can be ex- number of trade-offs among the biophysical sustain-
pected to be within farmers’ reach in the medium ability parameters themselves: compared to the actual
term (say, during the next 25 years or so) were scenario, the soil N mining and all N losses to the
also included. environment increased, whereas the soil K mining
3. Soil. The same as the base scenario, but with a and the use of biocides and its environmental hazard
zero soil mining constraint, thus excluding unsus- decreased. Because of the negative soil N and K bal-
tainable land use options with a negative soil nu- ances, most of the selected land use systems in the
trient balance. base run are still unsustainable from a biophysical
4. Environment. The same as the base scenario, but point of view, and yield levels (and therefore eco-
with the constraint that the values of all biophysi- nomic surplus) are likely to decrease over time. To
cal sustainability indicators do not exceed 50% of address this problem, sustainability in the sense of
the values in the base scenario. This scenario ex- closed soil nutrient balances was introduced in the
plores the scope for reducing environmental con- soil scenario, resulting in a 10% increase in economic
tamination as caused by biocide use, N leaching surplus as compared to the actual scenario. Addition-
loss and greenhouse gas emissions etc. ally, biocide use was reduced, though (de)nitrification
5. Soil/Environment. A combination of the above soil losses (proxy for NO and N2 O greenhouse gas emis-
and environment scenarios: in addition to zero soil sions) increased somewhat. Sustainability gains in
mining, biocide use, N leaching and greenhouse the soil scenario are mainly achieved by a decrease
gas emission indicators are not allowed to exceed in cultivated area and changes in pasture technolo-
50% of the values in the base scenario. gies (Bouman et al., 1999). From a comparison of
The SOLUS framework as set-up for the whole the results of the base scenario with those of the soil
Northern AZ was used (i.e., using the technical coeffi- scenario, one may conclude that foregoing the possi-
cient generators LUCTOR and PASTOR, and the lin- bility to achieve a higher economic surplus through
ear programming model REALM); details are given soil mining (admittedly temporary because of ex-
in Tables 1 and 2. pected future yield decreases!) comes at a cost which
68 B.A.M. Bouman et al. / Agriculture, Ecosystems and Environment 75 (1999) 55–73

Table 6
Results of SOLUS for five scenario analyses (see text for explanation)
Scenario Actual Base Soil Environment Soil/environment

Economic surplus (106 US$) 220 292 241 210 189

Labora (%) 75 72 69 39 38
Land useb
Forest (%) 38 38 69 67 90
Pasture (%) 43 43 19 23 2
Crops (%) 19 19 12 10 8
Per crop type:
Pineapple (%) 0.7 0.7 0.7 0.5 0.5
Palm heart (%) 2.8 1.6 1.7 1.2 1.6
Tree plantation (%) 0.0 4.1 0.0 0.0 0.0
Banana (%) 10.0 7.9 8.1 3.9 4.2
Plantain (%) 0.5 0.4 0.2 0.2 0.1
Cassava (%) 4.4 4.6 1.7 4.4 1.8
Beans (%) 0.0 0.0 0.0 0.0 0.0
Maize (%) 0.0 0.0 0.0 0.0 0.0
Number of animals 313 219 410 508 172 356 207 001 39 629
Sustainability indicators
Soil nitrogen balance (kg ha−1 ) −51 −62 0 −29 0
Soil phosphorus balance (kg ha−1 ) 3 1 0 0 0
Soil potassium balance (kg ha−1 ) −113 −20 0 −10 0
Nitrogen lost by (de)nitrification (kg ha−1 ) 15 41 37 20 20
Nitrogen lost by leaching (kg ha−1 ) 90 94 92 47 44
Nitrogen lost by volatilization (kg ha−1 ) 15 24 23 12 12
Pesticide active ingredients used (kg ha−1 ) 12 8 7 4 4
Biocide index (–) 398 264 98 132 25
aLabor employment in the primary production sector as % of the total agricultural labor pool.
bLand use as % of the total of 447 000 ha. The land use class forest includes 38% surface area permanently under forest in protected
areas and on nonsuitable soil, plus the land left fallow.

is equivalent to 23% of the maximum achievable eco- ances with those on the environmental effect indica-
nomic surplus. On the other hand, besides halting soil tors in the soil/environment scenario leads to an even
mining and preventing future yield decreases, the soil further reduction in economic surplus. Constraints
scenario (as compared to the base scenario) also leads on the environmental indicators effectively preclude
to other sustainability benefits including decreases extensive use of fertilizers to satisfy the zero soil
in both biocide use and greenhouse gas emissions. nutrient balance constraint. As a result, a significant
Limiting the environmental indicators in the environ- portion of agricultural land is taken out of production
ment scenario to half of their corresponding values because, without fertilizers, the constraint of zero soil
in the base scenario is relatively costly in economic nutrient balances can not be satisfied. Consequently,
terms (as evidenced by a 28% decline in economic there is scope for forest regeneration, which might be
surplus), even though much less so relative to the Ac- considered an extra environmental gain. The regional
tual scenario. Again, environmental gains are mainly economic surplus stays at a relatively high level due
achieved by significant decreases in pasture area. to the fact that even though crops account for only
The cost of reducing biocide use and (de)nitrification a relatively small part of agricultural land use, they
losses is born to a considerable extent by agricultural are responsible for most of the economic surplus in
laborers who see their employment opportunities cut the area (particularly the cultivation of bananas for
in half. Combining constraints on soil nutrient bal- export).
 B.A.M. Bouman et al. / Agriculture, Ecosystems and Environment 75 (1999) 55–73 69

Fig. 4. Percentage of suitable land area under crops (%) in the base run (a) and in the soil/environment run (b), of the SOLUS implementation
for the Northern AZ.

5.2. Results for sub-regions of Northern AZ tural land generally decreases in most sub-regions in
comparison with the base scenario result. However,
The spatial distribution of results of scenario anal- in sub-regions 1 and 3 (see Fig. 2 (c)), the average
yses are best illustrated with maps created with GIS per hectare application of biocides increases because
(Bouman et al., 1998b). For example, Fig. 4 (a) and of shifts in crop production technologies (data not
(b) present maps for the percentage land used for shown). Figs. 4 and 5 illustrate the geographical di-
crops in the base and the soil/environment scenario, mension of potential conflicts and trade-offs that exist
respectively. In the base scenario, the percentage land between farmers’ interests, who have a strong market
used for cropping is highest in the southern zones orientation (Kuyvenhoven et al., 1995) and have an
and lowest in the northern zones because of increas- interest to crop the available land to maximize eco-
ing transport costs from south to north. While in the nomic surplus, and environmentalists’ interests, who
soil/environment scenario considerably less land is are after minimizing environmental pollution caused
used for crops than in the base scenario, sub-regions by the use of biocides. Moreover, they visualize geo-
9 and 12 (see Fig. 2 (c)), which had the highest per- graphical ’hot-spots’ where local goals (at the level
centage cropped land in the base scenario, have no of the sub-region) may conflict with regional goals
longer any area with crops because of relatively high (at the level of the entire Northern AZ).
transport costs. The latter are caused by relatively
poor infrastructure, despite the close proximity of
both sub-regions to the main road in the Northern AZ 5.3. Results for individual land units in Northern AZ
(see Fig. 2 (b)). Fig. 5 (a) and (b) show the amount of
biocides (in terms of active ingredients) applied, as an Economic and biophysical sustainability trade-offs
average per hectare of used agricultural land, for the at the level of the land unit are illustrated with an ex-
base scenario and the soil/environment scenario, re- ample for fertilized pasture (Estrella, Cynodon nlem-
spectively. Differences among sub-regions are related fuensis) (Fig. 6 (a)) and fertilized cassava (Manihot es-
to land use types and associated technologies (data not culenta Crant) (Fig. 6 (b)). The horizontal axis in these
shown). In the soil/environment scenario, the average figures indicates production (used here as a proxy for
amount of biocides applied per hectare of agricul- economic surplus as well), while the vertical axis in-
70 B.A.M. Bouman et al. / Agriculture, Ecosystems and Environment 75 (1999) 55–73

Fig. 5. Biocide use (kg active ingredients ha−1 used land) in the base run (a) and in the soil/environment run (b), of the SOLUS
implementation for the Northern AZ.

Fig. 6. Calculated biocide use (left axis; black markers) and N loss by (de)nitrification (right axis; white markers) as function of dry
matter production of fertilized pasture (a) and fertilized cassava (Manihot esculenta) (b). Pasture was Estrella (Cynodon nlemfuensis) with
a stocking rate of three animal units per hectare. All data are annual values. Biocide use was derived from field survey data, whereas N
loss was derived from loss fractions obtained from literature and dynamic simulation models (Plant and Bouman, 1999).

dicates the corresponding, calculated, use of biocides ing a trade-off between the economic and biophysical
and N loss via (de)nitrification. For both fertilized pas- dimensions of sustainability. Nitrogen (de)nitrification
ture and cassava, increases in production are accom- losses were higher in fertilized cassava than in fertil-
panied by increases in (de)nitrification losses, imply- ized pasture since higher production levels imply cor-
 B.A.M. Bouman et al. / Agriculture, Ecosystems and Environment 75 (1999) 55–73 71

respondingly higher fertilizer application rates (lead- into the framework. For instance in the Northern AZ
ing to increased (de)nitrification losses). In cassava, implementations, separate studies were made of im-
higher yield levels are accompanied by the use of more portant socio-economic issues such as human health
crop protection agents, and total biocide use increases effects of biocide use (Jansen et al., 1999) and gen-
with increasing production. However, in fertilized pas- der issues in labor distribution (Dijksterhuis, 1998).
tures, biocide use is a decreasing function of produc- After careful deliberation, these indicators were not
tion. This is explained by the fact that the only biocides incorporated into SOLUS because the specific nature
used in pastures are herbicides: at higher levels of pro- of these issues warranted another approach than the
duction, pastures are more competitive against invad- linear programming-based methodology of SOLUS.
ing weeds and therefore less herbicides are needed for So far, SOLUS has been developed and imple-
weed control (Myers and Robbins, 1991). Again this mented mainly by scientists. Regional problem anal-
demonstrates that, next to the well-known trade-off yses to direct the development of SOLUS and to
between the economic and biophysical dimensions of construct scenarios were derived from literature (e.g.,
sustainability, there may also exist trade-offs between SEPSA, 1997) and from interviews with stakeholders
the various dimensions of biophysical sustainability (Wilhelmus, 1998). Results of SOLUS were reported
themselves. in workshops with government representatives and
in meetings with farmers organized in the area. Of
late, however, the concepts and tools of SOLUS were
6. Conclusions judged to be sufficiently well developed to adapt
and apply the framework in an interactive way with
The SOLUS framework is a powerful tool to eval- stakeholders. Researchers and stakeholders, ranging
uate land use scenarios and to illustrate trade-offs from farmers to regional planners, are increasingly
among economic and biophysical sustainability pa- becoming aware of the need for close interaction in
rameters at different levels of scale. Its tools and developing and executing land use studies (Bouma,
models are sufficiently generic to allow implemen- 1997). In 1997, REPOSA accepted an invitation of
tation to other regions with different bio-physical its counterpart the Costa Rican Ministry for Live-
and (socio-)economic conditions. SOLUS, however, stock and Fishery (MAG) to jointly adapt and apply
is not a blue print, and should always be adapted to SOLUS in a less-endowed pilot area on the Pacific
the specific characteristics and objectives of the case side of Costa Rica. An important objective of this
study under consideration. For instance some of the study is the feasibility of using SOLUS as a tool to
biophysical sustainability indicators employed in the aid discussion among regional stakeholders with con-
Northern AZ will not always be relevant for other ar- trasting objectives (planners from different MAG de-
eas and might thus be discarded, while other indicators partments, extensionists, cooperative coffee farmers,
may need to be added (such as erosion hazard and wa- sugar cane industry, water boards and nature conser-
ter conservation in more hilly and drier areas than the vatists). Though successful examples of such an ap-
relatively flat Northern AZ). On the socio-economic proach have been reported for Europe (Bouma, 1997;
side, the sustainability indicators presented in this Rossing et al., 1997), it is a major challenge to see
paper were regional income and employment. An whether it could also work in the Costa Rican setting.
additional important factor in many tropical environ-
ments is that of risk caused by temporal fluctuations
in yields and prices, and farmers attitudes towards it. Acknowledgements
Hazell and Norton (1986) presented several ways of
dealing with risk in linear programming modelling, Many CATIE, MAG en WAU collaborators and stu-
some of which were incorporated and tested in SO- dents over the years have contributed to the devel-
LUS for a Guácimo application (La Rovere, 1997). opment and implementations of SOLUS. Especially
In any attempt to adapt SOLUS for a specific case D.M. Jansen and J.J. Stoorvogel have played a large
study, however, it should be borne in mind that not role in conceptual development and in elaborating the
all issues relevant in land use analysis can be ‘forced’ Neguev and Guácimo cases. L.O. Fresco, S.B. Kroo-
72 B.A.M. Bouman et al. / Agriculture, Ecosystems and Environment 75 (1999) 55–73

nenberg, A. Kuyvenhoven and M.A. Van Ittersum are indicators. In: Finke, P.A., Bouma, J., Hoosbeek, M.R. (Eds.),
thanked for valuable suggestions and continuing sup- Soil and Water Qualities at Different Scales. Kluwer Academic
Publishers, Dordrecht, The Netherlands, pp. 13–22.
port.
FAO (Food and Agriculture Organization of the United Nations),
1993. Guidelines for land use planning. FAO Developments
Series 1, FAO, Rome, Italy, 96 pp.
References FAO (Food and Agriculture Organization of the United Nations)
1995. Planning for sustainable use of land resources. Towards
Alfaro, R., Bouma, J., Fresco, L.O., Jansen, D.M., Kroonenberg, a new approach. Background paper to the FAO’s Task
S.B., van Leeuwen, A.J.C., Schipper, R.A., Sevenhuysen, R.J., membership for Chapter 10 of the Agenda 21 of the United
Stoorvogel, J.J., Watson, V., 1994. Sustainable land use planning Nations Conference on Environment (UNCED). FAO Land and
in Costa Rica: a methodological case study on farm and regional Water Bulletin 2. FAO, Rome, Italy, 60 pp.
level. In: Fresco, L.O., Stroosnijder, L., Bouma, J.,Van Keulen, Fresco, L.O., 1995. Agro-ecological knowledge at different scales.
H. (Eds.), The Future of the Land; Mobilising and Integrating In: Bouma, J., Kuyvehoven, A., Bouman, B.A.M., Luyten, J.C.,
Knowledge for Land Use Options. Wiley, Chichester, UK, pp. Zandstra, H.G. (Eds.), Eco-Regional Approaches for Sustainable
183–202. Land Use and Food Production. Kluwer Academic Publishers,
Bessembinder, J., 1997. Uncertainty and temporal aspects in Dordrecht, The Netherlands, pp. 133–143.
long-term explorations of sustainable land use; with reference Hazell, P.B.R., Norton, R.D., 1986. Mathematical Programming
to the Northern Atlantic Zone of Costa Rica. Ph.D. for Economic Analysis in Agriculture. Macmillan, New York,
Thesis, Wageningen Agricultural University, Wageningen, The 400 pp.
Netherlands. Hengsdijk, H., Quak, W., Bakker, E.J., Ketelaars, J.J.M.H.,
Bouma, J., 1997. The role of quantitative approaches in soil science 1996. A technical coefficient generator for land use
when interacting with stakeholders. Geoderma 78, 1–12. activities in the Koutiala region of south Mali. DLV Report
Bouman, B.A.M., Nieuwenhuyse, A., 1999. Exploring options for No. 6, AB-DLO/Department of Development Economics,
sustainable beef cattle ranching in the humid tropics: a case Wageningen, The Netherlands, 40 pp. + appendices.
study for the Atlantic Zone of Costa Rica. Agric. Syst. 59, Hengsdijk, H., Nieuwenhuyse, A., Bouman, B.A.M., 1998.
145–161. LUCTOR.: Land use crop technical coefficient generator. A
Bouman, B.A.M., Nieuwenhuyse, A., Hengsdijk, H., 1998a. model to quantify crop systems in the Atlantic Zone of
PASTOR: A technical coefficient generator for pasture and Costa Rica; version 2.0. Quantitative Approaches in Systems
livestock Systems in the Humid Tropics; version 2.0. A Analysis 17. AB-DLO-PE, Wageningen, The Netherlands. 65
Users Guide. Quantitative Approaches in Systems Analysis 18. pp. + appendices.
AB-DLO-PE, Wageningen, 59 pp. + appendices. Hoekstra, S.L., 1996. Field research on transportation costs,
Bouman, B.A.M., Nieuwenhuyse, A., Schipper, R.A., Hengsdijk, farm-gate, and farmer market prices in the Atlantic Zone of
H., Jansen, H.G.P., 1998b. An integrated methodology for Costa Rica. Field report 110, REPOSA, CATIE, Turialba, Costa
sustainable land use exploration using GIS. In: Proc. First Int. Rica, 89 pp.
Conf. Geospatial Information in Agriculture and Forestry, vol. Jansen, W.G., 1995. Aggregating economic knowledge for use
II. ERIM, Ann Arbor, USA, pp. 230–237. at national, regional, sector, and farm level. In: Bouma, J.,
Bouman, B.A.M., Schipper, R.A., Nieuwenhuyse, A., Hengsdijk, Kuyvehoven, A., Bouman, B.A.M., Luyten, J.C., Zandstra, H.G.
H., Jansen, H.G.P., 1999. Quantifying economic and (Eds.), Eco-Regional Approaches for Sustainable Land Use and
environmental trade-offs in land use exploration at the regional Food Production. Kluwer Academic Publishers, Dordrecht, The
level: a case study for the Northern Atlantic Zone of Costa Netherlands, pp. 143–166.
Rica. Ecol. Model. 114, 95–109. Jansen, D.M., Schipper, R.A., 1995. A static descriptive approach
Crissman, C.C., Antle, J.M., Capaldo, S.M., 1997. Economic, to quantify land use systems. Netherlands J. Agric. Sci. 43,
Environmental, and Health Trade-Offs in Agriculture: Pesticides 31–46.
and the Sustainability of Andean Potato Production. Kluwer Jansen, H.G.P., Stoorvogel, J.J., 1998. Quantification of
Academic Publishers, Dordrecht, The Netherlands. aggregation bias in regional agricultural land use models:
De Wit, C.T., van Keulen, H., Seligman, N.G., Spharim, I., application to Guácimo county, Costa Rica. Agric. Syst. 58,
1988. Application of interactive multiple goal programming 417–439.
techniques for analysis and planning of regional agricultural Jansen, D.M., Stoorvogel, J.J., Schipper, R.A., 1995. Using
development. Agric. Syst. 26, 211–230. sustainability indicators in agricultural land use analysis: an
Dijksterhuis, H.L., 1998. Female participation in the agricultural example from Costa Rica. Netherlands J. Agric. Sci. 43, 61–82.
sector of the Northern Atlantic Zone of Costa Rica. Field Jansen, D.M., Stoorvogel, J.J., Jansen, J.G.P., 1997. A quantitative
Report 129, REPOSA, CATIE, Turialba, Costa Rica, 46 tool for regional land-use analysis. In: Teng, P.S., Kropff, M.J.,
pp. + appendices. ten Berge, H.F.M., Dent, J.B., Lansigan, F.P., van Laar, H.H.
Dumanski, J., Pettapiece, W.W., McGregor, R.J., 1998. Relevance (Eds.), Applications of Systems Approaches at the Farm and
of scale dependent approaches for integrating biophysical and Regional Levels, Kluwer Academic Publishers, Dordrecht, The
socio-economic information and development of agroecological Netherlands, vol. 1, pp. 399–412.
 B.A.M. Bouman et al. / Agriculture, Ecosystems and Environment 75 (1999) 55–73 73

Jansen, H.G.P., Uytewaal, E., Stoorvogel, J., 1999. Health Schipper, R.A., Jansen, D.M., Stoorvogel, J.J., 1995. Sub-regional
externalities and pesticide use in the Atlantic Zone of linear programming models in land use analysis: a case study of
Costa Rica: an economic evaluation. Ambio, submitted for the Neguev settlement. Netherlands J. Agric. Sci. 43, 83–109.
publication. Schipper, R. A., Jansen, H.G.P., Bouman, B.A.M., Hengsdijk, H.,
Keller, M., Veldkamp, E., Weitz, A.M., Reiners, W.A., 1993. Effect Nieuwenhuyse, A., Saenz, F., 1998. Evaluation of development
of pasture age on soil trace-gas emissions from a deforested policies using integrated bio-economic land use models:
area of Costa Rica. Nature 365, 244–246. applications to Costa Rica. Field Report 136, Program paper 16,
Kruseman, G., Ruben, R., Hengsdijk, H., van Ittersum, M.K., 1995. REPOSA. CATIE/UAW/WAU, CATIE, Turialba, Costa Rica.
Farm household modelling for estimating the effectiveness of 80 pp.
price instruments in land use policy. Netherlands J. Agric. Sci. SEPSA, 1997. [Secretarı́a Ejecutiva de Planificación Sectoral
43, 111–230. Agropecuaria]Polı́ticas del sector agropecuario (revisión y
Kuyvenhoven, A., Ruben, R., Kruseman, G., 1995. Options for ajuste). Ministry of Agriculture and Livestock (MAG), San
sustainable agricultural systems and policy instruments to reach José, Costa Rica.
them. In: Bouma, J., Kuyvehoven, A., Bouman, B.A.M., Luyten, Stoorvogel, J.J., 1993. Optimizing land use distribution to minimize
J.C., Zandstra, H.G. (Eds.), Eco-Regional Approaches for nutrient depletion: a case study for the Atlantic Zone of Costa
Sustainable Land Use and Food Production. Kluwer Academic Rica. Geoderma 60, 277–292.
Publishers, Dordrecht, The Netherlands, pp. 187–212. Stoorvogel, J.J., 1995. Linking GIS and models: structure and
La Rovere, R., 1997. Risk, income and land use in the operationalisation for a Costa Rica case study. Netherlands J.
Atlantic Zone of Costa Rica: an assessment with a linear Agric. Sci. 43, 19–29.
programming model. REPOSA report 125, CATIE/MAG/AUW, Stoorvogel, J.J., 1998. BanMan: a decision support system for
CATIE Turialba, Costa Rica, 93 pp. + appendices. banana management. In: Stoorvogel, J.J., Bouma, J., Bowen,
Myers, R.J.K., Robbins, G.B., 1991. Sustaining productive pastures W.T. (Eds.), Information Technology as a Tool to Assess Land
in the tropics 5. Maintaining productive sown grass pastures. Use Options in Space and Time. Quantitative Approaches
Trop. Grasslands 25, 104–110. in Systems Analysis 16. AB-DLO-PE, Wageningen, The
Nieuwenhuyse, A., 1996. Soils, geology and soil-related Netherlands. pp. 13–22.
sustainability aspects of the perhumid tropical Limon basin, Stoorvogel, J.J., Schipper, R.A., Jansen, D.M., 1995. USTED: a
Costa Rica. Serie Tecnica. Informe Tecnico 272. CATIE, methodology for quantitative analysis of land use scenarios..
Turialba, Costa Rica, 85 pp. Netherlands J. Agric. Sci. 43, 5–18.
NRC, 1996. Nutrient requirements of beef cattle. National Tims, 1995. Responses. In: Bouma, J., Kuyvehoven, A., Bouman,
Academy Press, Washington, D.C., 242 pp. B.A.M., Luyten, J.C., Zandstra, H.G. (Eds.), Eco-regional
Penning de Vries, F.W.T., van Keulen, H., Rabbinge, R., 1995. Approaches for Sustainable Land Use and Food Production.
Natural resources and limits of food production in 2040. In: Kluwer Academic Publishers, Dordrecht, The Netherlands, pp.
J. Bouma, A. Kuyvehoven, B.A.M. Bouman, J.C. Luyten and 113–122.
H.G. Zandstra (Eds.), Eco-regional approaches for sustainable Van de Steeg, J., 1997. A farm typology for the Atlantic Zone.
land use and food production. Kluwer Academic Publishers, Field report 117, REPOSA, CATIE, Turialba, Costa Rica. 44
Dordrecht, The Netherlands, pp. 65–88. pp.
Plant, R.A.J., Bouman, B.A.M., 1999. Modeling nitrogen trace gas Van Ittersum, M.K., Rabbinge, R., 1997. Concepts in production
emissions from current and alternative pastures in the Northern ecology for analysis and quantification of agricultural
Atlantic Zone of Costa Rica, J. Environ. Quality 28(3). input-output combinations. Field Crops Res. 52, 197–208.
Rabbinge, R., van Latesteijn, H.C., 1992. Long term options for Van Latesteijn, 1995. Scenarios for land use in Europe:
land use in the European Community. Agric. Syst. 40, 195–210. agro-ecological options within socio-economic boundaries. In:
Roebeling, P.C., Jansen, H.G.P., Schipper, R.A., Saenz, F., Bouma, J., Kuyvehoven, A., Bouman, B.A.M., Luyten, J.C.,
Castro, E., Ruben, R., Hengsdijk, H., Bouman, B.A.M., 1999. Zandstra, H.G. (Eds.), Eco-regional Approaches for Sustainable
Farm modelling for policy analysis at the farm and regional Land Use and Food Production. Kluwer Academic Publishers,
level. In: B.A.M. Bouman, H.G.P. Jansen, R.A. Schipper, A. Dordrecht, The Netherlands, pp. 43–64.
Nieuwenhuyse and H. Hengsdijk (Eds.), Tools for Land use Van Loon, M., 1997. Characterization of beef cattle farming
Analysis at Different Scale Levels,With Case Studies for the in the Atlantic Zone of Costa Rica. REPOSA Report 127,
Atlantic Zone of Costa Rica. Kluwer Academic Publishers, CATIE/MAG/AUW, CATIE Turialba, Costa Rica. 46 pp.
Dordrecht, The Netherlands, in press. Veldkamp, A., Fresco, L.O., 1996. CLUE-CR: an integrated
Rossing, W.A.H., Meynard, J.M., van Ittersum, M.K., 1997. multi-scale model to simulate land use change scenarios in
Model-based explorations to support development of sustainable Costa Rica. Ecol. Model. 91, 231–248.
farming systems: case studies from France and The Netherlands. Wilhelmus, C.M., 1998. Supporting consensus on land use policy’
Eur. J. Agron. 7, 271–283. the merits and limitations of computer modelling. A case study
in Costa Rica. Field report 131, REPOSA, CATIE, Turialba,
Schipper, R.A., 1996. Farming in a fragile future. Economics
Costa Rica, 45 pp.
of land use with applications in the Atlantic Zone of Costa
Rica. Ph.D. Thesis, Wageningen Agricultural University, The
Netherlands, 282 pp.