1
Collaborative Multi-Agent Modelling to improve Farmers’ Adaptive Capacity
to Manage Water and Migrations Dynamics in Northeast Thailand1
W. Naivinit2, G. Trébuil3, M. Thongnoi4, and C. Le Page3
2
Faculty of Agriculture, Ubon Rajathanee University; Ph.D. student at Chulalongkorn University, Thailand and Paris X – Nanterre University, France;
3
Cirad, UPR Green, Montpellier, F34000 France, and
CU – Cirad ComMod Project, Chulalongkorn University, Bangkok, Thailand;
4
M.Sc. student, Faculty of Agriculture, Ubon Rajathanee University, Thailand
Abstract
Northeast Thailand has the largest rainfed lowland rice (RLR) ecosystem in the kingdom and is notoriously known for its high rate of poor smallholders. The low and unstable rice productivity as a consequence of an unfavourable agro-ecological environment drives resource-poor farmers to migrate and look for more profitable employment. Labour migration is an adaptive strategy practiced by local farmers to cope with climatic uncertainty. But the relationship between labour migration and land and water management on the farms is still poorly documented.
Therefore, we used the Companion Modelling (ComMod) approach to improve the understanding of this key interaction and to reinforce stakeholders’ adaptive capacity to deal with uncertainty linked to water dynamics and labour management in the Lam Dome Yai watershed of Ubon Ratchathani Province. The cyclic ComMod process is made of iterative loops comprising field investigations, modelling, and participatory simulations relying on combinations of Role-Playing Games (RPG) and Agent-Based Models (ABM) used with stakeholders.
In this case study, five ComMod loops were carried out to better understand the problem being examined, stimulate exchange of points of view among stakeholders and enhance the creativity of the participants while lessening the black box effect of computer models. The RPG mainly helped the stakeholders to understand the features, rules and operation of ABM simulations. The ABM helped the stakeholders to better understand their situations and to examine the causes of other players’ actions. After its validation by the users, the ABM will also be used for discovery learning by exploring possible future scenarios selected by the local farmers.
Introduction
The Northeast or “Isaan”, the largest plateau in Southeast Asia, covers one third of Thailand (Fig. 1). The regional topography is undulating with elevation ranging from 140 to 200 meters above mean sea level. This is the most populous region in the kingdom where 21 millions inhabitants belonging to 5.5 millions households reside (NSO, 2005). Around 2.7 millions households are working in the agricultural sector and most of them are resource-poor farmers with a 3.2 ha average size of holding (OAE, 2005).
In Thailand, the total area planted to rice covered almost 20% of the country in 2005. With 42% of the total rice growing area, the Northeast region shared the largest area but harvests only 28% of rice production (Fig. 2). More than 80% of rice areas in the Northeast are in the rainfed lowland rice (RLR) ecosystem. The RLR refers to the level to slightly sloping, bund fields; non-continuous flooding of variable depth and duration; submergence not exceeding 50 centimetres for more than 10 consecutive days; rice transplanted in puddled soil or direct seeded on puddled or plowed dry soil; alternating aerobic to anaerobic soil of variable frequency and duration (Zeigler and Puckridge, 1995).
Key agro-ecological constraints that cause low and unstable rice production are the erratic rainfall distribution and poor soil quality. Though the region has an average annual rainfall comparable to other parts of the country, Northeast farmers still suffer from frequent droughts alternating with floods. Even in the rainy season from June to October, the occurrence of rainfall is
1
2 Fig. 1. Annual rainfall distribution in northeast, Thailand and location of the research site in Lam Dome Yai watershed, Ubon Ratchathani Province.
Northeast (3,229) East (372) South (369) North (2,510) Central (1,192) Northeast (5,307) South (869) East (1,068) Central (4,530) North (7,064)
Source: Office of Agricultural Economics, Ministry of Agriculture and Co-operatives, Bangkok.
Fig. 2. Rice-growing area and production by region, Thailand; 2005 crop year.
still unpredictable. There is high possibility that this region would encounter an early-season drought in June and July when rice seedlings are generally transplanted and a late-season drought in October when rice is in its reproductive phase (Fig 3). Furthermore, soils characterized by low fertility with average organic matter content at 0.85%, and coarse-textured with 90% of sand (Harnpichitvitaya et al., 2000) causing low water holding capacity, worsen the biophysical conditions for agricultural production. These soils, particularly the low water holding capacity, magnify the water availability problem.
Although the water table that commonly rises up close to the surface in August mitigates the effects on the rice crop of the dry spells, a severe flood may occur if exceptionally high rainfall takes place. The groundwater level generally fells quickly after flowering making rice susceptible to the late season drought causing significant yield reduction. Supplementing water during the dry spells is an way to alleviate the drought stress. However, these sandy soils limit the development of irrigation infrastructure. Therefore, irrigated areas of the largest rice growing region of the Northeast cover only 6% of the farmland while approximately 50% of agricultural land of the central region is irrigated system (OAE, 2005).
3 0 50 100 150 200 250 300 350 400 450 mm per 10 days 5th quintile 4th quintile 3rd quintile 2nd quintile 1st quintile PET/2 PET
Fig. 3. Frequential analysis of rainfall and rainfed lowland rice cropping patterns in Ubon Ratchathani Province, lower Northeast Thailand; 1954-2003.
Rice being the staple cereal in the local household food systems, its production contributes only 17-20% of the total cash income of farmers but contributes to reduce household’s expenses. The poverty rate of the Northeast is all time highest compared to other regions as indicated by poverty incidence (Fig. 4). The 1997 economic crisis resulted in an increase of poverty incidence in this region. The high rate of poverty and underemployment in villages drive many young male and female farmers to cities (Fig. 5) causing labour management problem on local farms.
Under the poverty alleviation national policy, several interventions have been undertaken to strengthen grass root economy and reduce poverty incidence. However the success of the Thai government’s attempts to mitigate regional poverty through water resource improvement to increase farm intensification is limited by the undulating topography of the Northeast that are not well suited to large-scale irrigation projects. Very recently, the Thai government planed to take off the shelf a very ambitious and costly project costing up to 10 billion euros to divert water from the Mekong River to supply 19 provinces in the Northeast region through the Chi-Mun system. But the past and recent projects are implemented without prior understanding of relationship between land & water use and labour migration.
How can we better understand the interaction between land & water use and labour migration to improve farmers’ adaptive capacity to manage these agro-ecological and socioeconomic dynamics? Our assumption is that such better understanding could be achieved through co-learning during a process of iterative, collaborative modelling between researchers and RLR growers. Based on a shared understanding of the interaction between land & water use and labour migration, various land & water management scenarios to deal with uncertainty of climatic conditions could be explored with local farmers through series of participatory simulations.
This communication starts with the description of existing knowledge on the problem at the study site. The Companion Modelling (ComMod) and the development of tools used are introduced, and followed by the organization of series of participatory workshops and the selection of participants. Finally, various effects of this participatory modelling and simulation process on farmers’ learning and adaptive capacity described and discussed.
4 7.8 5 1.6 17.6 55.9 16.2 48 44.9 39.2 15.1 1 10 100 1988 1990 1992 1994 1996 1998 2000 2002 2004 %
Northeast North South Central Bangkok
Economic crisis
Source: Statistical Year Book 2006, National Statistical Office, Ministry of Information and Communication Technology, Bangkok.
Note: Thailand measures poverty incidence at household level by comparing per capita household income against poverty line which is the income level that is sufficient for an individual to enjoy the society’s minimum standard of living.
Fig. 4. Evolution of poverty incidence by region of Thailand; 1988 to 2004.
0 200 400 600 800 1,000 1,200
Northeast Central North South
x1000 persons
Source: Labour Force Survey Quarter 4/2005, National Statistical Office, Ministry of Information and Communication Technology, Bangkok.
Fig. 5. Number of out-migrants by region of Thailand; 2005.
Existing knowledge on the problem at the study site
The RLR ecosystem of Northeast Thailand is made of “mini-watersheds” which are identified based on the landform and consist of uplands and upper, middle and lower paddies. Rice-based cropping systems occupy the middle and lower paddies, and some years in more risky upper paddies. Farmers still rely on human-labour at rice transplanting and harvesting due to their poor economic status and the difficulty of using machinery on undulating land and small paddy fields. But labour availability is often inadequate. Under the same climatic situation, various farm management practices are used by heterogeneous farmers to handle such conditions with different labour limitations.
In 2004, a farm survey was carried out at the study site to identify the different main types of household-based agricultural production systems, and to uncover the differentiated set of management criteria and the determining factors of labour migrations among these different categories of farming households (Naivinit and Trebuil, 2004). The current farm types are categorized into three main groups based on the availability means of production, their socioeconomic objectives, and related farming strategies (Table 1).
5 Table 1 . Characteristics o f the main types o f farmi n g households in Ban Mak M ai, D et U dom district , U bon Ratchathani P rovince. Ty p e Fa rm si z e F a m ily s iz e F a rm ed are a per f a m ily labo ur Ob je ct iv e Strategy Su b-ty p e s A1 : Lo w wa te r a v aila bili ty . V e ry po or w a te r a c cess ib ili ty (a v e rag e 5 7 s q .m . of p ond a rea) or pad dies lo cate d in upl and areas. R e c e ive lo w re m it ta n c e (l ess th an 400 e u ro pe r y e a r). Pe rm ane nt a nd sea s on al off -f a rm labo ur. W a te r is sav ed for alle v iate drou ght to st abil iz e rice p rodu c tivity . A2 : Po or acc e ssibil ity (av e rage 15 0 s q .m . of po nd area ). Raise po ultry f o r sale . Receiv e m e dium to hi gh rem ittan c e (av e rage 45 0 euro p e r y ear). Perm anen t off -fa rm labo ur. W a ter is sav ed f o r a llev iate dro ugh t to stabil iz e rice pro ducti v ity . A3: G o od w a ter av ail abi lity becau s e of pon ds (av e ra ge 650 s q .m . of po nd area ), and p addi es lo c a ted i n low e r land. Rec e iv e low rem ittance (le s s tha n 400 e u ro per y ear). Sea s o nal off -f a rm lab our. W a ter in po nds is used e v e n in w e t y e ar to establi s h rice nurserie s . B 1 : Good w a ter a v aila bili ty (aver a g e 7 20 s q .m . of p ond a rea). R e lativ e hi gh re m ittance (900 -1,300 e u ro per y ear) used f o r chil dren e ducation an d h o m e im prov e m ent. P e rm a nen t of f-f a rm labou r. B2 : Ve ry go od w a ter ava ilab ility ( a ve rage 80 0 s q .m . of po nd are a ). No m igrants. Raise live s tock f o r s a le. C1: Uplan d c rop s e.g. ca ss a v a , ma s s pla n ta ti on e.g. cashew nut, p a ra rubb er and fi sh prod uctio n are used to m a xim iz e lab our prod uctiv ity . No m ig rant. C2: Larg e ric e -based , m a rk e t-o ri ent ed farm s w ith re la ti v e hi g h re m ittance of p e rm a nent of f-f a rm labou r (av e rag e 1,5 00 e u ro per y ear). T a k e p e rm a nent of f-f a rm em p loy m e n t or be e n trepre neurs. Ca s h fl ow f a ci liti es Av era ge ann ual gro s s incom e Sha re of of f-f a rm i n com e Sha re of Debt 3 1 % P rodu c e no n-gl utinou s rice f o r sale an d increase land pro ductiv ity th roug h integra te d farm s 2.2 - 4 h a 2,10 0 euro 22% 31% Insuf fic ient c a sh f low f o r any inv e stm ent 1,15 0 euro Grow one o r tw o v a rie ti e s of ri c e , late-m aturing (n on-g lutino u s , KDML1 05) ri c e for s a le , an d earl y -m a tu ring (tradi ti ona l g lutino u s) ri c e f o r fam ily con s um ption. Cattle a nd buff a lo are raised an d s e ll th em in cas e o f c a s h -n eed e m erge ncy s u c h as s ick n e s s of f a m ily m e m b ers. Fis h a nd pou ltry is for f a m ily consum ption an d s a le . Motori z ed f a rm equ ipm e nt is used f o r land an d w a ter m a n agem ent. O ff -farm em p loy m e n t is im po rta n t source of in com e to s u p port on-farm acti v iti es. Fam ily staple food security an d regul ar incom e thro ugh selli ng s u rplu s rice a nd f u ll em p loy m ent of f a m ily labo ur 6 - 8 m e m bers 7.2 h a B : S m al l ri ce-b as e d m a rk et -ori en te d f a rm s w it h in te g rat ed f a rm s f o r sel f-c o n sum p ti o n Lim ited due to th e deb t Grow tw o o r thre e va rietie s of rice, la te-m a turing (n on-g lutino u s , KDML1 05 and g lutino u s , RD 6 ) ri c e f o r sal e , an d earl y -m a tu ring (tradi ti ona l g lutino u s) ri c e f o r fam ily c o nsum p tio n. O fte n hire l la bou rers to harv e s t l a te m a turin g . P rodu c ts of in tegrated farm s a re used fo r f a m ily c o ns u m ptio n a n d se ll su rp lu s. L iv e s to ck is r a is e d fo r s a v in g a nd s a le. W a ter in po nds is used to estab lish rice nu rs e ri e s . 66 % 3.2 h a A: Very s m a ll rice-ba s e d , self-s ub s is tence orie nted f a rm s 0.64 - 1.6 ha 4 - 6 m e m b ers 4 ,9 00 euro 6% 4% C: L a rge m a rk et orie nted f a rm s w ith d iv e rsif ic atio n out o f rice or high re m itta n ce 1.7 - 2 ha 6 - 8 m e m bers 8.6 ha In c re a s e l abou r p rodu c tiv ity throu gh div e rsific a ti on out of rice. G row tw o or m o re v a rieties o f rice, la te -m aturin g (non -glutin ous, K D ML 105 an d glutin ous, RD6) rice f o r s a le, a nd ea rly -m a turin g (tra dition al glu tin ous) ric e for f a m ily consum pti on. W a ter av ail abi lity is v e ry goo d (av e rag e 800 sq.m .). U s u a lly us e fa rm m a c h in es e .g . s m al l tr ac to rs , rice thre s h ers to all e v iat e lab our shortage . W a te r in pon ds i s used to e s tabl ish ric e n u rs e ries. Hig h
6 Rice production practices and water use strategies
Because the incidence of drought is more frequent and serious than flooding, drought escape is the most important mechanism to stabilize the rice yield. Escape is achieved by matching variety duration with rainfall distribution so that crop passes critical development stages (Setter et al., 1995). To cope with the climatic variability, photoperiod-sensitive rice varieties are widely grown because regardless of seeding date, they flower at about the same date after the peak rainfall period has passed and climatic conditions are favourable for the harvested grain to ripen and dry (Mackill et al., 1996). The late-maturing variety such as RD6 and KDML105 (aromatic jasmine rice) are grown in the lower and medium paddies to ensure adequate water availability before harvest in late November while the early-maturing varieties are usually planted in the upper paddies and are harvested in late October. Regarding the use of supplementary water from farmponds, type A farmers (very small holders) are not likely to use water in wet years even if its quantity is sufficient. The use of water is carefully planned since the main purpose of having ponds is to store water and use it to alleviate drought effects and to feed fish. Unlike type A farmers, type B and C farmers often use water from ponds to produce their rice seedlings and avoid delayed transplanting.
Labour management and migration in relation to RLR production and farm types
Farmers belonging to type A play an important role to supplying labour to the community because their land per labour ratio is relatively low. However, it is risky and unsecured to mainly rely on the off-farm employment. Therefore, seasonal migrants who return to help their family to grow rice are commonly found for this farm type.
A key constraint of type B and C farms is labour shortage. These farm types play a major role in employing hired labour during the periods of peak labour demand in particular rice transplanting and harvesting. Migrants generally do not return home at the beginning of rice season but they remit money back to the village so that their families can use it to hire additional labour. These farm types have relatively high capability to invest in agricultural production through the saving assets (livestock herd) and high remittances.
The diversity of farming systems plays a major role in supporting local labour availability at the community level. During the 6 months of the dry season, most of the young labourers migrate to off-farm jobs to increase household incomes. The labour migration is seen as an adaptive strategy practiced by local farmers to cope with the uncertainty of rainfall and its distribution and the low productivity of their RLR cropping system (2-3 t/ha of paddy and only one crop cycle per year). Therefore, it is important to understand the interaction between this RLR ecosystem and labour management across different farm types prior to initiate any development.
Companion Modelling (ComMod)
The ComMod approach stems from adaptive management for collective ecosystem management. The ComMod aims to strengthen local stakeholders by improving their adaptive management capacity with regard to their renewable resources management (Bousquet and Trebuil, 2005). It also aims at the collaborative group model building integrating various stakeholders’ perceptions to develop a shared representation of the systems to be managed and of the problem to be examined the platform of collective learning. For this research, the ComMod is used to better understand the interaction between water dynamic and labour management of the RLR ecosystem in Ubon Ratchathani Province, lower Northeast Thailand by facilitating dialogue, shared knowledge, and collective learning with farmers. For the ComMod, the human-environment interaction is investigated through the stakeholders’ decision-making process, and assessed through the computer simulation based on the Multi-Agent Systems (MAS).
The MAS emphasizes the investigation of the emergent properties as a consequence of the interactions among agents and between agents and their environment. The MAS is based on the idea that it is possible to represent the behaviours of entities (agents) active in the world in a computerized form, and that it is possible to represent a phenomenon as the outcome of the interactions among an assembly of entities with their own operational autonomy (Ferber, 1999).
7 The ComMod requires a continuous, evolving and iterative confrontation between theories and field circumstances. Therefore, it is based on repetitive back and forth steps between the model and the field situation to comprehend interactions between ecological and social dynamic in complex systems (Barreteau, 2003). The combination of two complementary tools, Role-Playing Game (RPG) and Agent-Based Model (ABM)—for the creation of the MAS is used to capture such interactions through the co-learning with stakeholders during the cyclic ComMod process.
Materials and methods
Conceptual model formalization
The preliminary findings of the field survey integrating with interdisciplinary scientists’ point of views were used to formalize an initial conceptual model by building a set of Unified Modelling Language (UML) diagrams. Based on these UML diagrams, the RPG were designed and implemented to be used in series of participatory modelling and simulation workshops. Fig. 6 shows conceptual model in an UML Class diagram. The main spatial component is land use types incorporated with vertical arrangement of tanks: ponding tank, root zone tank, subsoil tank, and water storage tank. All tanks have their own hydrological processes such as infiltration, percolation and runoff. Another important component is the “Household” rule-based agents that functions as a decision-maker regarding rice production and labour migration.
Description of the Role-Playing Games (RPGs)
The RPGs were designed and constructed based on the initial conceptual model and its gradual modification after each field workshops giving the opportunity to the participants to request changes in the representation of the system. The use of RPGs was also to facilitate mutual recognition and social validation of the representation of rice production and labour migration by different farm types under different climatic situations and water availability conditions. Moreover, the RPGs were used to improve local farmers’ understanding of the interaction under study and to acquaint them with the structure and rules of ABM.
Fig. 6. UML Class diagram showing the structure and the key components of RLR production system in Det Udom District of Ubon Ratchathani Province, Northeast Thailand.
RPG sessions were based on simplified sequences of actions by the participants playing their own role (farmers). The scheduling of actions and some external information like rainfall conditions were provided by the facilitator during the RPG sessions. One round of the game represented one crop year and included three steps; i) RLR crop establishment, ii) Rice harvest, and iii) activities during the dry season including migration.
The RPG session in July 2004 started with the allocation of rice-growing areas for two varieties, KDML 105 (for sale) and RD6 (for self-consumption) on the 2D game board representing
8 individual farms (Fig. 7a). In the 2nd step, the players were allowed to find labourers to transplant rice from a job broker played by a research assistant. Then the players had to identify where to allocate their family labour on the map of Thailand when they moved into the 3rd step. The 4th step was purposely set to let players migrate as they wished. In the final step of a gaming round, they received fake money on the computation (in an Excel spreadsheet) of their annual rice production and sales. 4 rounds of play with different rainfall conditions were completed. The second RPG session in April 2005 was similar to the first one but new migrant players who just returned home could participate. Furthermore, this time, we introduced to the participants a computer simulation “playing the game” as a replay version of the RPG session during the plenary discussion.
The last RPG session in October 2005 was organized to specifically address the issue of farmers’ perceptions of water availability. First, the players were requested to delimit rice growing area for KDML105 and/or RD6 on a “grid” where 1 square was equal to 0.16 ha (or one “rai” the Thai land unit). To simulate rainfall for each successive week starting in April, a drawing card was revealed by the facilitator. Each player had then to decide whether or not they started to work in their paddy fields. If so, he/she had to draw a cross on the corresponding week on his/her form recording his/her decisions (Fig. 7b). Every time a player decided to use water from his/her pond, he/she had to draw water levels in both pond and paddy fields, before and after using water. Finally when all players had announced the end of the crop establishment phase, each of them had to allocate family members in one of the 4 following categories: farm labour; returned migrants; dependent; migrants in cities. Once it was time to harvest, the players had to allocate family labour and decided to hire labour by negotiating with other players and/or asked the facilitator to hire labour from other villages. For the dry season activities, the players decided to allocate family members as they did in the crop establishment phase. Two scenarios were set according to their actual situations: availability of individual ponds or a community pond. 6 rounds of game were played.
The decisions made by the players were recorded and analyzed during individual interviews and plenary discussions after the gaming sessions to clarify the players’ decision-making processes and actions. The results were used to enrich and modify the initial conceptual model and then to construct the ABM.
Fig. 7. 2D game board used for the 1st and 2nd RPG (a), and an abacus to record players’ decisions used in the 3rd gaming session (b).
Co-construction of the Agent-Based Model (ABM)
The ABM called “BanMakMai” was co-designed and built with local farmers to represent the human-environment interactions regarding water dynamic and labour management of farmers working in Ban Mak Mai Village, a typical RLR production area of lower Northeast Thailand. The model is built under the CORMAS (COmmon-Pool Resources Multi-Agent System) platform, which is a programming environment dedicated to the creation of Multi-Agent systems, with a focus on resource management and is using the Smalltalk object-oriented language (Bousquet et al., 1998).
The BanMakMai model integrates knowledge obtained from the farm survey and previous participatory workshops. The model consists of five main interacting components; Climate, Hydrology, Spatial setting, Household and Rice (Fig. 6). The spatial settings represent land uses; paddies, pond, and human settlement (house, village, city). A pixel (smallest unit) is equal to 0.16 ha. Each pixel has its elevation and soil type as attributes. Pixels are aggregated into a rice field. Fields are aggregated into farms with houses. Two small farms (3.3 ha), and two large farms (6.5 ha) with different size of farmpond are displayed (Fig.8). The spatial entity integrates a hydro-dynamic module
9 that was developed to simulate the availability of water in paddy fields and ponds depending on daily rainfall patterns (Lacombe and Naivinit, 2005). The actual daily rainfall and potential evapotranspiration (PET) recorded by the meteorological centre in Ubon Ratchathani are used to provide climatic values to the BanMakMai model to feed the hydrological processes.
A “Household” is made of heterogeneous “Member” agents having different demographic characteristics (age, gender, and marital status). The “Household” is a central decision-maker responsible for assigning specific roles (farmer, seasonal migrant, more permanent migrant or dependent) to its “Member”. Once rice is planted, it will grow from seedling stage to maturity. Three main decision-making processes are i) decisions during nursery establishment and transplanting, ii) decisions at harvest, and iii) decisions after rice harvesting, including migration.
Organization of series of participatory workshop and selection of participants
Five loops of the cyclic ComMod process were implemented for this research since 2004. The successive sequences of key activities are shown in figure 9. Each loop addresses specific objectives to serve the model’s co-design purpose. To achieve this purpose, the joint use of RPG and ABM tool was adopted to work with local farmers (Table 2).
The players were carefully selected to cover all farm types based on the typology of farming systems in this agrarian system built in the initial farm survey. 11 farming households were invited to participate, including three large farming households belonging to type C (1), and type B (2) plus eight households belonging to type A took part as they are the most frequent type in this area. These same households participated through the five loops of the ComMod process. But some members from the same family were reshuffled when the former participants could not join again.
Fig. 8. Spatial interface of the BanMakMai model displaying land use for each of the four farms, the village and city.
3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 910 11 12 1 2 3 4 5 6 7 8 910 11 12 1 2 3 4
0 Problem identification 1 Review/assess existing knowledge 2 Sensitizing activities 3 Survey 4 Model conception 5 Model implementation 6 Calibration/verification 7 Validation 8 Scenarios
Legend: Field activity Laboratory activity Mixed activity
Month Month 2006 2007 2008 Month Year 2004 2005 Step Month Month
Fig. 9. Successive sequences of main activities in the Companion Modelling process implemented in the Lam Dome Yai watershed of Ubon Ratchathani Province; 2004-2008.
10 An activity to evaluate the learning effects of farmers participating in the ComMod process in the Lam Dome Yai was carried out after three successive participatory workshops held in Ban Mak Mai village (2006-2007). The evaluation consists of producing transcripts based on individual interviews with participants, and analysis of the contents of the interviews. The analysis was based on the coding of the transcripts.
Results
Table 3 and 4 display the different kinds of effects generated by the ComMod activities on the different types of participating farmers. We show only the results obtained from the sub-group of 7 out of 11 farmers who were the most at ease during the gaming and computer simulation activities: 4, 2 and 1 for type A, B, and C respectively.
Table 3 shows that in the case of type A resource-poor farmers, the kind of new agro-ecological knowledge gained from the collaborative modelling process deals with the effects of rainfall and water availability on the pattern of rice production. Even if technical topics are not explicitly examined in this process, the exchanges taking place among the participants seems to lead to the acquisition of new knowledge in this field as well, for example on direct seeding of rice and vegetable production after rice. We know that these are the small farms providing under-employed workforce to other larger holdings or migrants looking for wage-earning jobs, therefore it is not surprising to see that they found interest in getting a better understanding of the pattern of labour migration in the village during the ComMod process.
In term of changes in personal perceptions of the system represented in the models, most of the type A farmers would like now to get a better understanding between the various commodities they can grow, especially if more water is available (such as vegetable production after rice, or mixed farming systems around farm ponds), and the relevant markets. They also think that they are now better prepared to face drought events and have learned how to plan rice establishment and the use of agricultural water from their small ponds. They mention three interesting opportunities that emerged from this learning process. The first one is the possibility to diversify their agricultural production out of RLR into vegetable cash cropping after rice and mixed farming. Another one deals with the possibility to exploit underground water to increase the volume of the resource available for farming activities; again, this was not explicitly discussed with the research team during the field workshops but from more informal exchanges among the participants during and after these short events. This is because they also recognize here the usefulness of sharing ideas and knowledge among farmers to generate such opportunities. It seems from these results that this has already an effect on the way some of them interact with others. While others claim that the learning cum modelling process has already led to some changes in their farm practices (go back to RLR production, adoption of direct seeding and vegetable production).
Table 4 show the same kind of results but this time for the larger type B and C holdings which are the most affected by the lack of labour during the peak demand periods of RLR transplanting and harvest. The topics they cite regarding the new agro-ecological and technical knowledge gained from the collaborative modelling process are similar to the ones already mentioned by type A farmers. But due to their labour constraints, it is not surprising to see them mentioning the effects of a shortage of labourers on the economic results of RLR production. Regarding changes in personal perceptions of the system, like the type A farmers, they emphasize their wish to better understand the market opportunities for their farm products. The type B farmers also mention their better ability to plan their farm operations, and one of them appreciated the opportunity to merge their farmer empirical knowledge with the more academic agricultural one from the research team. Like for type A farmers, these three larger favourable farms seem to be ready to increase and diversify their agricultural production if water is made available. But for the time being, no effects related to change of behaviour or farm practices could be noticed in their case.
11
Discussion
The various types of effects generated by the ComMod activities on this sample of participating farmers confirm the hypothesis that such an interactive collaborative modelling process can trigger both individual and collective learning through the intensive exchanges facilitated and stimulated by the evolutionary gaming and simulation models used along the process. Even at this stage, some farmers mention limited change in their decision-making, behaviour and farming practices that they relate to their participation in the ComMod activities. We must also say that this is valid for the research team too because its members learned a lot from these exchanges, especially on farmer decision-making processes related to the adaptation of RLR production practices to rainfall distribution, and the management of the labour force on the different types of farming households.
But it was observed that not all the representatives from the eleven farming households could be active participants feeling comfortable in the simulation workshops. While the gaming sessions where more inclusive, the participatory simulation ones were obviously more difficult to follow for some of the older, male or female, farmers. More attention should be given to this aspect when selecting the participants in this type of community.
This relatively long process could seem costly, both in term of time and funding, but it is because the team leaders in the field are Ph.D. and M.Sc. students who are still on the learning curve about how to design, implement, monitor and evaluate a ComMod process, and how to conceive and implement ABM models. For example, the long periods of time separating each of the successive field workshops were due to the unavailability of the team leader while he was tight by academic duties, or was spending time to master the modelling and simulation platform.
Conclusion
The ComMod process enhances our understanding regarding the interaction between land & water use and labour migration at the study site through its evolving iterative, collaborative modelling. The farm diversity plays an important role in labour availability at village level. The interaction across farm types is clearly seen during wage negotiation. Type A farmers who mainly grow small-scale rice production areas with less farm diversification take part in the local labour system as hired labour, and they seasonally migrate for off-farm jobs. In contrast, the larger farms belonging to type B and C play a key role in employing hired labour due to their high land and labour ratio. Moreover, through the co-learning and knowledge sharing during series of participatory workshops, participants’ adaptive management capacity was generated. Better understanding the relationship between water availability and rice production was gained leading to be better prepared to face droughts with the use of ponds. New ideas, for instance mixed farming and the use of underground water, were emerged. Some participants even changed their farming practices as a result of learning process through collaborative modelling.
Further steps
The following steps will be made of a series of participatory simulation sessions to (i) validate the most recent version of the ABM models, (ii) identify scenarios dealing with water and labour use that the farmers are interested to explore together with the research team, and (iii) simulate these scenarios and assess their results with the different types of farmers.
12
Table
2. Successive tool
s used in Companion Mod
elli n g pro cess implemented in La
m Dome Yai wate
rshed, Ubon Ratch at h ani Pr ovi n ce. T ool RPG 1 R PG 2 an d ABM 1 R PG 3 ABM 2 C o lle c ti v e illu s tr a ti n g al gori th m s of rule -bas ed age nt ABM 2 ABM 2 Date 9-10 J u ly 2005 20-21 Apri l 2006 10-1 1 Oc to ber 2006 24 Apr il 2 0 06 5-6 Augus t 2 0 07 5-6 Febr u a ry a n d 19 -2 0 Mar c h 2008 8-9 April 2008 Objec ti v es T o val idat e the res earc h team un der s tandi ng of t h e in ter a c ti on bet w een land/w a ter us e and labou r m anagem ent on the dif ferent ty pe s o f fa rm s. T o inv e s ti gate t h e i m pac t of ex c e pt ional pr ol ong ed drought an d ac ces s to an ir riga ti on c anal on p lay ers ’ fa rm m anage m ent an d to pres ent the pl ay ing gam e c o m put er sim u la ti on fo r c o lle c tiv e l earni ng. T o ac qui re the k now led ge of pl ay ers ’ w a ter perc epti on acr o s s f a rm ty pes a nd to pr ov id e the pl ay ers an in te rm edi ate s tage of l e ar ni ng s im u la ted c o n d iti o ns and s c e narios thr o u gh RPG. T o val idate the c o m p re h ens iv e pr oc e s s of in ter a ct io n betw een w a te r dy na m ic s and l a bour m igration by us ing the ABM s im u lat ion. T o des ign th e m o del w ith fa rm er s by us in g the d raw ings tr a n s lat ed f rom dec is ion-m a k in g al gori thm s of rule -b a s e d agents a s a t o ol f o r col lec tive validat ion . T o v a lid a te th e L a m D o m e Y a i Ag e n t-Ba s e d Mo d e l (L DY m o d e l) regardi n g ri c e -gr o w ing prac ti c e s and la bour m a nagem en t in c ludi ng m ig ratio n w ith fa rm e rs . T o us e the f inal L a m Dom e Y a i Agent-B as ed Mo d e l ( L D Y m o d e l) to ex pl ore d e s irabl e sce n a ri o s w it h fa rm e rs. Know led ge p roduc ti on • Enric h m ent of the in it ia l c onc eptual m odel regar d ing fa rm er’ s de c is ions ab out f a rm m anagem e n t w hen di ff er en t c lim at ic c o nd it io n s . • T he div e rs it y of fa rm in g sy ste m p la y s a m a jo r ro le i n s u p por ti ng l o c a l labour m a rk et s inc e di ff e rent fa rm t y pe s hav e di ff er en t c ropp ing c a le n dar and f a rm s iz e . • Sm a ll h o ld er s we re m o re adap ti ve to tak e adv antage o f bet ter a cce ss t o w a te r w h ile th is had l it tle i m pac t on m e d ium an d l a rg e f a rm hol di ngs . • T he RPG h e lp s the fa rm ers to unders tand the ru les and s equ enc e of ABM s im u lati on w h ile the A B M he lp s t o be tt e r unde rs tand s e lf -s it uation and ex am in ed c a us es of ac ti ons of ot he r p lay ers . • Fo r v e ry sm a ll h o ld e rs, wa te r in i n di vi d u a l fa rm po nds w a s m a in ly c o n s er ve d to be u s ed o n ly in c a s e of a v e ry dr y y ear to reduc e r ic e -y iel d l o s s c a us e d by d rou g h t. • A c o m m uni ty pon d s c e nar io ef fe c ti v el y provok ed the in c reas e of fa rm i n te n s if icat ion of very s m all -f a rm ing hou s ehol ds . • T he hy dr ol ogi c al proc es s e s i m pl em en ted in the m o del s e em t o be s u ff ic iently val id ated to re p res e n t wa te r le ve l of pa ddi es c o m pos ed w it h sa n d y so il. • R ice -n u rse ry est abl is hm en t is bas e d on the r e la tio n s h ip b e tw ee n topos e quenc e and av ai la bil it y of w a ter . Other rice-g row ing a c ti vi ties (t rans pl antin g , harves tin g etc .) depend on t h e rel a ti ons h ip betw e en w a ter an d labou r avai la bi lity . • Al gorithm s of agent’ s d e c is io n -m ak ing r u le s ar e sim ila r to the ac tual f a rm ers ’ p ra c ti ce s. • T h e al gor it hm s of dec is ions to hi re la bour and m igr a ti o n is st ill not tr an s par en c y to the fa rm e rs. T he LD Y m o del w a s c a lib ra ted t o m a tc h t h e ac tual rice-g row ing pr a c ti c e s purpos ed by fa rm ers duri ng the w o rk s h op. But ther e w a s no m a jor di ff e renc e betw een dec is ions m ade by rul e -b a s ed c o m p uter agen t and f a rm er s . Part ic ipants R e se a rch te a m , 1 1 fa rm ing hous e hol ds , an agri c ultural ex tens ion of fi c e r, a NGO pers onnel , and lo c a l a d m in is tr a ti v e of fi c e rs . Sam e as the RPG 1 ex c ludi ng a NG O pers onnel . Sam e as the RPG 1 ex clud ing NGO per sonnel and an agric u ltur a l ex tens io n of fi cer . Sam e a s the R P G 3. Resear c h team . Farm er s w e re d iv id ed i n to 3 s m all gr ou ps b a s e d on f a rm ty p e s a n d they w e re i n v ite d t o p a rt ic ip a te separ a tely . S a m e a s th e co lle ct iv e ill u s tr a ti ng al gor ithm s of ru le -bas ed agent. Sam e as the RPG 1 .
13 F a rm t y p e A1 A2 A 3 A 4 1. Kn o w le dg e acq u isi ti on Agr o -e c o logi c a l • W a te r a v a ilabi lit y in re la tio n t o loc a tio n of paddy f ie lds • Chan ge of w a ter le v e l i n f ie ld/ pond depe ndin g o n r a inf a ll • W a te r a v a ilabi lit y in r e la ti on to loc a ti on of padd y f ie lds • I m pa c t of r a inf a ll d is tr ibut ion on r ic e pr o duc ti on • I m pa c t of dr oug ht on r ic e p roduc ti on T e c h n ic a l • Us e of dir e c t s eedi ng in v e ry dr y y ear • Dif fe re nt r ic e -g ro w ing p rac ti c e s bet w een upp er a nd low e r pad dies • Us e of pond t o gr ow r ic e • Vege tab le pr odu c tio n fo r s a le in dr y se aso n • Dr o ught m it igat ion b y h a ving m o re po nds to s tor e w a te r • In tegr at ed f a rm ing pr ac ti c e Ec onom ic • Benef it s of b e tt er ac c e s s t o w a te r • How t o s h ar e land be tw ee n RD 6 ( fo r fa m ily c ons um pt ion) and KDML1 05 ( for sa le ) rice v a rie tie s Soc ial • Lab our m igr ati on s itu at ion in th e vil lage • T oge th er nes s am ong par ti c ip ant s a s a re sul t o f co lle ct iv e d iscu ssi on s • La bour m igr at ion pat te rn s in th e vil lage 2 . C h a nge i n pe rc e p ti on/ Ow n W a n t to i m pro v e • B e tt e r e s ti m a te o f fa rm i n co m e • Unde rs ta nding t he r e la tio n s h ip be tw ee n fa rm pr od uc ts a nd m a rk ets • Und e rs tandin g t h e r e la ti ons hi p b e tw e en fa rm pr o duc ts a nd m a rk et s • Un der s tandi ng the r e lat ions h ip bet w een fa rm p roduc ts and m a rk et s • W ant to bu ild a pond Cap a c it ies • Bett er pr ep ar ed t o f a c e d rough t • T im in g of w a te r pum ping f rom p onds f o r high er benef it • Plan f o r r ic e tr ans pla n ti ng • Bet ter pr e par ed t o f a c e dr oug ht • Ne got iat ion is nee ded to avoi d w a te r sha ri n g co nfli ct O ppor tunit ie s • Pos s ibil ity t o gr ow veget abl es a fter r ic e • Unde rg roun d w a te r m o re s u it able t han ir ri ga tio n c anal • Shar ing m y ideas w it h oth e r f a rm er s • Mor e inc o m e c ould b e m ade du ri ng th e dr y s e a s on • Shar in g k now le dge is bet te r t han us ing onl y o w n o n e • Unde rg roun d w a ter is s u it ab le fo r m y fa rm • In tegr at ed f a rm ing pr ac ti c e • Pond an d under gr ound w a te r is s u it abl e fo r my fa rm 3 . C h a nge i n pe rc e p ti on/ Othe rs • Bett er s o c ia l n e tw o rk in g i n t he villa ge • Shar ing k now led ge and ide a s helps t o un der s tand ot her par ti c ipant s ' f a rm m anage m ent s k ills 4. Ch an g e in d e c isi o n -m aki ng • Mor e f a m ily l abour t o w o rk on-fa rm if m o re w a te r is availa ble • W ill pr od uc e m o re f a rm c o m m odit ies if m o re w a ter is avail able • Mor e f a m ily labo ur t o w o rk on -f ar m if m o re w a te r avail able • Buil d a ne w po nd or im pr ove t he ex is tin g on e to s tor e m o re w a te r • Mor e f a m ily labo ur to w o rk on -f ar m if m o re w a te r is av ai la b le 5 . C h a nge i n be ha v iour • Mor e s har in g o f fa rm ing ide a s w ith ot her f a rm e rs • Spend m o re ti m e to w o rk i n p addi es 6. Ch an g e in a c ti on ( s ) • Re tur n ed to r ic e p roduc ti on in s tea d of lea s ing t he land t o neigh bour s • Sta rte d to gr ow veget abl es in dr y s eas o n • Us e d ir e c t s e eding t e c h n ique f o r ri c e pr o duc ti on Table 3. L earning effect s of small and resour ce-poor f ar m ers (type A) parti cip at ing in the ComMod collaborat ive modelli n g
process in the Lam Dome
Yai waters hed, U bon Rat cha thani P rovince in 200 6-2007.
14 F a rm t y p e s B1 B2 C 1 . K now le dge a c qui s ition A g ro-e cologi c a l • Relat ionship b e tw e en r a in fa ll and w a ter level in fi eld/p ond • Dur a ti on of r ice gr o w ing c y c le in r e lation to w a ter av ailabi lit y • T he d if fer ence b e tw e en sandy and c la y ey s o ils in rel a tion to w a ter av ailabi lity in pa ddy f ield s • Rela tionship betw e en r a inf a ll distr ibut ion an d r ice pr o duction T e chnical • Rice tr an s p lantin g pr actice in r e lati on to fi eld lo c a ti on • Usef u lness of h a ving pond s an d tim e to use w a te r f rom them • Rice pr o duction in r e la tion to w a te r dy n a m ics i n pad dy f ield/ pond depen ding on r a inf a ll E c ono m ic • Lo w r ice p roduction because of lack of labou r to loo k af ter r ice • Im p a ct of labou r s h or tage on f a rm pr odu c tio n • L e s s incom e f ro m r ice is expected if dr ou g ht occur s S o cial • La bour m igr atio n patt e rn s in th e vill age 2 . C h a nge in pe rc e p ti on/Ow n W a nt to i m prove • B e tte r estim a te of fa rm incom e an d inves tm e nt • Unde rs ta nding the r e lation s h ip betw een f a rm pr odu c ts an d m a rk e ts • W a n t to un der stand r ic e tr adi ng sy stem • Unde rs tandin g the relatio n s h ip be tw een f a rm pr od uc ts an d m a rk e ts Cap a c ities • T im ing of w a ter pum pin g f rom pon ds fo r h igher benef it • M a k e an nual f a rm o per at ion pl an Opp o rt uni ties • Le ar n to m e rge ac a dem ic agric u ltur a l k now le dge w ith o w n expe ri ence 3 . C h a nge in pe rc e p ti on/Othe rs Cap a c ities • Nego tiatio n pr ocess r egar d ing w a ter sh a ri n g Opp o rtuni ties • Obser v e and analy z e the ac ti ons of other f a rm er s w hose m e ans of pr odu c ti on ar e s im ila r 4 . C h a nge in de c is ion-ma k ing • W ill p roduce m o re f a rm com m o d ities if m o re w a te r i s a v a ila b le • M o re f a m ily labo ur to w o rk on -f arm if m o re w a te r i s a v a ila b le • W ill pr odu c e m o re f a rm com m o dities if m o re w a te r is a v a ila b le Table 4. Learn ing effe cts of m
edium sized (type B) and
la rge (type C) f armers partic ipati n g in t h e Com M od collaborative modelli ng p rocess in t h
e Lam Dome Yai waters
h ed, Ubon Rat chathani P rovince in 20 06-20 07.
15
Acknowledgements
The authors would like to thank the Challenge Program for Water and Food (CPWF) and Echel-Eau Project for their financial support.
References
Barreteau, O. 2003. Our companion modelling approach. Journal of Artificial Societies and Social Simulation 6(1): 7. [Online] Available from: http://jasss.soc.surrey.ac.uk/6/2/1.html
Bousquet, F., Bakam, I., Proton, H., and Le Page, C. 1998. CORMAS: common-pool resources and multi-agent systems. In International Conference on Industrial and Engineering Applications of Artificial Intelligence and Expert Systems. Benicasim (Spain). Berlin (Germany): Springer-Verlag. Bousquet, F., and Trebuil, G. 2005. Introduction to companion modeling and multi-agent systems for integrated natural resource management in Asia. In F. Bousquet, G. Trebuil, and B. Hardy (eds.), Companion Modeling and Multi-Agent Systems for Integrated Natural Resource Management in Asia, pp. 1-20. Los Banos, the Philippines: IRRI.
Ferber, J. 1999. Multi-agent systems: an introduction to distributed artificial intelligence. New York: Addison-Wesley Longman. 499 p.
Harnpichitvitaya, D., Trebuil, G., Oberthur, T., Pantuwan, G., Craig, I., Tuong, T. P., Wade, L. J., and Suriya-Arunroj, D. 2000. Identifying soil suitability for subsoil compaction to improve water- and nutrient-use efficiency in rainfed lowland rice. In T. P. Tuong, S. P. Kam, S. Pandey, B. A. M. Bouman, and B. Hardy (eds.), Characterizing and understanding rainfed environment pp. 97-110. International Rice Research Institute, Los Banos, Laguna, Philippines.
Lacombe, G., and Naivinit, W. 2005. Modeling a biophysical environment to better understand the decision-making rules for water the use in the rainfed lowland rice ecosystem. In G. Trebuil, F. Bousquet, and B. Hardy (eds.), Companion Modeling and Multi-agent system for Integrated Natural Resource Management in Asia, pp. 191-210. Los Banos, Philippines: IRRI.
Mackill, D. J., Coffma, W. R., and Garrity, D. P. 1996. Rainfed lowland rice improvement. Manila: International Rice Research Institute. 235 p.
Naivinit, W., and Trebuil, G. 2004. Interactions between water use and labour migrations in northeast Thailand: Context, methodology, and preliminary findings. In 4th International and Interdisciplinary Seminar of the Common Program on Irrigated System (PCSI): Hydraulic Coordinations and Social Justice, 25-26 November 2004. Agropolis, Montpellier, France.
NSO, 2005. Key statistics of Thailand. p. 140.City: National Statistical Office (NSO), Ministry of Information and Communication.
OAE, 2005. Agricultural statistics of Thailand, crop year 2004/2005. p. 151.City: Office of Agricultural Economics (OAE).
Setter, T. L., Ingram, K. T., and Tuong, T. P. 1995. Environmental characterization requirements for strategic research in rice grown under adverse conditons of drought, flooding, or salinity. In K. T. Ingram (ed.), Rainfed lowland rice: Agricultural research for high-risk environments, pp. 3-18. Manila, Philippines: International Rice Research Institute (IRRI).
Zeigler, R. S., and Puckridge, D. W. 1995. Improving sustainable productivity in rice-based rainfed lowland system of south and southeast Asia. GeoJournal 35(3): 307-324.
