Impact of On-farm Water Storage for Increasing Agricultural Productivity of Rice Farms in Thailand

by

Sutapa Amornvivat

A.B., Applied Mathematics, Harvard-Radcliffe Colleges (1997)

Submitted to the Department of Civil and Environmental Engineering in partial fulfillment of the requirements for the degree of

Master of Science MASSACHUSETTS INSlTUTh in Civil and Environmental Engineering OF TECHNOLOGY MAR 0 6 2002 at the LIBRARIES MASSACHUSETTS INSTITUTE OF TECHNOLOGY

February 2002

@ Massachusetts Institute of Technology 2002

Signature of A uthor ...... Department of Civil and Environmental Engineering . I (\ January 15, 2002

C ertified by .. .,. ... - -.-...--...--... -...... Dennis B. McLaughlin H.M. King Bhumibol Professor of Water Resources Management Thesis Supervisor

Accepted by...... 4V Oral Buyukozturk Chairman, Department Committee on Graduate Studies Impact of On-farm Water Storage for Increasing Agricultural Productivity of Rice Farms in Thailand by Sutapa Amornvivat

Submitted to the Department of Civil and Environmental Engineering on January 15, 2002, in partial fulfillment of the requirements for the degree of Master of Science in Civil and Environmental Engineering

Abstract The effect of local hydrology on the availability of water is critical to successful agriculture. Both too much and too little water can adversely affect crop yields. Flooding can damage crops through submergence or uprooting while drought can cause desiccation or poor plant development. In central Thailand, both these conditions are sometimes experienced during the growing season. This thesis demonstrates that construction of on-farm ponds, which collect local runoff in order to mitigate these problems, can be both technically feasible and welfare improving. In order to test the technical viability of on-farm ponds, this thesis introduces a model to predict crop yields given hydrologic variability and a certain set of farm characteristics. In addition, the economic welfare implications of on-farm ponds are investigated for a hypothetical farm in Saraburi Province, Thailand. The economic analysis compares operations without supplemental irrigation and with an "optimally designed" irrigation system. Supplemental irrigation is found to be welfare improving through the reduction of risk and the increase of net income.

Thesis Supervisor: Dennis B. McLaughlin Title: H.M. King Bhumibol Professor of Water Resources Management

2 Contents

1 Introduction 10 1.1 Thailand's agricultural challenges ...... 11 1.1.1 Meteorological Variation ...... 1 1

1.1.2 Irrigation Development ...... 1 4 1.1.3 Rice Availability and Prices ...... 1 5

1.1.4 Rice Production ...... 1 5

1.2 His Majesty's "New Theory" Development Scheme ...... 17

1.2.1 Solution: the New Theory Project ...... 1 7

1.2.2 Necessary Conditions ...... 20

1.3 Objectives and Overall Approach ...... 20 1.3.1 Crop simulation model - CROP ...... 2 1 1.3.2 Hydrologic model - HYDRO ...... 2 1 1.3.3 Utility Evaluation - UTILITY ...... 23

2 Literature Review 24

2.1 Water Harvesting ...... 24

2.2 Crop simulation models ...... 25

2.3 Agricultural Investment Under Risk ...... 26

2.4 Summary of Contributions ...... 26

3 Study Area 28

3.1 General Description ...... 28 3.2 Soils & Surface Topography ...... 3 1

3 3.3 Water Resources ...... 3 2 3.4 Meteorological Profiles ...... 33 3.5 Agricultural Practices ...... 33 3.5.1 Types of Crop ...... 33 3.5.2 Irrigation System ...... 39

4 Methodology 40 4.1 Overview ...... 40 4.1.1 HYDRO-CROP Integration ...... 40 4.1.2 Model Validation ...... 42 4.1.3 UTILITY ...... 43 4.2 Hydrologic Models - HYDRO ...... 44 4.2.1 Concept ...... 44 4.2.2 Farm Configuration ...... 44 4.2.3 HYDRO : Inputs and Outputs ...... 47 4.3 Crop Simulation Program - CROP ...... 48 4.3.1 DSSAT Software ...... 48 4.3.2 CROP: Experimental Setting ...... 51 4.3.3 CROP: Inputs and Outputs ...... 51 4.4 Utility Evaluation - UTILITY ...... 52 4.4.1 Constant Relative Risk Aversion ...... 53 4.4.2 Expected Utility ...... 53

5 Results 55 5.1 Simulated Yields ...... 55 5.1.1 Statistics ...... 55 5.1.2 Model Validation ...... 56 5.1.3 Yield VS. Pond Size ...... 59 5.2 Implications for Investment ...... 61 5.2.1 Price of Rice Production...... 61 5.2.2 Optimal Pond Size ...... 63

4 5.2.3 Risk Aversion ...... 6 5 5.2.4 Pond Construction/Maintenance Cost ...... 6 7 5.2.5 Subsistence Constraint ...... 70 5.3 Sensitivity Analysis...... 70 5.3.1 Risk Aversion Parameter ...... 7 1 5.3.2 Rice Price ...... 7 1 5.3.3 Subsistence Constraint ...... 74

6 Conclusions 78 6.1 Evaluation of On-farm Pond Installation ...... 78 6.2 Recommendations for Future Research ...... 80 6.2.1 Multiple-cropping system ...... 80 6.2.2 Supplementary Water Resources to the On-farm Pond 82 6.2.3 Opportunity Cost of Agricultural Labor ...... 84 6.2.4 Household's Wealth Accumulation Beyond Subsistence 84 6.2.5 Government-aid Programs ...... 84

A HYDRO - MATLAB Code 86

B CROP - DSSAT Code 89

5 List of Figures

1-1 Rainfall distribution and variability common to the semi-arid areas in the central region of Thailand...... 12 1-2 Annual rainfall distribution illustrating the extent of variability (dotted line) associated with monsoon onset and withdrawal periods...... 13 1-3 In addition to irrigation, the pond can be utilized for fishery to enhance income. 18 1-4 New Theory demostration farm at Wat Mongkolchaipattana Royal Development Study Center, Saraburi province...... 19 1-5 Diagram illustrates overall approach of this study...... 22

3-1 Location of the study site in Chaleom Phrakiat District Saraburi Province, north of Bangkok...... 29 3-2 Topography of our study site and the watershed area...... 30 3-3 Average monthly rainfall and Class-A pan evaporation in Saraburi Province. . . 34 3-4 Maximum and minimum temperature at the study site...... 35 3-5 Crop pattern suggested by Praputthbaht Field Crop Experiment Station, Dept. of Agricultural Extensions in Saraburi Province...... 38

4-1 Interaction between the crop simulation program (CROP) and the water balance analysis (HYDRO)...... 41 4-2 Plan view of model farm showing rice paddies with surrounding ditches. Arrows represent important horizontal fluxes...... 45 4-3 HYDRO farm configuration: Vertical cross-sections through pond and rice paddy water storage compartments showing fluxes into and out of each compartment. . 46

6 4-4 Elements of the Decision Support System for Agrotechnology Transfer (DSSAT) 50 4-5 Timeline for rice planting processes with respect to growth. The period of each phasic development is averaged over 35 simulated cropping seasons...... 52

5-1 Histograms to fit normal distribution of yield: rainfed and irrigated rice farms. 57 5-2 A: CDFs of yield per unit area with and without irrigation. B: CDF of yield increase with pond installation...... 58 5-3 Simulated and observed grain yield for irrigated and rainfed condition...... 60 5-4 Plot of yield per unit area versus pond size from our model simulation. The continuous curve fits the plot...... 62 5-5 Average revenue with respect to pond size. Optimal pond size is at the peak of curve...... 64 5-6 Plot of expected utility of revenue versus pond size. The curve results from interpolation of HYDRO-CROP simulated data (green dots)...... 66 5-7 Histograms of revenue under irragated and rainfed conditions. The dotted line represent estimated normal distributions using parameters derived from both sets of revenue...... 68 5-8 Cumulative distribution functions for revenue (top) and utility (bottom) from rice production under irrigated and rainfed conditions...... 69 5-9 Impact of the risk aversion coefficients on utility...... 72 5-10 Maximum cost of a 5% pond in which the on-farm pond scheme is attractive to

investment in relation to different levels of farmer's attitude towards risk. ... 73 5-11 Relationship between the maximum cost of a 5% pond and price of rice produc- tion. Here we assume the price of rice equal to 4.3 baht/kg ...... 74

6-1 Multi-cropping system in the New Thoery Study Center in Saraburi Province. . 81 6-2 The complete irrigation network of the New Theory: A big reservoir supplies a small; a small reservoir supplies a pond...... 83

7 List of Tables

3.1 Source: Department of Agriculture and Royal Irrigation Department, Ministry of Agriculture and Cooperatives ...... 32

5.1 Simulated and observed grain yield for irrigated and rainfed conditions...... 76 5.2 Area allocated for rice paddy in an average Saraburi rice farm with respect to percentage of pond area ...... 76 5.3 Statistics of simulated revenue for rice production, generated from 35 years of actual weather data (1961-1995) ...... 77 5.4 Input values used in sensitivity analysis ...... 77 5.5 Summary of sensitivity analysis results ...... 77

8 Acknowledgments

I would like to thank all of those who have taught and inspired me over the years: my deepest appreciation goes to my advisor, Professor Dennis McLaughlin, for his patience and support during my inevitably awkward early years of interdisciplinary research. This thesis might have suffered a dismal existence were it not for our discussions-discussions that helped me sort through the mental chaff and crystallized my ideas. I feel deeply indebted to His Excellency Dr.Chaovana NaSylvanta and Dr.Sippanondha Ketudat for their invaluable advice and continuing interest in my professional development. In addition, I want to acknowledge Dr.Djitt Laowattana who gave me the opportunity to work on this meaningful project and has done so much for my academic pursuits Julie, Freddi, and other past and present members of Parsons Laboratory deserve special thanks for directly influencing the course of my graduate project. Mon, Fred and Helani provide unfailing friendship and quality times during our weekly dinners. P'Euay, P'Nong and everyone else in TSMIT community has been a source of inspiration and comfort. I am grateful to my dear friend Jan whom I owe many thanks for restoring my peace of mind through tough times and sharing excitement and joy when things turn our all right. This thesis is dedicated to my parents, Police General Sawat and Khunying Catliya, with the hope that they see this as only a tiny part of the love and gratitude I feel. Their understanding and dedication have made this long journey a rewarding success.

9 Chapter 1

Introduction

Monsoon Asia is generally characterized by distinct wet and dry seasons, which are associated with wide seasonal fluctuations in agricultural production, farm income, labor use, wages, and prices. In Thailand, the most persistent constraint to increasing rice productivity has been identified as the lack of water. Although many of the rice cultivation areas in Thailand, mainly those in the central region, receive fairly high total annual rainfall, the rains are highly concentrated within the monsoon months. Excess water that may be available during part of the growing season is unavailable at critical rice development stages and cannot be used by other crops. Irrigation is perhaps the most obvious instrument for attempting to reduce the fluctuations between peak and slack seasons. The effect of local hydrology on the availability of water is critical to successful agriculture. Both too much and too little water can adversely affect crop yields. Flooding can damage crops through submergence or uprooting while drought can cause desiccation or poor plant development. In central Thailand, both these conditions are sometimes experienced during the growing season. Strong monsoon rains and low permeability soils combine to cause damaging floods. However, sporadic dry periods within the wet season and uncertainty in the time of onset of the wet season create drought hazards for agriculture in this region. Proposals for the development of irrigation in this region include the construction of numerous small farm ponds, which collect local runoff and may be supplemented by larger storage reservoirs. In order to test the viability of such a plan, this thesis models the hydrology and agricultural response of the region. We introduce a model framework to predict crop yields given hydrologic

10 variability and a certain set of farm characteristics. We also investigate the impact of pond size on farmer's investment decisions. This chapter examines the problems of water scarcity that Thai rice farmers in the central region encounter and the ongoing government effort to tackle such challenges. The objectives and overall approach of the thesis are also briefly summarized. Chapter 2 surveys literature on water harvesting schemes, crop simulation models, and agricultural planning under risk. Chapter 3 describes the study site. Chapter 4 explains the methodologies developed in this thesis in detail while Chapter 5 discusses the results and carries out sensitivity analyses. Lastly, Chapter 6 summarizes the study and suggests directions for future research.

1.1 Thailand's agricultural challenges

1.1.1 Meteorological Variation

Agricultural drought may be defined as an unfavorable water balance for optimum crop pro- duction. Since different crops require different soil-water regimes for optimum yields, drought problems must be looked at in relation to the crops to be grown. Most often, rainfall variation and unpredictability affect rainfed lowland rice yields more than the shortage of water does in aggregate sense. In Thailand, even with the high concentrations of rainfall events during the main rice season, rainfed rice in many areas and in most years suffers from in-season drought and consequent loss of yield. Another constraint farmers face is the variation in the timing of the onset and termination of the monsoon. The extent of such variations for Saraburi province in the central region of Thailand is shown in Figures 1-1 and 1-2. Such a situation gives farmers little choice but to adopt a very low-risk, low-input cropping system, which is directed to giving a stable rather than a high production. Water shortage prevents most farmers from growing a non-rice cash/commercial crop during the post-rice season. Even a marginal improvement in the soil-water status in this season can make a significant difference in cropping opportunity and/or yields.

11 40 n I I I I I I I I I I I

35C- std. deviation

30C- mean (1961-1997)

25C-

0a

15C-

10+

50-

0 A I I i- J F M A M J A S 0 N D

Figure 1-1: Rainfall distribution and variability common to the semi-arid areas in the central region of Thailand.

12 35C I Ithdawa I I I I

301

25C- withdrawal std.dev = 24 days

20C- E E onset3..y 15C std.dev = 36 ys

10C

50-

0 J F M A M J J A S 0 N D

Figure 1-2: Annual rainfall distribution illustrating the extent of variability (dotted line) asso- ciated with monsoon onset and withdrawal periods.

13 1.1.2 Irrigation Development

Early investments in irrigation schemes in Thailand were large projects and concentrated on major water distribution systems in the Central Plain. The investments were largely to increase the yield of the wet-season rice crop in the central region by improving the reliability of the water supply. In the early 1970s the emphasis was shifted to smaller projects which provided water distribution networks at the farm level. The benefits of irrigation include not only double cropping of rice in the Central Plain and a modest expansion of upland crop cultivation in certain parts of the north and northeast, but also an increase in rice yield per hectare in both the wet and dry seasons. The increase in rice yield has come about because irrigation has induced a shift in planting technique from broadcasting to transplanting, a shift from traditional to modern rice varieties, and increased use of fertilizer and other chemical inputs. The wet-season rice yield per hectare in irrigated areas during the first half of the 1990s averaged 2.8 tons, almost double the rainfed yield of 1.7 tons. Total rice production increased from an average of 13.9 million tons in 1971-75 and 22.1 million tons in 1991-95. Dry season rice yield increased from 6.3 percent of the total crop to 19.4 percent during the same period. Conventional irrigation facilities, however, are not likely to be developed for the vast ma- jority of cultivated areas in the foreseeable future. Investment in irrigation infrastructure has been declining in the last decade and much of the rainfed lowland areas is unavailable for large- scale irrigation projects because of topography, soil drainage, or water storage problems. Yet the pressure to produce more crops from the land is increasing as the population continues to grow at relatively high rates. On-farm reservoir systems offer a promising way to alleviate drought. These systems are used to harvest rain falling on the pond and surplus runoff produced in the catchment area, and to store the water for subsequent use. Farm ponds have been in use in rural areas since ancient times. But today this time-tested technique has acquired a new critical dimension because recent advances in agricultural technology has opened up new opportunities for farmers. For instance, new crop cultivars which require less irrigation-intensive treatment have been introduced and adopted by local farmers. Recent research results have shown that scientific management of water resources within a microwatershed, using an on-farm reservoir as the

14 centerpiece, is economically attractive and socially desirable [3, many examples of the on-farm impoundment system are discussed ].

1.1.3 Rice Availability and Prices

The increased rice output resulting from irrigation has increased the supply of rice available for domestic and foreign consumers. The growth rate in rice production has permitted modest growth in Thailand's per capita domestic rice consumption, together with large increases in export volume.

At an aggregated level, expansion of irrigated area may reduce seasonal rice price variation.

Two related effects may contribute to reducing the magnitude and increasing the predictability of seasonal price spreads. First, an increase in the absolute level of rice production in the dry season may reduce the storage requirements and cost of holding much of the wet-season production through to the following wet season. Second, seasonal price changes are influenced by the deviations in rice production from expectations. If irrigation can reduce variability in production, and hence the deviations of actual from expected production, it will contribute to mitigating the inter-year instability in seasonal price changes [36].

At the farm level, irrigation may have large effects on seasonality in cropping practices, yield and production, farm income, and labor demand. When irrigation permits dry-season cropping, it will tend to smooth the seasonal pattern of production, income and labor use.

1.1.4 Rice Production

Planting Methods

In the rainfed rice-growing areas of Thailand, only dry soil broadcast planting is practiced, but in irrigated areas there has been a shift to transplanting and, to a small extent, to puddled soil

broadcasting. Land consolidation tends to reinforce the shift to puddled broadcasting. This

new broadcasting method, coupled with increases in application of fertilizer and chemicals, provides a high yield per hectare. This technique requires a well-leveled field and a good water

control system to drain out water before broadcasting so that the water level can be increased

with the height of the growing plants. In irrigated areas without land consolidation, puddled

15 soil broadcasting can be, and was, practiced over more limited areas (28 percent in the wet season and 30 percent in the dry season) depending on the degree of water controllability.

Net Cash Income of Farm Households

The largest source of income to local farm households is the sale of paddy rice, contributing over 90 percent of the total income in each cropping season in irrigated areas without land consolidationi, 80-85 percent in land-consolidated areas, and only 65 percent in the wet-season rainfed areas. Other crops and livestock account for 24 percent of total cash income in the rain- fed areas in the wet season, and livestock alone contribute about 23 percent in the dry season. Livestock and nonfarm employment are important sources of income for rainfed households, not only in the dry season when rice cannot be grown, but also in the wet season, because paddy sales bring in only 65 percent of total income2 For the land consolidated farms, other important sources of income are employment on other rice farms, especially in the wet season, and tractor rentals, mainly in the dry season. Since the broadcasting techniques used in land-consolidated areas require less labor than the transplanting method, household laborers are hired to work on other farms in irrigated areas without land consolidation. This hiring out occurs before and after broadcasting and harvesting on these laborers' own farms. Income from tractor rentals is mainly derived from renting out small tractors for land preparation, not only on rice farms but also in areas growing upland crops in the dry season. Consequently cash income from sources other than paddy rice is much larger for land-consolidated farms than for households in other areas. This study, however, emphasizes the impacts of on-farm reservoirs on non-consolidated rainfed farms in semi-arid areas of the central region, where transplanted rice is predominant. Household income for such farms comes mainly from the sale of rice.

Land consolidation under private farmland ownership refers to an exchange of the private ownership and location of spatially dispersed parcels of farms to form new holdings containing just one (or as few as possible) parcel(s), with the same (or similar) wealth in land as that before the exchange. Land consolidation can turn farms from fragmented to compact enterprises, enlarge parcel size, and make sale, lease, and other forms of joint use of land physically easier. But it does not enlarge farm size; e.g., a farm previously composed of 10 dispersed parcels (on average 0.1 ha each) could now hold one compact parcel of 1 ha. However, only farmers with a medium-size or larger land holding are likely to undertake such legal transactions. 2The statistics of husbandry and nonfarm incomes are adapted from Center for Agricaltural Information (1998) and Somrith (1998).

16 1.2 His Majesty's "New Theory" Development Scheme

From past to present, the shortage of water supply for agricultural activities has been a major problem facing Thai farmers. The impact has been more severe in agricultural areas where farming activities depend heavily on rainwater. Unfortunately, such areas constitute a pre- dominant part of the country. Hydrologic variability enables farmers to plant only once a year during the rainy season. Moreover, farmers are exposed to risks and damages due to adverse conditions of the soil, climate, and inconsistent rainfall patterns. Although efforts have been made to counter the water shortage problem by digging ponds or enlarging channels to store water, clear design guidelines have never been established. Besides, there are also other fac- tors that magnify the water shortage situation. Examples of such factors are unsystematically planned crop cultivation and the mono-cropping farming system, under which only rice or field crop is planted.

1.2.1 Solution: the New Theory Project

Aware of the situation, His Majesty King Bhumibol Adulyadej initiated a development scheme, called "Trisadee Mai" or "New Theory" in direct translation, intended to provide guidelines for small-scale farmers for optimum management of their land and water resources. The concept requires division of a small piece of land into distinct parts. It entails technical calculations to determine the amount of water needed to support cultivation all year round. The idea provides individual farmers with a comprehensive plan that consists of three phases:

Phrase I The first phase of the New Theory requires dividing the land into four parts under a certain ratio: Part 1 is designated for the pond; Part 2 is set aside for rice cultivation; Part 3 is used for multi-cropping farming; Part 4 is allocated for housing, animal raising, and other activities.

Phrase II The second phase suggests that farmers pool together their efforts and resources in the form of a group or cooperative to execute the following activities: 1) production (crop va- rieties, soil preparation, irrigation system, etc.); 2) marketing (sundry area, silo, rice mill, product distribution); 3) welfare (public health, loan); 4) education (school, scholarships); 5) society and religion.

17 Figure 1-3: In addition to irrigation, the pond can be utilized for fishery to enhance income.

Phrase III The third phase involves making the necessary contacts and coordination to establish a fund or ensure the source of funding from banks and companies in order to assist them in the investment of activities to improve the quality of their lives.

This royal-initiative development scheme first started in 1995 at Wat Mongkol Chaipattana Royally-initiated Development Study Center in Saraburi Province. Since then, the program 3 has been extended to various geological areas in other Royal Development Study Centers [9, The history and implementation of the New Theory concept is described]. This study, however, uses the physical properties of Saraburi area as a case study for the model which is discussed extensively in Chapter 3.

3 Royally-initiated Wat Mongkol Chaipattana Development Project in Saraburi Province, Khao Hin Sorn Royal Development Study Centre in Chachoengsao Province, Royally-initiated Khlong Si Siad Reservoir Development Project in Nakhon Nayok Province, Huai Sai Royal Development Study Centre in Phetchaburi Province, Kung Krabaen Bay Royal Development Study Centre in Chanthaburi Province, Puparn Royal Development Study Centre in Sakon Nakhon Province, Baan Dan Samakkee New Theory Demonstration Project in Kalasin Province, Pak Thong Chai Royally-initiated New Theory Demonstration Project in Nakhon Ratchasima Province, Huai Hong Khrai Royal Development Study Centre in Chiang Mai Province, Pikun Thong Royal Development Study Centre in Narathiwat Province

18 * *--U44 %* - -

* - -

Figure 1-4: New Theory demostration farm at Wat Mongkolchaipattana Royal Development Study Center, Saraburi province.

19 1.2.2 Necessary Conditions

There are three necessary conditions for the success of the New Theory concept: (1) cooperation among participating farmers, (2) technical feasibility and (3) economic feasibility. The main idea of the New Theory is to serve as a production system that allows farmers to become self-sufficient and self-reliant. To be viable, this concept requires the unity and willingness of the community to work with and assist each other in order to reduce expenses, similar to the traditional rice harvesting practice. However, this requirement is beyond the scope of this study and is subject to research in the area of social sciences. This study only looks at the hydrologic and economic feasibility of this proposed project presuming the best possible cooperative effort of participants. Another important aspect of the New Theory is to use water storage to supply farming during dry seasons or dry spells. Therefore, the concept requires that part of the land is set aside for the construction of a pond. This study is to find guidelines for the suitable pond size to suit each area's local characteristics, including soil characteristics, the amount of rainfall, and the general environment. For instance, in the southern region where rainfall is more plentiful or in areas where sources of water are available to continuously replenish the pond, it will be possible to reduce the size of the pond and allocate the surplus land for more productive purposes.

1.3 Objectives and Overall Approach

The New Theory concept has been widely implemented in other areas beyond demonstration farms. It is necessary that operators thoroughly and clearly identify the conditions in which this concept is technically viable and economically attractive. The main objective of this study is to propose a simplistic approach for determining the feasibility and benefits of the on-farm pond, suggested in Phase 1 of the New Theory concept. This thesis brings together scientific work on hydrology, as well as soil science and plant biology, with economic modeling. The goal is to model the physical system on its own, and then to integrate that system with the investment decisions of the households. Profit, risk, and variability of yields will all influence choice of pond investment, as farm households are generally risk averse.

20 We focus on an area around the Wat Mongkol Chaipattana Study Center to demonstrate our approach. More detailed methodology is discussed in Chapter 4. The work described here is a stepping-stone to establish the link between hydrology and agricultural productivity under the New Theory development scheme. More careful economic analyses are expected in future research to extend the model. Our model can be divided into three modules: CROP, HYDRO, and UTILITY. The first two modules, CROP and HYDRO, operate simultaneously to simulate rice yields under rainfed and irrigated irrigation systems. Then the process of utility evaluation is included to translate agricultural production into household welfare and to determine whether the on-farm pond irrigation system is an attractive investment. The overall model interaction is illustrated in Figure 1-5.

1.3.1 Crop simulation model - CROP

The crop simulation model integrates the aggregate effect of the underlying processes that de- termine the behavior of complex agricultural systems. It is based on CERES-Rice, a component of the Decision Support System for Agrotechonology Transfer (DSSAT) software [57]. DSSAT is a collection of crop models and computer programs integrated into a single software package in order to facilitate the application of crop simulation models in research and decision making. This software is a product of the International Bench-mark Sites Network for Agrotechnology Transfer (IBSNAT) project and is widely used in agricultural planing and agronomic research. CROP is a site-specific file, set up in DSSAT standard format, which enables use to inves- tigate the effect of different irrigation schemes on rice yields in our study area. The required

inputs consist of weather, cultivar, soil structure, irrigation management, fertilizer application, planting and harvesting methods. The CROP model simulates growth and development as well as grain yield under different irrigation treatments: rainfed and irrigated under an on-farm reservoir scheme.

1.3.2 Hydrologic model - HYDRO

HYDRO is a quantitative water balance analysis model which takes in hydrological information and irrigation needs, computed in CROP. Its main function is to investigate the feasibility of

21 fluxes & storages

simulated yields

F r -

Figure 1-5: Diagram illustrates overall approach of this study.

22 supplying a small on-farm pond exclusively from runoff generated within the farm boundary. Water stored in the pond is diverted to support rice field irrigation. The HYDRO model quantifies water fluxes entering and leaving the farm system assuming all runoff is harvested and given a specified irrigation rule. Another task is to decide whether the irrigation rule can be sustained. If the pond cannot supply enough water to satisfy irrigation demand; a new irrigation rule is assigned to the CROP module. Crop and hydrologic information is shared as inputs by both modules while water fluxes and storage values are passed between the two modules. The CROP and HYDRO modules operate simultaneously and complement each other in simulating grain yield from actual me- teorological data. All interactions are carried out manually; the two computer programs are run independently. The simulation process converges when water use in CROP is consistent with water supply from HYDRO. After convergence, the results of the simulation are stored for utility evaluation.

1.3.3 Utility Evaluation - UTILITY

The UTILITY module evaluates the impact of on-farm pond investment, given simulation results from the HYDRO-CROP integrated model. The assumed utility function is used to determine the conditions in which on-farm investment is attractive to local farmers. The utility evaluation module is a combination of break-even analysis and sensitivity analyses. The results allow the analyst to understand how changes in parameters affect investment choices.

23 Chapter 2

Literature Review

There is an extensive literature concerned with the impact of investment in water resources to alleviate drought and increase productivity in rainfed lowland rice farms. However, only a few of the studies reviewed below establish a link between the technical viability and welfare impli- cations of such investment. The main objective of this thesis is to simultaneously investigate the technical and economic feasibility of on-farm irrigation in central Thailand. The economic analysis compares operations without the supplemental irrigation and with an "optimally de-

signed" irrigation system. Our overall approach combines the concepts of water harvesting, on-farm storage, crop simulation, and agricultural investment in the presence of risk.

2.1 Water Harvesting

Water harvesting, defined in its broadest sense as the collection of runoff for productive use

[51], is an ancient art practiced in the past in many parts of North America, the Middle East, North Africa, and Asia. More precisely, water harvesting can be defined as the process of concentrating rainfall as runoff from a larger catchment area to be used in a smaller target area. This process may occur naturally or artificially [40]. Water harvesting supports a flourishing agriculture in many dry areas, where rainfall is low

and erratic in distribution. Examples are given, among others, by Oweis and Taimeh (1996), Suleman et al. (1995), Krisna et al. (1987) and Oswal (1994). However, for any agricultural water development to be successful, it must be economically viable. The sustainability of the

24 various water harvesting techniques is found to depend largely upon the timing and the amount of rainfall ([10],[48], and [4]). A good historical review of rainwater harvesting for agriculture is given in UNEP (1983). Although the revival of water harvesting techniques began in the early 1930s, little construc- tion and research activity began before the late 1950s [19]. In the 1960s, various governmental, private, and university research organizations, particularly in arid and semiarid areas, initiated studies to develop and evaluate water harvesting techniques with lower installation costs and improved system reliability. As a result, new water harvesting techniques have appeared in the literature during the last two decades [11]. Recently, there has been renewed interest in water harvesting in South and Southeast Asia, probably as a result of the lack of water in times of need, which is the dominant obstacle to increasing productivity of most rainfed ricelands [3].

2.2 Crop simulation models

Crop modeling enables researchers to integrate knowledge from different disciplines in a quanti- tative way. That, in turn, helps researchers understand the underlying processes that determine the behavior of complex agricultural systems. Mathematical models are caricatures of real- world systems. Solving the equations enables a numerical description of the system to be produced. Ultimately, 'what if' questions can be asked about the functioning of a system, and numerical answers are provided [42]. Hansen and Jones (2000) surveyed issues and approaches related to applying crop models at scales larger than the plot. Crop models are particularly useful in rainfed lowland rice ecosystems, which are charac- terized by high temporal variability and spatial heterogeneity. A limited number of field ex- periments cannot provide a reliable basis for management strategies under the myriad possible conditions that exist. Simulation models can replace expensive and time-consuming experi- ments due to their ability to generalize experimental findings and help interpret the results of a few selected experiments ([56], [49], [61], and [5], for examples). Modeling is especially useful in yield gap analysis, a method for identifying constraints to agricultural production in different agroclimatic zones. Pinnschmidt et al. (1996) concentrated

on ameliorating those factors that contribute to the gap between farm yield, potential farm yield,

25 and potential experiment station yield.

2.3 Agricultural Investment Under Risk

Previous studies have addressed the economic feasibility of on-farm impoundments. Burt and Stauber (1971) studied such investments. However, they focused on irrigation scheduling without addressing water availability, water supply reliability, impoundment size or crop choice. Krishna et al. (1987) studied the economic feasibility of on-farm ponds in east Texas, but assumed deterministic water supply and irrigation demand along with fixed impoundment size and irrigated land area. Dudley et al. (1971a, b, 1972) developed short, intermediate, and long- run optimizing models using stochastic water supply and demand to determine the combination of reservoir size and irrigated acreage for a single crop. Their approach did not allow for irrigated/dryland crop substitution or multiple crop substitution. Other studies have assumed availability of sufficient water ([1] and [41]). Furthermore, the effect of risk attributes has not been comprehensively examined in the irrigation system investment context. The reduction in income variability can be an important impetus for irrigation adoption [1]. Pender (1992) proposes welfare analysis for investment in shallow wells; whereas Featherstone and Goodwin (1993) discuss the risk aversion effect on farmer's choice in soil conservation. Ziari et al. (1995) uses a risk-sensitive model which simultaneously considers water supply, irrigation system investment, irrigation scheduling, and crop mix selection to evaluate the economic feasibility of small impoundments.

2.4 Summary of Contributions

In attempt to fulfill the gaps left in the aforementioned studies, this thesis has the following objectives:

o to derive stochastic series of simulated yields directly from actual meteorological data;

o to derive an on-farm pond design for subsistence rice farms in central Thailand, accounting for the roles of uncertainty and risk.

26 These objectives are achieved by using a modeling approach which simultaneously considers impoundment size, irrigation technique, variable rainfall and farmers' attitudes towards risk.

27 Chapter 3

Study Area

In this Chapter we give a general description of the case study area's soil, water resources, and agriculture. We then explain existing irrigation management in the rice fields.

3.1 General Description

Saraburi Province is located on the edge of the Chao Phraya Basin in the central region of Thailand. Our study site is located to the south of Wat Mongkol Chaipattana Royally-initiated Development Study Center in Chaleom Phrakiat District of Saraburi Province (Figure 3-1). The land is a rice field that belongs to Mr. Tongsuk and Mrs. Sawang Pimsan. Because the farm is outside the Royal Projects Study Center, it is not directly connected to the Huai Hin Khao reservoir supply system (Site B in Figure 3-2). The Pimsan farm is suitable for a study site for a number of reasons:

1. The farm grows primarily transplanted lowland rice; its paddy fields used to be strictly rainfed until recently.

2. It applies the New Theory concept and uses an on-farm pond which, since its construction, has been able to contain sufficient water to insure a higher yield paddy rice crop during the wet season and to irrigate some dry season and perennial cash crops.

3. Most of the water in the pond appears to be generated locally. It does not need supple- mental irrigation from outside its boundary. It is uncertain whether some of the water in

28 Figure 3-1: Location of the study site in Chaleom Phrakiat District Saraburi Province, north of Bangkok.

29 _t

117 r~ 4.7 e TeF1 M - ----

~K.;

- L bA1-

4- - - ; .*#c~J~~j -

m V

- - mIFE - -

S- -dwg~

bfL*UtWaN bIW M

WIeng -~~i~aw -' IAa.LE N

0 1 3 Jeilometars

Figure 3-2: Site B illustrates the watershed area, downstream of Huai Hin Khao reservoir. Its boundary concides with that of Wat Mongkol Chaipattana Development Study Center. Pimsan farm is shown in a red star on the south border of the Royal Study Center.

30 the pond comes from runoff from adjacent farms supplied by the Huai Hin Khao system. However, this issue is not important to our study because we only evaluate the feasibility of an on-farm impoundment which captures runoff generated within the farm boundary. If the pond system is feasible in the absence of supplemental irrigation, additional runoff will add extra benefit to the system.

Physical characteristics and cultivation practice serve as inputs to our computer model, which combines hydrologic model and crop simulation program.

3.2 Soils & Surface Topography

The farm and the Royal Projects site are on the northeastern edge of an extensive rice-growing region or the Noi-Lopburi floodplain, approximately 8 km north of the main district of Saraburi Province. The watershed area of our study site occupies roughly 4.3 square kilometers; this area is shown as site B in Figure 3-2. Huai Hin Khao reservoir (14'36'N and 100'55'E) was constructed in 1995 to collect runoff from the catchment area and limestone outcrops. The area further north of the Royal Study Center has somewhat different soil and is devoted mostly to maize cultivation. From our interviews with local farmers, the area grew almost entirely rice until the last decade when some farmers changed to maize for animal feeds in order to avoid the adverse effects of dry spells. The area is relatively flat with slopes of less than 30, except for scattered limestone outcrops up in the north. It is situated approximately 15 meters above mean sea level (MSL). The basic soil group is quite uniform across the Royal Study area. It is sandy clay loam suitable for transplanted paddy rice as the infiltration loss is quite low. Low in organic matter and soil nutrients, the soil is prone to waterlogging and seasonal flooding during the wet season. The soil is categorized as Khao Yoi Series. Soil properties for the area in the Study Center around Wat Mongkol Chaipattana is shown in Table 3.1.1 Both soil layers are slightly basic.

'Electrical Conductivity (EC) is water's ability to conduct an electrical current and is directly related to the concentration of dissolved salts. The greater the EC, the more dissolved salts are present in solution. Cation Exchange Capacity (CEC) is an indication of the number of exchange sites within a soil that may temporarily hold positively charged ions. It is generally determined by the amount and type of clay and the amount of organic matter.

31 Low cation exchange capacity (CEC) indicates inefficiency of chemical fertilizers and thus, lower crop productivity

Depth from surface 0-1 m below 1 m pH 7.6 7.8 EC at 25 0 C [millihos] 0.113 0.075 Organic Matter [%] 1.24 0.64 P [ppm] 2 1 K [ppm] 20 15 CEC [me/100gm] 9.5 6.5 Bulk Density [gm/cc] 1.78 2.36 Particle Density 2.17 2.17 Porosity [%} 34.32 13.56

Table 3.1: Source: Department of Agriculture and Royal Irrigation Department, Ministry of Agriculture and Cooperatives

3.3 Water Resources

Water in the pond has ph = 7.5 because the lower soil layer (below 1 m) originates from Calcareous soil. The level of calcium contamination in pond water is insufficient to be a hazard to plant production and fishery. Farms in the Study Center replenish their pond through a pipeline network from Huai Hin Khao reservoir. Gravity does the work as the reservoir is located uphill of the area. Nevertheless, our study area in practice is not directly connected to the network. The pond captures in situ rainfall as well as runoff generated near and on the farm. The watershed in Site B of Figure 3-2 appears to be in an area where the clay is primarily underlain by a layer of sandstone/shale/quartzile which has some interbedded layers of lime- stone. There are also some regions with a marl deposit. These marl and sandstone layers range from 50 to 100 meters thick. Beneath these layers there is an extensive limestone deposit which outcrops in isolated regions. The limestone is shown to extend at least to about 200 m

below the ground surface. Further south-near Wat Mongkol Chaipattana and our study site, the marl and the sandstone layers are absent. To look at a larger region than the boundary of Site B, the big area seems like a limestone bowl, which is filled in by other deposits in our study site [6].

32 Hand-dug ponds can reportedly reach no deeper than 4-6 meters due to a limestone layer which lies near the surface. Local groundwater is inappropriate for crop use due to high contamination of heavy metal elements and low yield. Further complicating the matter are the legal issues surrounding groundwater regulation. The usage of groundwater for irrigation is regulated through a well licensing system. Users are granted an entitlement to pump a fixed volume of water per year. However, since the pumped amount is difficult to monitor, the groundwater system is at high risk of overuse. In addition to the common pool problem, the fear that high usage of well water upstream may lead to the flow of saline groundwater into downstream regions, particularly the Bangkok area, has put a moratorium on the issuing of new pumping allocations.

3.4 Meteorological Profiles

Agroclimatological data was obtained from two closest stations operated under Meteorological Department: Muang District, Saraburi (station 414001, 14'31'N and 100'56E and Phu Khae Botanical Garden (station 414012, 14 0 40'N and 100'55E). Daily series are available for 46 years (1952-1997) of rainfall; 17 years (1982-1998) of Class-A pan evaporation; 38 years (1961-1998) of temperature. The area has average annual rainfall of 1200 - 1400 mm with high variation during wet seasons as shown in Figures 1-1 and 1-2. The average number of rainy days is 84. Annual pan evaporation is slightly greater than the amount of annual precipitation. Like rainfall and solar radiation, temperature shows an obvious seasonal cycle; this characteristic is typical of monsoonal tropics. The wet season typically starts in June and lasts until November. The monthly precipitation and pan evaporation are shown in Figure 3-3. Solar radiation during the monsoonal period naturally drops due to cloudiness.

3.5 Agricultural Practices

3.5.1 Types of Crop

The type of crop that can be grown is constrained by the soil's nutrients. Rice seems to be a prominent commercial crop in the study area since it is not as adversely affected by poor soil

33 30C - average pan evapora > = 7waverage rainfall

250

200

-150 00x0

100-

50k

O'L- 0 J F M A M J J A S 0 N D

Figure 3-3: Average monthly rainfall and Class-A pan evaporation in Saraburi Province.

34 40 + average temperato

35-

30k 0 ++ CD + + I ) a) 25-

20k

151 J F M A M J J A S 0 N D

Figure 3-4: Maximum and minimum temperature at the study site.

35 nutrients as other cash and field crops [52]. However, as farmers have applied more fertilizers, the nutrient constraint has naturally reduced. The area further north of the Royal Study Center has somewhat different soil and is devoted mostly to maize cultivation. From our interviews with local farmers, the area grew almost entirely rice until the last decade when some farmers changed to maize for animal feeds in order to avoid the effects of dry spells. In this section, we will discuss each important crop cultivated in this area and the time schedule for planting and harvesting.

Lowland Rice

Rainfed lowland rice grows in a bunded fields that are flooded for at least part of the cropping season to water depths that may exceed 50 cm for no more than 10 consecutive days. Rainfed lowlands are characterized by lack of water control, with floods and drought being potential problems. Adverse climate, poor soils, and a lack of suitable modern technologies keep farmers from being able to increase productivity. Currently, yield constraints for rainfed lowland rice include frequent drought and flood, soil properties, diseases such as blast and bacterial blight, and insect pests like gallmidge and stem borers. The current breeding objectives are to develop cultivars that are tolerant to adverse environmental conditions and resistant to disease and insects, while maintaining the superior grain characteristics of the traditional cultivars and their appropriate maturity characteristics [53]. The most common cultivar for domestic consumption and local market in the Central Plain is non-glutinous RD7 while that for export is fragrant Khao Dawk Mali (KDML) 105. Since our study area mostly consists of small-scaled farmers, very few of them grow rice for export. Transplanting and direct seeding are the two primary methods of growing rice. For trans- planting, seedlings are first grown in a seedbed for a month while the main fields are being ploughed along the bunds surrounding each paddy. When each field has been prepared, the seedlings are pulled from the seedbed and then transplanted in the field by hand. In broad- casting, seeds are sown directly in an upland field or lowland paddy. In upland fields, seeds are sown only by dibbling on dryland preparation. The reasons for changing from transplanting to broadcasting are (1) to save time and labor cost, and (2) to avoid the adverse effects of drought during the transplanting period.

36 Without irrigation, farmers wait for heavy rains to puddle the land before they transplant seedlings, no matter how long rains are delayed. If the heavy rains do not come, the rice farms are either not planted at all or the crop may fail after transplanting to dry soils. To avoid the problem of transplanting in dry soils, farmers may change to the dry-seed broadcasting at the beginning of the rainy season or during the dry period. If the crop fails due to adverse conditions, reseeding is easier and less costly than retransplanting [37]. Most farmers prefer the transplanting method. Transplanted rice usually yields higher productivity due to more control over seedlings. Irrigation allows farmers in water-deprived areas to choose transplanting methods. Rainfed lowland rice ecosystems may be divided into five subecosystems: (1) favorable rainfed lowland, (2) drought-prone, (3) submergence-prone, (4) drought/submergence-prone, and (5) medium-deep water. Technologies for the irrigated rice sector can be applied in both the favorable rainfed lowland and drought/submergence-prone subecosystems. Our study area belongs to drought/submergence-prone subecosystem.

Farmers can control their transplanting or broadcasting time if irrigation water is available, regardless of the onset time of the rainy season. Lowland rice is grown mainly during July to November as illustrated in Figure 3-5. Broadcasted rice matures approximately the same time as transplanted rice. The second line in Figure 3-5 indicates broadcasted rice. Harvesting for lowland rice generally starts in November. Upland rice is primarily planted by the broadcasting method, which starts in June. With its maturity 50-day longer than that of lowland rice, upland rice is harvested in December for 2-3 weeks.

Perennial Trees

From our field surveys, most farmers in the area keep a small orchard of perennial fruit trees: banana, mango, and papaya2 . Few orchards grow only one variety, while many keep a combi- nation of many types of fruit even though the orchard area is usually small. The produce is largely for domestic consumption. In case of an abundant harvest, production excess is sold on the spot. Since most farmers grow similar varieties of fruit, the price is normally low in years of

2Besides the three major fruits, examples of fruit trees and perennial plants observed in Saraburi area are tamarind, jackfruit, sapodilla, orange, custard apple, santol, sesbania (Sesbania grandiflora), horseradish, neem tree, and cassod tree.

37 JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC

Upland Rice

L owland Ric Lowland Rice ..-

,, 'Field Crop & Vegetabl_ Crop &VgSeedling growth z Growing period Harvest

Figure 3-5: Crop pattern suggested by Praputthbaht Field Crop Experiment Station, Dept. of Agricultural Extensions in Saraburi Province. The parallelograms represent farmer's adjust- ment to weather conditions. abundant yield due to oversupply. As the orchard is not their prime source of income, farmers rarely tend it or invest in fertilizer application. Therefore, irrigation for orchards mainly relies on small shallow ditches surrounding each plot. These capture runoff from rainfall in excess of infiltration and evapotranspiration, instead of diverting water from ponds or nearby canals.

Cash crop - flowers & vegetables

Local farmers who practice wet-rice agriculture in valley bottoms also cultivate vegetable gar- dens, mainly for domestic consumption. Beans, corn, and native vegetables are planted in late January and harvested in May-June. Garden vegetables and herbal plants are gaining more significance as a prime source of income for some households. This option requires en- trepreneurship and intensive labor as garden crops require considerably more care than field crops or rice. Research and promotion is much needed to find garden crops suitable for the area's conditions and marketing infrastructure.

38 3.5.2 Irrigation System

The lowland communities in central Thailand have developed an agricultural system adapted to, and partially determining, the distinctive ecosystems of their areas. Water harvesting is combined with well-selected crops to ensure sufficient water. Farmers follow methodical rules of water harvesting for rice fields so long as there is enough water to sustain the system. The seedbed occupies a very small plot of land with a water level of 2-4 cm. After transplanting, the depth of water in the paddy is maintained at least 4.5 cm. The bunds surrounding paddy plots are 12 cm in height. Excess water spills to the surrounding ditches. When the ditches run dry, water from the pond is diverted to the paddies. If both the pond and rice ditch system are full, the excess water is spilled off-farm (down-gradient.) Farmers stop watering during the harvest period.

39 Chapter 4

Methodology

The model we use to determine the feasibility of the on-farm reservoir technique is divided into three separate modules: HYDRO, CROP, and UTILITY. This Chapter begins by showing how the hydrologic model (HYDRO) is integrated with the crop simulation program (CROP), and then explains a formulation for each of the two modules. The coupled HYDRO-CROP system provides simulated rice yields for a specified irrigation system and meteorological history. The yield results are used in UTILITY to evaluate the utility provided by the farming enterprise.

4.1 Overview

4.1.1 HYDRO-CROP Integration

All three modules are bundled together as illustrated in Figure 1-5. However, the first two modules-HYDRO and CROP-operate simultaneously and deserve a special attention to un- derstand the flow of information between them. Figure 4-1 shows input data and interactive processes within the two modules. The crop simulation program (CROP) relies on meteorological data series, soil properties, plant variety, and agricultural management information. In the diagram, the irrigation rule is separated from other management practices because it is adjusted during the model convergence process, which is carried out manually. In this study, transplanted rice is the only crop; CROP uses the CERES-Rice model included in DSSAT to simulate crop development. Section 4.3 discusses the simulation process in more detail. With all the required data, the model simulates

40 Weaiher

Gul~varPond fluxes and storage FM storages Havesti..ngI

Default ~Irrigation og maaeetor irrigation? irrigation reaclustment

Figure 4-1: Interaction between the crop simulation program (CROP) and the water balance analysis (HYDRO). The final output produced by the HYDRO and CROP modules is simulated yield per unit area. grain yield of rice for each cropping season. In addition, the crop simulation computes daily water requirement for the specified irrigation rule. Fluxes and storage values derived from CROP are passed on to an quantitative water balance analysis model, called HYDRO. The required meteorological inputs for HYDRO include daily series of Class-A pan evaporation and precipitation. The crop's evapotranspiration is calculated in CROP. The hydrologic module calculates on-farm water storage in the pond and in ditches surrounding the rice fields and indicates whether there is a water deficit for every daily step. If the water remaining in the pond and ditches is not sufficient to support the irrigation rule specified in CROP, this irrigation re-adjustment rule is manually changed, the CROP and HYDRO modules are rerun, and the water deficit is checked again. We repeat the entire cycle until the irrigation rule is satisfied and mass is conserved (i.e. until the simulation converges). The final result is a simulated grain yield which satisfies the modified irrigation rule without exceeding daily water supply. It is important to note that uncertainty in simulated yields results directly from meteorological variation.

41 4.1.2 Model Validation

An important facet of the modeling precess is to apply appropriate statistical tests to evaluate model accuracy. Dent and Blackie (1979) suggested a t-test to determine whether the slope and intercept of linear regression between model simulated and observed values are different from unity and zero. Willmolt (1982) contended that although the model is less sensitive to extreme values, bias (Equation 4.1) and root mean square error (RMSE) (Equation 4.2) which provides the "best" overall measures of model performance.

1 N Bias (Si- Oi) (4.1)

1N RMSE = (S, - Oj)2 (4.2)

where Si = simulated value, Oi = observed value, and N = number of observations. Graf et al. (1991) used a standardized bias (R) (Equation 4.3) and a standardized mean square error (V) (Equation 4.4) to test goodness of fit and evaluate visual comparison simulated data corresponding with observed data for dynamics of rice production.

N E (S -0) R = N (4.3) zoi i=1 N E (Si - 0)2 V = =1N1 (4.4) 0? i=1

Where N number of field observation, Oi and Si are observed and simulated values, respectively. At the ith observation R and V are estimate for the overall error of the method with regard to field data. R quantifies the model's ability to reproduce the observed yield pattern. Negative deviations (Si - Oi < 0) compensate for positive deviations (Si - Oi > 0) and vice versa (Equation 4.3). On the other hand, V is a measure that reveals the model's

42 tendency to generally overestimate or underestimate simulating field observation. However, both procedures give heaviest weighing to large values. For the HYDRO-CROP model to be properly validated, the simulation results need to be tested against observed data. We validated the simulated yield results from the HYDRO-

CROP model by using experimental results from Praputthabaht Field Crop Experiment Station, Saraburi Province. The experiments were conducted from 1989 through 1994 during the wet season under irrigated conditions. For the rainfed condition, we gathered historic data on average aggregate rice production in the non-irrigated areas of Saraburi and Lopburi Provinces, reported from 1985-1995. We performed a number of statistical tests to compare simulated and real-world data. Bias and Root Mean Square Error (RMSE) in Equations 4.1 and 4.2 also used as measures of model performance. In addition, goodness of fit was evaluated visually and by computing a standardized bias (R) and a standardized mean square error (V); the formulas are given in Equations 4.3 and 4.4. The results of these are discussed in Chapter 5. Mankeb (1993) calibrated the genetic coefficient for rice cultivar RD7 by using data from yield trial experiments which used RD7 from San Pa Tong Rice Experiment Station, San Pa

Tong, Chiangmai Province (18'37N and 98'54'E). Date on weather, soil, nitrogen fertilizers, irrigation data and management practices were entered into the CERES-Rice model. The minimum requirement for daily weather data includes solar radiation, rainfall, maximum and minimum temperature. Simulated and observed tiller numbers [m- 2], LAI, and above ground biomass were compared to assess the model's ability to predice rice growth over time. The accuracy in predicting the phenology of selected rice cultivars in Thailand's condition was statistically acceptable [34].

4.1.3 UTILITY

The final issue we consider in our simulation is the welfare or utility provided by on-farm water harvesting and irrigation system. Utility function notation may be used to describe how the farmer values the yield obtained from a rice farm. Generally, higher income means higher utility. However, revenue can vary depending on rainfall and other factors. This variability subjects the farmer to risk. Some years income may be lower than average and some years it may be higher. A risk-averse farmer maximizes his

43 expected utility; he sacrifices high income in order to reduce variability and risk. Our study analyzes a typical small subsistence rice farm. Few of these farms have enough savings to smooth out their consumption during financial shortfalls caused by low agricultural productiv- ity. Income generated from rice production is devoted mainly, if not all, to consumption and agricultural inputs. We can compare the expected utilities of consumption from both irrigated and rainfed cases. If utility from the irrigated case is greater, the pond is welfare improving. In practice, a pond incurs maintenance as well as operational costs (for instance, electricity for pumps.) The UTILITY module, in addition to assessing impacts of the on-farm pond on farmer's utility, estimates the maximum cost which the farmer is willing to pay to maintain the pond. Sensitivity and break-even analyses help to identify the extent to which the on-farm pond scheme is a sound investment for a typical farmer in our study.

4.2 Hydrologic Models - HYDRO

4.2.1 Concept

The purpose of this submodel is to investigate the possibility of supplying small on-farm ponds exclusively with water harvested from runoff generated within the farm boundary. The concept is to divert local runoff to the pond and to rice paddies and ditches, where it is stored for later use. McLaughlin et al. (1999) tested the New Theory project in Saraburi Province and suggested that it is indeed possible to grow a reliable rice crop in the Saraburi region without supplemental water from a large reservoir. The report provides a quantitative water balance analysis of on-farm water harvesting; which serves as a fundamental framework for our hydrologic model (HYDRO) in this study. Note that our earlier study did not look into the effect of the pond on crop yield or into the economic feasibility or risk [35].

4.2.2 Farm Configuration

HYDRO keeps track of water entering and leaving the pond and surrounding ditches in a rice- only farm. This is illustrated in the simplified sketch of Figure 4-2. The arrangement and relative sizes of different land uses (i.e., non-farming, pond, paddies, and ditches) are similar to those found in local rice farms in Saraburi. The rice paddy is excluded from HYDRO since

44 paddy fluxes are computed by the crop simulation model (CROP).

I itcheg / Rice/pond pond spills housing

Irrigation

Runoff from I paddy II Tb II rice paddy rice paddy 'I rice paddy / 01* 4-- W,. Off-farm (down gradient)

Figure 4-2: Plan view of model farm showing rice paddies with surrounding ditches. Arrows represent important horizontal fluxes.

In order to understand the movement of water as shown in Figures 4-2 and 4-3, we should begin during harvesting time when the water left in the paddy area is diverted to the pond. In the following dry season, the paddy is kept fallow and parts of the surrounding bunds are breached to allow water to drain out. Precipitation on the paddy during the fallow period is partitioned into infiltration, evaporation, and runoff to ditches which encircle the rice fields. Ditches receive water from both runoff and rainfall, and lose water to evaporation and infiltra- tion. If ditches are filled, though not likely in the dry season, water will spill over to pond. The pond captures water from runoff generated in the non-crop area, precipitation, and spills from ditches in addition to the water diverted from rice paddy during harvest. The pond loses water to evaporation and infiltration. In most years, the pond can maintain some depth throughout

45 Pond and Non-crop Area

Precipitation Evaporation

Rice irrigation Runoff from frcm rice non-crop area Spills to and ditches

*4m

storage _Pond I 4 V Groundwater recharge

Rice and Rice Ditches

Precipitation

Spills from Evaporation paddy

fromfrrigation pond Spills to and from irrigation pond Q fro n ditches Puddle I

30 cm Ditch storage Paddy

Groundwater recharge

Figure 4-3: HYDRO farm configuration: Vertical cross-sections through pond and rice paddy water storage compartments showing fluxes into and out of each compartment. 46 the dry season. The monsoon can come as early as late April (Figure 1-2). The pond starts to collect water and almost reaches its full capacity at the onset of the rice growing season. According to local practice in rice cultivation, the bunded paddy is maintained at a depth of 45 mm while the bund is 120 mm high. Whenever the paddy depth falls below the minimum level, farmers pump water first from ditches and then from the pond as the last resort. This operational rule is followed throughout the growing season; it is violated only when the combined water in the pond and ditches is insufficient to maintain the desired water level in the paddy. When pond and ditches are both full, the excess water spills off-farm. Finally, water is drained from the paddy to the pond at the end of the growing season to prepare for harvesting. The entire routine then comes to full circle.

4.2.3 HYDRO : Inputs and Outputs

Inputs for the hydrologic submodel (HYDRO) can be categorized into three sets. First, HY- DRO requires daily weather information such as precipitation and pan evaporation. These meteorological data are also shared with the CROP module. Next are water inflows and out- flows computed by CROP. These include (1) surface runoff from the non-crop area; (2) spill from the paddy; and (3) water required for irrigation to maintain the specified operational rules. The third set of inputs is the land use allocation. We assume a fixed ratio of paddy to ditch areas and a fixed area allocated for housing and walkways (non-crop.) The pond can vary in size. Evaluation of performance for each pond size entails a separate HYDRO run. HYDRO records daily storage in the pond and ditches under the default irrigation rule. The HYDRO submodel automatically indicates whether water stored in the pond and ditches is enough to sustain the irrigation rule for every daily step. If a water deficit occurs the irrigation rule must be modified and both HYDRO and CROP rerun, as discussed in Section 4.1.1.1

1HYDRO was written in MATLAB language; the codes are shown in Appendix A.

47 4.3 Crop Simulation Program - CROP

One application of crop simulation programs is to enable policy makers, farmers and other decision makers to assess the consequences of actions today on outcomes in the future. By allowing users to explore the future, a model enables them to choose those actions which are more likely to produce desirable outcomes. Here, crop simulation allows us to predict grain yields under irrigated as well as rainfed conditions by controlling all other factors. Many crop simulation programs are available commercially for agricultural research and development projects. Here we describe the crop simulation package used in this thesis; then we discuss our experiment file CROP, set up in DSSAT format, which simulates the potential values of the on-farm pond investment.

4.3.1 DSSAT Software

Decision Support System for Agrotechnology Transfer (DSSAT) is a collection of crop models and computer programs integrated into a single software package in order to facilitate the application of crop simulation models in research and decision making. This software is a product of the International Bench-mark Sites Network for Agrotechnology Transfer (IBSNAT) project and is widely used in agricultural planing and agronomic research. The DSSAT system consists of several components: (1) crop models; (2) soil and weather data; (3) collaborators' experimental data; and (4) application programs to enter and retrieve data, link the models with site and experimental data files, and analyze the observed and simulated data for specific objectives. Figure 4-4 illustrates the linkage between the crop model and the collaborator experiment, weather, and soil data; and the programs for providing graphical outputs and data analysis results to collaborators. The input and output files for crop models in DSSAT are arranged to facilitate comparisons with field data and sensitivity analyses. For example, different weather, soil, cultivar, planting date, irrigation management, row spacing, and fertilizer management can be changed. Simu- lated results can then be plotted from any of the runs for comparison with real experimental treatments or for evaluation of hypothetical treatments [24]. The crop model of interest here is

48 CERES-Rice 2 , which is a part of DSSAT and has been tested over a wide range of environments by the IBSNAT and its collaborative institutes.

The CERES-Rice was developed by International Benchmark Sites Network for Agrotech- nology Transfer (IBSNAT) [47] and modified for transplanted rice by researchers at the In- ternational Fertilizer Development Center (IFDC) [20]. The model can simulate the yield of different rice varieties for different management and agroclimatic scenarios. The following are the main processes simulated in the model:

" Growth and development.

" Biomass production and partitioning

" Root system dynamics

" Effect of soil water deficit and nitrogen deficiency on the photosynthesis and photosynthate

partitioning in the plant system.

" Water and nitrogen budgets at the surface and in the root zone

The model assumes that growth limiting factors such as weeds, insects, diseases, and nutrient deficits are completely controlled. CERES-Rice requires the following inputs:

* Daily weather, at least for the duration of the cropping season, which includes solar

radiation, maximum and minimum air temperature, and precipitation

" Soil initial conditions and properties

* Management practices such as cultivar, plant density, planting date, irrigation, and fer-

tilization

" Latitude of the production area to evaluate day length during the cropping season

" Variety-specific coefficients that account for differences in the response of different geno-

types to environmental factors.

2 CERES is an acronym for Crop Environment REsource Synthesis.

49 Data Generating Soil Crop-Soil Collaborators Characterization Modeling Collabordors Collaboraors

I I I

Expe ,mon, -*p SUrriMWY D

4

|CropMode,

11v

-oCollaborators

Figure 4-4: Elements of the Decision Support System for Agrotechnology Transfer (DSSAT) 50 4.3.2 CROP : Experimental Setting

The CROP submodel is actually a collection of CERES-Rice inputs. Two different types of CROP simulations are considered in this thesis: (1) rainfed cultivation and (2) pond-based irrigation. Rice is cultivated in a paddy setting with the same cultivar, soil properties, weather, and other crop managements in each scenario. DSSAT allows us to carry out simulations of crop sequences which alternate between rice and fallow conditions. CROP keeps track of soil moisture depletion during this sequence, rotating through 35 years (1961-1995) of actual weather data.

4.3.3 CROP : Inputs and Outputs

In our simulations, all CERES-Rice inputs are based on data from the Saraburi study site. According to Somrith (1996), the most popular rice genotype grown in this area is RD7. The genetic coefficients for RD7 were calibrated to fit actual experiments for paddy rice in Thailand by Mankeb (1993). Crop management inputs such as planting date, fertilizer application rates and harvesting date also follow local practice. Figure 4-5 shows the cultivation and growth stages for our case study. High fertilizer inputs are imposed to ensure that growth of stressed plants was depressed by drought only. A basal application equivalent to 100 kg N, 40 kg P, 40 kg K ha- 1 , is mixed into each plot one day before transplanting. Additional ammonium sulphate equivalent to 60 kg N ha' is added at mid-tillering and panicle initiation and 40 kg N at flowering. Harvesting is set at 100 days after transplanting to allow room for delay in maturity for some dry years. CROP produces four major outputs: (1) dates describing phasic development or growth stage duration, (2) biomass production, (3) water and chemical balances, and (4) water and chemical deficits. Water fluxes and storages are used to primarily determine the convergence of the HYDRO-CROP integrated model. The converged HYDRO-CROP model delivers a series of simulated annual yields for each pond size. These yields per unit area are then translated into total yield by multiplying by the paddy area, which is equal to total land holding area less housing and pond. Therefore, a farmer faces revenue loss for each unit of land he allocates to the pond. On the other hand, his increase in revenue from higher yield per unit area may sufficiently offset or even exceed that from the loss of cultivation land due to the pond.

51 panicle seedling initiation flowering maturity 25d 27d 34d 29d GROWTH STAGES

transplanting harvesting PLANTING N fertilizer applications in paddy

Figure 4-5: Timeline for rice planting processes with respect to growth. The period of each phasic development is averaged over 35 simulated cropping seasons.

With results from HYDRO-CROP run, we can plot average yields vs. pond size. Once a proxy pond size is selected, two series of annual yields-one from an irrigated plot and the other from a rainfed plot-are passed on to UTILITY which transforms yield to farmer's utility

(see diagram in Figure 1-5.) The recommendation is then made about the adoption of this on-farm surface storage. The sensitivity and break-even analyses will be discussed in the next section. 3

4.4 Utility Evaluation - UTILITY

Economic theory states that an individual acts to maximize his utility or welfare. The farmer's utility in our case depends on crop revenue and his attitude toward risk. We can estimate crop revenue from crop yield and the local market price of rice. Income in its own right gives no direct utility. It is rather only when this income is spent on consumption goods that any utility results. Our surveys indicated that the farmers in our study area barely keep saving accounts and devote most of their income to consumption. For this reason, consumption and income are essentially equivalent in our study. Farmers are perceived to be risk averse. That is, they are willing to accept less income in order to avoid risk. In the context of agriculture, farmers are willing to invest in water storage infrastructure if such investment can sufficiently reduce the uncertainty of crop yields.

3 Details of the CROP submodel are given in Appendix B.

52 4.4.1 Constant Relative Risk Aversion

We need to assume the nature of the farmer's attitude toward risk in order to see how he trades off between risk and income. Economists make frequent use of the constant relative risk aversion (CRRA) function, which describes how utility (U) depends on revenue (R)

U(R) = . (4.5) 1 - -Y The degree of curvature of this function is determined by the value of -y. The more curved the function the more the farmer is willing to pay to insure against risk. The coefficient of relative risk aversion is defined as the percentage change in the marginal utility of income or consumption arising from a one percent change in the level of income or consumption, i.e. it is equal to the following expression:

S RU R) (4.6) U' (R)

If y = 0, then U(R) is linear and the farmer is said to be risk neutral.

4.4.2 Expected Utility

We are considering evaluation of the mean utility when the revenue is a random variable. Any differentiable utility function U(R) can be expanded about the mean revenue as follows:

2 2 OU(R) - 8 U(R) (R -R) U(R) = U(R) (R -R )) (4.7) OR R 2

Equation 4.6 can be substituted into this expression and higher-order terms can be truncated if the revenue is Gaussian with a sufficiently small variance. The mean utility is then:

U(R) =U(R) 2 U(R) (R - R)2 OR (R2 2 O2 U(R) 0.2 = U(R) + R R OR 2 2 U'(R) 0.2 = U(R) - Y R(4.8) R 2

53 - -7y R 2

= ~- -(4.9) 1- 2 R?

Since U'(R) in Equation 4.8 is always positive, this expression shows that mean utility is reduced by a factor proportional to a2, the variance of the revenue. This variance is often called the income risk. As the income risk increases the farmer's average utility decreases. This effect is greater for larger 7. In our study, the on-farm pond increases the likelihood of sufficient water supply during dry spells. This lowers the income risk. When the farmer's risk measure (X) is higher he is likely to invest in an on-farm pond. Of course, the pond also has a maintenance cost. We can disregard the construction cost for now as all pond construction is currently subsidized by government agencies. This cost will be taken into consideration when we interpret results. Factors relevant to the utility evaluation process include the price of rice production, the cost of the pond, and the farmer's attitude towards risk (see Equation 4.9). These are all considered in the UTILITY submodel.

54 Chapter 5

Results

We now examine the results obtained by the methodology outlined in Chapter 4. To ensure the accuracy of these results, we carry out a model validation by comparing our simulation outputs with actual observations and with results from the literature. We also assess the implication of the simulation results for investment in the on-farm reservoir scheme. The optimal pond size is identified and we discuss the effect of a pond investment on the farming household. Our results indicate that the on-farm water harvesting and storage strategy advocated in the New Theory is technically and economically viable for a small-scale subsistence farm. We perform sensitivity analyses to identify range of conditions for which the pond investment is attractive.

5.1 Simulated Yields

5.1.1 Statistics

The HYDRO-CROP model produced simulated yields given 35 years of meteorological data, crop and soil properties for both rainfed and irrigated conditions, for various pond sizes. The statistics of simulated yields, recorded in Table 5.2, indicate average yield per unit area increas- ing with pond size until the pond is expanded to 5% of the total area. As a typical land-holding area of Saraburi farm households is 15 rail, or 2.4 hectare, a 5% pond requires 1200 square

'Rai is a Thai unit for land area. 1 rai is equal to 1600 square meters or 0.16 hectare.

55 meters of land. This means an on-farm pond at 5% can adequately provide for the default irrigation rule, specified in the HYDRO-CROP modules. A 5% pond is used as the proxy size to represent the forecasted yields gained from the on-farm scheme. First, a 5% pond sufficiently provides for the default irrigation rule; which means it maximizes expected yield per unit area. Secondly, a farmer should not allocate a larger area than 5% for a pond. With a larger pond, he would forgo his precious cultivation area without gaining higher yield per unit area. His average revenue from rice production immediately decreases as the marginal loss from smaller rice fields outweighs the marginal gain from a pond. However, even if a 5% pond produces the highest expected revenue, it does not necessarily maximize utility since it does not account for risk. Figure 5-1 illustrates histograms of yield per unit area for both irrigated and rainfed plots. These are derived from the sample of 35 years simulated for each alternative. The dotted lines represent a normal probability density fit to each histogram. The irrigated plot not only generates a higher average yield, but also shows lower yield variability than the rainfed plot. It follows that the on-farm reservoir scheme reduces risk on rice production. This effect is even more apparent in Figure 5-2a, which plots the two yield cumulative distribution functions. Yields from the rainfed plot never exceed those from the irrigated plot and there are only a few years that both plots give equal yield. Figure 5-2b shows the cumulative distribution function of the increase in yield obtained when the on-farm reservoir is in place. Note that one-fifth of the 35 annual simulated yields exhibit no improvement in rice productivity. Such events occur in extremely wet growing seasons. In these years the paddy does not need supplementary irrigation from the pond. It is worth noting that these results consider only yield per unit area, not the total rice production. The revenue from rice production will be discussed in section 5.2.

5.1.2 Model Validation

In terms of model validation, the rainfed case performed poorly but the irrigated case performed exceptionally well. The HYDRO-CROP integrated model overestimated grain yield across years of available data by an average of 8.3% and 68.8% (bias of 471 and 2,048 kg/ha; and root mean square error = 625 and 2,198) for irrigated and rainfed plots, respectively (Table 5.1).

56 7777

20 irrigate

15 -

C-, Cr 10k 1 ' '4- . 5 00 10F.-'9.* 0 3000 3500 4000 4500 5000 5500 6000 6500 7000 7500 8000 kg/ha

20, I C3 rainfed 15-

C I

C. r CD

.*0 5 *d-

0 3000 3500 4000 4500 5000 5500 6000 6500 7000 7500 8000 kg/ha

Figure 5-1: Histograms to fit normal distribution of yield: rainfed and irrigated rice farms.

57 1 A 0.81

0.61 LI. 0.4

0.2 - irrigate< - rainfedl C 3000 3500 4000 4500 5000 5500 6000 6500 7000 7500 kg/ha

1 B 0.8

0.6 x 0.4

0.2

n "0 500 1000 1500 2000 2500 3000 3500 4000 4500 kg/ha

Figure 5-2: A: CDFs of yield per unit area with and without irrigation. B: CDF of yield increase with pond installation.

58 Although both management alternatives (rainfed versus irrigated) exhibit positive deviations (Si - Oi > 0) for all cropping season, the sharp difference in the magnitude of these deviations is obvious in Figure 5-3. Provided that all deviations were positive, V measures the model's tendency to overestimate rice production. The model's prediction performed more poorly in the rainfed condition due to a potential downside bias of the observed data. The irrigated experiments that we obtained from Praputthabaht Field Crop Experimental Station used fertilizer inputs identical to our inputs for the CROP module. Furthermore, the rice plants were protected from pest and disease thanks to regular weeding and spraying of pesticide. On the other hand, the large discrepancy between observed and simulated yield under the rainfed condition appears to be a result of uncontrolled fertilizer application. Information on rice yield from rainfed plots was adapted from aggregate rice production records from the non-irrigated area of Saraburi and Lopburi provinces. Although soil and crop properties and planting/transplanting dates in the CROP module resembled the local practice, information on fertilizer applications was not available and reportedly non-existent. In addition, there was a history of pests and diseases destroying rice production in parts of the survey area. For these reasons, we can expect substantially higher estimates from the simulated data than from the observed data. In spite of these uncontrolled factors, the observed yields had a lower variation than the simulated ones. Although the HYDRO-CROP simulation did not account for pest and disease risk, the simulated result follows the same trend as the observed result. This is possibly because soil nutrient limitation crippled plant development. Since the simulation does not account for these limitations it may be more sensitive to variability in rainfall, which the primary factor controlling yield in the simulated case.

5.1.3 Yield VS. Pond Size

We separately ran HYDRO-CROP modules for different pond sizes; the simulation results are reported in Table 5.2. At 5% of total area, the pond can provide the irrigation water required for the default operation rule. Thus, there is no need to simulate for a larger pond. A larger pond will only give yield as high as a 5% pond. Provided that there is no other benefits accrued from pond (e.g., fishery), a farmer should not sacrifice more cultivation land for the pond area.

59 Irrigated

700c simulate + observej Cu 600(

0) 500

*0 a) 400

300c I I I I I I I I 5 86 87 88 89 90 91 92 93 94 95

Rainfed

700( I- Cu 600C -c 0) 500C *0 *13 400C

300(

200g 5 86 87 88 89 90 91 92 93 94 95 Year

Figure 5-3: Simulated and observed grain yield for irrigated and rainfed condition.

60 Figure 5-4 illustrates an attempt to fit a continuous curve to average yields obtained for several different pond areas. As the average yield approaches the maximum level, a saturation function of the following form seems appropriate for curve fitting:

Y = (Ymax - Ymin) - (1 - e ax 2 ) + Ymin (5.1)

where x a percentage of pond over total area Y rice yield per unit area

Ymax = 6, 123 kg/ha Ymin = 5,277 kg/ha a = 0.155. This continuous curve is useful for the sensitivity analysis in section 5.3.

5.2 Implications for Investment

Now that we can forecast reasonably accurate yields from actual meteorological data, we can judge the attractiveness of on-farm pond investment. In order to do this, we need to consider all parameters involved in utility maximization: (1) the rice price, (2) the optimal pond size, (3) the farmer's risk aversion measure, (4) pond construction and maintenance cost, and (5) the revenue required for subsistence. The price of rice production is necessary for translating rice yield into revenue. For practical purposes, the optimal size of an on-farm pond is chosen for a typical rice farm in the study site. The risk aversion measure can be estimated from observations of farmers' behavior. Construction and maintenance costs can be amortized over the time horizon of the pond's usage. Subsistence constraint is defined as the income level sufficient to sustain a household.

5.2.1 Price of Rice Production

The price of unmilled rice varies depending on existing supply in the market as well as on proximity to a higher-priced market. In most cases, small-scale farmers sell their rice production to the local millers because the transportation cost to town markets exceeds potential profit gain [50]. Despite possible fluctuations of the rice price across time, we want to keep hydrologic

61 620C-

0 610C alpha =0. 155 600C

CU .- 5900

0) CU 5800

_0 5700 a)

560C

5500

5400

5301 1 2 3 4 5 6 7 8 9 10 pond size %

Figure 5-4: Plot of yield per unit area versus pond size from our model simulation. The continuous curve fits the plot with respect to equation 5.1.

62 variability as the focus of our study. Consequently, we assume a constant price. Rice was sold to local millers at 4,300 baht per metric ton for Saraburi province in 2000 under the guaranteed price regulation which is still in effect today [23]. That is, farmer's income from rice production is guaranteed at 4.30 baht per kg unmilled rice. This is the price used in our analysis. However, an IRRI document stated that before devaluation of baht in 1997, the farm price was 4,800 baht per ton and after, 7,600 baht [26]. From our interviews with local farmers in 1999, farm price ranged between 4.30 and 5.00 baht per kg of unmilled rice depending on the market supply. During a high production season, the price of rice, like other commodity goods, tends to be at the lower range. Thus, we choose the lowest price possible (4.30 baht per kg) to estimate the lower bound of price range at which the pond scheme is economically feasible.

5.2.2 Optimal Pond Size

The average size of land holding areas for subsistence farms in Saraburi province is 15 rai, or approximately 2.40 hectares. Housing and roadways cover 5% of total land; whereas, ditches surrounding the paddy fields constitute approximately 5% of the total cultivation land left after subtracting non-crop area. Hence, the paddy area listed in Table 5.2 is calculated using the formula:

(95 - x) A = 2.4 x 0.95 x 100 (5.2) 100 where x is the percentage total land holding area devoted to pond

For the rice price specified above, expected yield estimates from Equation 5.1, and paddy areas from Equation 5.2, we can plot a continuous curve of average revenue from rice production as a function of pond size (see Figure 5-5). The maximum revenue under the on-farm reservoir investment from our calculation is approximately 54,000 baths per rice season. The optimal pond size, corresponding to the highest expected revenue, for a typical 15-rai rice farm is 4.22% or 0.63 rai. As discussed earlier, a farmer values both average income and low income variance. The

63 5.x

5.4-

0 u 5.3- Cl) 0. 0

S5.2-

5.1- 0

5-

x*= 4.22%

"0 1 2 3 4 5 6 7 8 9 10 pond size (%)

Figure 5-5: Average revenue with respect to pond size. Optimal pond size is at the peak of curve.

64 coefficient of relative risk aversion (-y) estimated by Paulson and Townsend (2001) for households similar to those in our study area is y = 0.671. The relationship of expected utility of revenue U(R) for this value of -y is plotted versus pond size in Figure 5-6. The utility-maximizing pond is slightly bigger than the revenue-maximizing pond due to the effect of lower revenue variability with a larger pond. The optimal size is 0.645 rai, or 1,032 square meters. However, when we ran a HYDRO-CROP simulation with a pond size of 4.30%, the simulation result yields a lower average utility than that of a 5% pond (see Table 5.2.) A blue square in Figure 5-6 corresponds to the expected utility of a 4.30% pond, which is significantly lower than the value estimated from curve. Accordingly, the optimal pond size lies somewhere between 4.30%

(1,032 M2 ) and 5% (1,200 m2 ). For simplicity, we choose a pond size of 5% as our candidate to represent the on-farm pond alternative. We can then carry out a detailed comparison between performance with this pond size and performance with no pond.

5.2.3 Risk Aversion

With a pond of the optimal size of 1,200 square meters, we can use our HYDRO-CROP model to forecast yields for 35 years of actual weather data. After translating into baht terms, the revenue probability density and cumulative distribution functions are plotted in Figures 5-7 and 5-8a, respectively. In a few high rainfall years the rainfed alternative actually provides higher yields and utility than the irrigated alternative because more land is available for rice cultivation. Although the on-farm pond improves average revenue, the increase appears rather insignificant. By considering levels of expected revenue alone, investment in a pond seems unattractive. However, an individual's decision is based on utility rather than revenue. The cumulative distribution function of the farmer's instantaneous utility is plotted in Figure 5-8b using the CRRA function in Equation 4.5. As marginal utility is higher when the utility level is low, the benefit of the pond is more pronounced at lower utility. Table 5.3 records statistics of simulated revenues from a rice production with a 5% pond and without irrigation. The expected gain from a 5% pond in comparison with that under a rainfed condition is only 4,878 baths per crop season. Despite the fact that the gain on expected revenue looks unimpressively small,

65 I I I I I I I

(1) I.3.

0

_0

x* = 4.30%

I I 0 1 2 3 4 5 6 7 8 9 10 pond size (%)

Figure 5-6: Plot of expected utility of revenue versus pond size. The curve results from interpolation of HYDRO-CROP simulated data (green dots).

66 the variance in revenue is narrower by 54%. Results of utility evaluation given farmer's E-V objective function shows that a pond improves the expected utility level as plotted in Figure 5-6. Hence, the on-farm pond appears to be a considerably more attractive investment once risk is taken into account. The results summarized above depend on several important parameters which are rather difficult to estimate. Furthermore, our study assumes that all variability in revenue is due solely to meteorological uncertainty. We need to look at uncertain parameters separately and determine the range of conditions under which this on-farm reservoir technique can be beneficial to a rice farm.

5.2.4 Pond Construction/Maintenance Cost

In order to generalize our analysis, we need to identify the maximum cost for the pond such that it does not become a bad investment. At this break-even cost (C), a farmer will be indifferent between investing in a pond installation and doing nothing. The nominal value of C may be considered as an annual payment for the pond; its total discounted sum across the pond's life includes both maintenance and construction costs. We can incorporate cost (C) into Equation 4.9 by replacing R with R - C:

R -C -y 2 U(R - C) (R C)Y - - R (5.3) 1 - -y 2 R - C Assume that the actual cost (C) for each pond size is constant for each pond size (x), the only uncertainty is in revenue from rice production. Therefore, the mean cost is equal to this constant cost so R - C = R - C and the variance of net profit-i.e., revenue less cost-is equal to the variance of revenue alone; that is, o-C = (. To obtain the break-even cost, we equate U(R - C) with the pond (x = 5%) to U(R - C) without the pond (x 0%). The result is equal to 5,326 baht for a 5% pond in spite of the fact that the gap between the average revenue under a pond scheme and that without pond is only 4,878 baht. This premium for a pond installation results from the risk aversion of the farmer. He is willing to pay more to insure against uncertainty.

67 15 -**"%. I I { irrigate

C **

4- 1-.

00S ---- 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 revenue (baht) x i d

C5 =3 13 rainfe

d-

cT

~ 5. - -

C 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 revenue (baht) x i d

Figure 5-7: Histograms of revenue under irragated and rainfed conditions. The dotted line represent estimated normal distributions using parameters derived from both sets of revenue.

68 irrigate - rainfed 0

10 .6-

0.4-

0.2--

.5 3 3.5 4 4.5 5 5.5 6 6.5 7 Revenue (baht), R

1.-- -- ir rigate ainfed

0.

0.4

0.2-

0 Utility

Figure 5-8: Cumulative distribution functions for revenue (top) and utility (bottom) from rice production under irrigated and rainfed conditions.

69 5.2.5 Subsistence Constraint

Each farm household faces a subsistence constraint on consumption and other expenses 2 . For the on-farm pond to be viable, the revenue under the pond scheme must be sufficient to cover both its cost and the necessary amount to sustain a household. Sensitivity analysis indicates that pond size and the expected utility level are independent of the changes in subsistence constraint, but the maximum price of the pond farmers are willing to pay is linearly affected.

5.3 Sensitivity Analysis

Sensitivity analysis enables the analyst to determine how changes in parameters affect simula- tion results. In our application, sensitivity analysis identifies the range of parameters for which the pond scheme is viable. There are a number of parameters on which it is worthwhile to undertake sensitivity analysis. Some of these are very uncertain while others are likely to have a significant impact on our conclusions regarding the attractiveness of an on-farm pond. For sensitivity analysis purposes we specify as a nominal case the 5% pond option discussed on Section 5.2. Three parameters are selected for our sensitivity analysis: the risk aversion measure, the rice price, and the income threshold for a subsistence household. Table 5.4 records the changes in these parameters; while Table 5.5 reports responses in utility level and the break-even cost at which the pond scheme is an attractive investment ceteris paribus. The results of sensitivity analysis indicate that utility is highly sensitive to the risk aversion coefficient. However, the maximum price the farmer is willing to pay only shifts slightly when this coefficient is changed. The rice price has an insignificant effect on the utility level but a strong effect on the break-even pond price. Although the farmer's utility is insensitive to the changes in the subsistence constraint, the maximum pond price is highly and conversely

2While the farmers wait for the harvest, they spent 3,500-4,000 baht per rai over the past four months on water pumping, soil preparation, seed growing, pesticides and fertilizer [23]. Somrith (1998) estimated total agricultural expenditure in the central region at 3,645 baht per rai. Therefore, a typical rice-only farming household with a 5% pond and a 5% housing area keeps 90% of total area (13 rai) for rice fields. They require 46,700 baht for aggregate rice cultivation activities. In addition to agricultural inputs, a family of four eats 12 kilograms of milled rice a month. This consumption amounts to 144 kg of milled rice annually; price of which lies in 10 baht per kg [26]. Rice consumption constitutes about 75% of food comsuption; as a result, total household consumption sums up to 1,920 baht per household annually. Because farmers need to purchase the rice after milling, we cannot simply subtract the rice out of their dry grain inventory. In sum, each household needs to maintain an income of 48,500 baht annually.

70 correlated to such changes.

5.3.1 Risk Aversion Parameter

Without considering cost, the average revenue under an on-farm pond is higher than that without a pond while the variance is lower (see Table 5.3.) Consequently, a farmer who has a normal upward-sloping utility function will prefer a pond regardless of his attitude toward risk. When the pond cost is included, risk becomes more relevant. The farmer's risk aversion coefficient measures his sensitivity to changes in income. These curves in Figure 5-9 indicates that a farmer with higher risk aversion measure is more sensitive to a small change in revenue since the slope of the curve is greater. Figure 5-10 illustrates that a more risk-averse person is willing to pay more for the pond in exchange for a lower variability of revenue. A risk-averse farmer is willing to pay the difference between the break-even cost and the average revenue in order to reduce risk (in the other word, income variance). A highly risk averse person values reduction in variance substantially. Nevertheless, he would not want a pond larger than 5%, as the decline in risk becomes insignificantly small. The optimal pond size remains within the same range of 4.30% - 5% despite the change in risk aversion measure.

5.3.2 Rice Price

We can incorporate a constant price multiplier (p) into Equation 4.9 as follows:

U'()S- U(pR) U(pR) -2 pR 2

I 7 pR p-Y - U(R) (5.4)

Since the price multiplier (p) is independent of pond area (x), x remains unchanged regardless of p. Although the level of the expected utility increases only slightly when the price of rice

71 10% increase gamma = 0.6 1 0% decreas

SI I I I I I 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 revenue (baht) x d

Figure 5-9: Impact of the risk aversion coefficients on utility.

72 ......

550C -e- average revenue -4e- break-even cost of p nd

530C-

520C A -c (U .0 51 W- premium paid for reduced risk 500-

490C

480(

0 0.2 0.4 0.6 0.8 1 1.2 multiplier of estimated risk aversion

Figure 5-10: Break-even cost of a 5% pond versus risk aversion coefficient

73 1100

1000--

2 900G-

C o 8000-

0. -6 0 0 E .E 600(- E 500(-

400G-

300A .6 0.8 1 1.2 1.4 1.6 1.8 2 multiplier of the estimated price

Figure 5-11: Relationship between the maximum cost of a 5% pond and price of rice production. Here we assume the price of rice equal to 4.3 baht/kg.

rises, the relationship between the price multiplier and the maximum cost (C) is positively and linearly correlated as illustrated in Figure 5-11.

5.3.3 Subsistence Constraint

The results presented in Table 5.3 indicate that there is a nearly 50% chance that a rainfed farming household will suffer a financial crisis. This risk explains urban migration and the cultivation of second crops during the dry season. On the other hand, the expected income from an irrigated farm is greater than the required expense by more than one standard deviation. The income from rice less production cost and consumption (5,400 baht) is still higher than

74 the maximum price a farmer is willing to pay for an on-farm pond (5,326 baht). With a lower pond cost, subsistence farmers attain a higher utility under the on-farm reservoir plan.

75 Table 5.1: Simulated and observed grain yield for irrigated and rainfed conditions. Year Grain Yield [kg/ha] Irrigated Rainfed Simulated Observed Simulated Observed 1985 5915 2792 1986 5550 2900 1987 4651 2374 1988 5799 3414 1989 6329 6222 4112 2934 1990 5944 5582 5196 3119 1991 5700 5548 4694 3002 1992 5349 4693 5349 2998 1993 6240 5543 5899 3278 1994 7151 6898 3070 2500 1995 5988 3991

Mean 6119 5648 5111 3027 Bias (eq 4.1) 471 2084 RMSE (eq 4.2) 625 2193 R (eq 4.3) 0.0834 0.6883 V (eq 4.4) 0.0120 0.5150 Source: Irrigated - Praputthabaht Field Crop Experiment Station Rainfed - Central of Agricultural Information, 1998

Table 5.2: Area allocated for rice paddy in an average Saraburi rice farm with respect to percentage of pond area Pond Size (x) Paddy Areaa (Ap) Average Yield (Y) Revenueb (R) [% of total area] [hectare] [kg/ha] [baht] 0.00 % 2.1660 5,277 49,147 2.00 % 2.1204 5,572 50,804 3.00 % 2.0976 5,921 53,405 4.30 % 2.0680 6,042 53,728 5.00 % 2.0520 6,123 54,027 6.00 % 2.0292 6,123 53,427 a total land-holding area is 2.40 hectare or 15 rai, with 5 % housing b price of rice production is 4.3 baht per kg at local market

76 Table 5.3: Statistics of simulated revenue for rice production, generated from 35 years of actual weather data (1961-1995) Statistics of Rainfed Plot Irrigated Plot 5% Revenue without cost with maximum costa [x104 baht] [x104baht] [x104baht] R 4.915 5.403 4.870 RN/2 5.124 5.337 4.804 Rmax 6.693 6.341 5.808 Rmin 2.806 4.546 4.013 O'R 0.945 0.482 0.482 Skewness -0.612 0.405 0.405 a The maximum cost (C) of 5,326 baht is calculated according to Equation 5.3 with risk aversion measure of 0.671.

Table 5.4: Input values used in sensitivity analysis Parameter CRRA (y)a Rice Price6 Subsistence Constraint' Change [ I [baht/kg] [baht/year] -120% -0.134 -100% 0 -20% 0.537 3.44 38,800 -10% 0.604 3.87 43,650 0% 0.671 4.30 48,500 10% 0.738 4.73 53,350 20% 0.805 5.16 58,200 100% 8.60 Source: The base case uses a 5% pond in a typical 15-rai farm. a estimated in Paulson and Townsend (2001). b local price of unmilled rice in Saraburi. C explained in footnote.

Table 5.5: Summary of sensitivity analysis results Parameter Measures of Interest A Utility A Maximum Cost (C) CRRA (y) highly sensitive slightly sensitive Rice Price slightly sensitive highly sensitive (linear) Subsistence Constraint insensitive highly sensitive (linear)

Original Solution 109.51 util 5,326 baht

77 Chapter 6

Conclusions

In this final Chapter, we evaluate the on-farm reservoir investment in terms of technical and economical feasibility and discuss directions for future work.

6.1 Evaluation of On-farm Pond Installation

This thesis proposes an integrated approach for investigating the feasibility of investing in an on-farm reservoir. The said investment follows the general principles outlined in Phase I of His Majesty the King's New Theory concept; i.e., farming households allocate a certain area of land for a pond which captures runoff generated within the farm boundary. Furthermore, since the on-farm reservoirs are individually owned and managed, organizational problems inherent to larger irrigation schemes are absent. This study considers not only the pond's capacity to harvest rainwater for supplemental irrigation to rice farms but also its economic impact for farming households' welfare. Our case study is based on the physical and hydrologic characteristics of a typical small-scale rice farm in the semi-aria area of central Thailand. The simulation results derived here confirm that the New Theory concept is technically viable. The pond can store enough rainwater to provide for successful rice farms during the wet season. A farm with an optimal-sized pond has higher rice yields than a farm with no pond. However, this high rice yield comes at a cost. The farmer sacrifices valuable cultivation area for the pond. The improvement in total rice production and thus in the revenue is therefore less than the improvement in the yield per unit area. Despite the small revenue difference,

78 the variation of yield is reduced by half under the irrigated scheme. Given higher expected revenue and lower variance, risk-averse farmers find the on-farm pond alternative an attractive investment if pond construction costs are ignored. It is more interesting when the pond's cost is taken into consideration. We performed a break-even analysis to calculate the highest cost farmers are willing to pay for a pond given the increase in their expected utility. This cost is estimated for a nominally equal annual payment throughout the pond's life and includes both construction and maintenance expenses. The expected net profit from the pond is close to that of the rainfed case once cost is incorporated. Nevertheless, the on-farm pond is still attractive since the revenue is still significantly lower. Our sensitivity analysis computes the range of pond costs which farmers can afford. Nat- urally a more risk-averse person tends to assign higher value to an on-farm pond. As the level of farmer's risk aversion increases, the maximum price the farmer is willing to pay for a pond rises. Furthermore, even a risk-neutral farmer finds the on-farm reservoir an attractive investment due to a slight increase in expected revenue from rice production. Since we do not take price uncertainty into account, changes in the price of unmilled rice at local market have a linear effect on the price of the pond. In accordance with the results from our sensitivity analyses, the range of a pond's cost exceeds the actual cost of maintaining the pond. In sum, the pond alternative meets the welfare utility criteria for a potentially good investment. In topographically flat areas such as our study site, on-farm reservoirs are constructed as dug-out ponds and maintained by hiring local labor to clean up sediments. Currently all costs involving a pond's construction are subsidized by either government agencies or state banks in various forms. Farmers, in actual practice, only need to cover maintenance cost, which is relatively low compared to the estimated break-even cost. In reality, the reservoir projects have selected or targeted farmers with relatively less cash, land, and labor resources. As such, these projects appear to have the positive strategy of improving small farmer equity. The overall approach described here should be applied to farms and watersheds in other parts of Thailand, including the more arid Northeastern region. Conclusions from the Saraburi site should not be generalized to the rest of Thailand until such investigations have been carried out.

79 6.2 Recommendations for Future Research

In recent years, several projects have applied the New Theory concept to various drought- prone locations. In spite of these advances, there is a lack of clear understanding of the range of physical, social, economic, institutional, and environmental factors that favor successful implementation of the on-farm pond system, and allow sustainable benefits to be achieved from use of the concept. The relevant issues are addressed below.

6.2.1 Multiple-cropping system

Several factors have to be considered in irrigation management, particularly in the case of multiple cropping. The relatively small gain in wet-season rice production from the pond installation is limited by a favorable weather conditions. Dry-season crops, on the other hand, can increase agricultural income with the help of supplemental irrigation from the pond. However, some pond adopters may not be able to plant a dry-season rice crop due to unfavorable agroclimatic conditions. A careful analysis of crop production is needed for each subecosystem. Multiple cropping, including cash crops and perennial trees, can also help to mitigate the risk of rice-only production and provide for domestic consumption. One of the key decisions to be made is how much water should be allocated to different cropped areas. The decision must be based on (1) the availability of land and water resources, (2) the reliability of the water supply and, (3) the expected benefit and uncertainty from crop production. There are two possible strategies for the application of water to the crops. The first is to apply irrigation water at a level which gives maximum net incomes. The approach may be used when there is no constraint on irrigation supplies. That is not the case in the semi-arid subecosystem in central Thailand where water is in short supply. When a water constraint exists, it is useful to provide alternative levels of irrigation water and thus cover a larger area, which may result in higher returns. In spite of an acute water shortage, a farmer may, in actual practice, irrigate more than required even for maximum production. This calls for optimum allocation and distribution of water along with scientific planning of cropping patterns.

80 Figure 6-1: Cash crops such as perennial fruit trees and vegetables can increase incomes for farm households, particularly during the dry season when water is insufficient to sustain rice.

81 6.2.2 Supplementary Water Resources to the On-farm Pond

Surface Reservoir

Different scales of water storage can jointly improve irrigation supplies for rice farms. Re- cent research results have shown that scientific management of water resources within a micro watershed, using an on-farm reservoir as the centerpiece, is economically attractive and so- cially desirable. Larger water storage facilities such as a communal reservoir similarly serve as supplemental water supplies. These supplies mitigate the risk of rainfall variation and per- mit dry-season cropping. A methodology appropriate for assessing the effect of alternative technologies is needed to establish the long-term feasibility and profitability to participating households. Combination of these storage facilities rather than a single one may be the optimal choice. Thai bureaucracies are in a transition from relying on large-scale hydraulic structures as the hope for all the countries' irrigation needs toward recognizing the need for a range of technologies and social organizations. Large-scale alternatives are not feasible everywhere. Small systems also cannot replace large systems in many circumstances, but large systems may also be inappropriate where many small systems are feasible. What is gained in engineering efficiency with the large system is sometimes lost in the complexities of actually managing large social systems. Small-scale alternatives which may be easier to manage, in practice, gain new currency.

Groundwater

Water balance analyses of Saraburi and other prospective sites for the on-farm reservoir scheme should include quantitative studies of shallow groundwater resources. The combination of an on-farm pond and a shallow groundwater pumping system for wider irrigation coverage should be assessed. In theory, constructing an unlined rainwater storage structure markedly improves groundwater recharge. Future studies should determine whether shallow aquifers in the region of interest could supply significant amounts of irrigation water and whether conjunctive use is more economically desirable than of either system used above.

82 0/0 ~0 I/ 0 9 * 0

0

0 __ * 0 0 0 big reservoir * 0 V

Figure 6-2: The complete system of the New Theory: A big reservoir supplies a small; a small reservoir supplies a pond. An experiment of this principle can be observed at Wat Monkol Chaipattana, Saraburi Province. Source: Chaipattana Foundation's The New Theory [9].

83 6.2.3 Opportunity Cost of Agricultural Labor

The on-farm pond technique increases agricultural income but also requires high labor input in farming and pond-maintenance. Labor inputs in crop farming conflict with the time required for other activities. A farmer pays an opportunity cost by choosing not to work in town where he can earn a fixed wage. Given the relatively less variable wage-earning labor, as compared to the agricultural labor, farming households prefer regular seasonal farm labor. We need to study the extent to which the on-farm reservoir technique changes the occupational choice of rural households so as to understand more realistic socio-economic impacts of this irrigated scheme.

6.2.4 Household's Wealth Accumulation Beyond Subsistence

The main objective of the New Theory concept is to improve the living conditions of rural and subsistence farming households. There is a need for economic impact evaluations of the on-farm ponds in order to establish the long-term feasibility and profitability of these reservoirs. Even though small ponds can provide irrigation water to temper the vagaries of the monsoon and thus mitigate the risk of water shortage, evidence shows that poor farmers fail to undertake profitable investment that they should, in principle, self-finance. Without subsidies from government, the non-divisibility of the pond investment puts it out of their reach. We need to look at credit structures that can encourage farmers to invest in desirable on-farm reservoirs. Once the financial infrastructure is in place, the farm household can attain wealth accumulation and thus a better standard of living.

6.2.5 Government-aid Programs

Cooperative Activities

Thus far, we have only discussed the benefit-cost structure on the on-farm reservoir scheme from an individual point of view. We also need to consider the aggregate effects of agricultural improvement under the New Theory. The second phase of the New Theory suggests that farm- ers put together their efforts and resources in the form of a group or cooperative to execute the agricultural activities. Cost-sharing activities such as soil preparation, irrigation management,

84 and product distribution can dramatically increase profit of each individual farming household. Government can help coordinate farmers who share the same pool of resources and establish such cooperative groups.

Financing

It is also necessary to identify under what conditions institutional credits are needed in order for small-scale farmers to benefit from on-farm ponds. Currently both on-farm and larger reservoir projects in Thailand are fully-funded by the government. Most areas who reject irrigation project do so because of the lack of financing. Small-scale irrigation systems are available only to economically advantaged farmer groups and very few subsidized communities. To make the system more applicable to other areas, government-aid programs may be needed to jump start the process and encourage its use by economically disadvantaged groups. For example, the Bank of Agriculture and Agricultural Cooperatives is considering a low-interest loan for a supplementary communal reservoir. The approach to this issue is to compare alternative financial options for funding the project.

Institutional Arrangement

The effectiveness of water management, including promoting adoption of new irrigation in- frastructure, requires a well-structured administrative body. In Thailand, there is overlapping of administration among the Royal Irrigation Department (RID), the Local Administration Department, and the water users association. The Chak leader elected by the water users association takes charge of water supply, drainage, and maintenance work within his area to ensure that water is allocated in an equitable manner. Although this Chak leader is the key

to water use management, he can be directed by the chairman of the water users association, by the zoneman of RID, or by the District officials of the Local Administration Department. This complicated structure results in lack of unity and brings about inefficiency in water man- agement. Further study of the institutional impact of the use of on-farm reservoir is needed.

85 99

------%

AOQI @q PO JI PH JH qTTnfo UIJDT2 GAPS

pug PuG '0(4'4O = sqn

'= (T+4) PO !PV*(T+;)PO = (4)40Tino ( 0(T4)P ' O

,do '( ';)xnTJ '(8'4)xflU; '(q)PH '-4)q;TP =E;2 '(T±;)PO '(T+;)'7H '(T+-4)PH] ,,3 (ql~xnJ =do 83TJ JO PUG % L 'E == (0001 '(4)x Oa)Pou JTa9TG !((-4)2H 'dV*(6';)xnTJ; 'X 'do '( 1-4)xn-[j '(8'4)xnT; '(q)PH 'qq4P= [(;)'I '(T+-4)PO '(T+-4)-H '(T+-4)PH] !(9-4xn-;= do UOS~S GOT3 % (L C > '(q)iH '( ';)xnTJ '(8';)xnTl; ';)puod = [X '(q)-TH] * (VT+q)H~qo'd = Iq

0C6ZT:T = 4 03:

! ( ': x TJ = )MCI TqxflJ TxnT9 aPGTO . 1 KvTxnug a.lxfTJ] = xflTJ !0= 4OTqtno

=0 Po !0 PH

pqEuooddq unj HIO~d 'Lqxfl[ 'TxflTJ % T304hm PPOT Uleul =X uoTqoufl

%M0S*TOT9drLDO Pup MNS* O0TdrIICQ '9661-T961 0 0% seT4qTP Pp puod go Tqdep A-;T-4u@PT 04 sindqno IXISSQI bUTsn *

OI3OD 9IVT~lVN - OHIH

V x-ipuaddV P03P @3J o apuoaad % !UdHOdO-DEHdSfOH-DEdONOd-OOT = DOddDIU bUTPTO4 PUPT aayqua oq Paap papqoao go abpuaoaad % !0 =3IddHD'dO bUTPTOq4 PuPI eJITcU@ 04l P~22 bU~snoq go e ;quaoIad % !9 = 3dWIdSflOH aZTepTJod PPOT

bUTTq PuPT au~qua Oq Pecu puod go abpuaojad % PdPg= DdHJQNOd %%

!009T = a uvbs UT UOT4POOTTP PuPT %

00000%%% ~ %%%0000000%%%%%%%%%%%%%%%%%%% oo%%%oo%%% %%%%%%%%%% %%000%%%0o0 % W'MVqddi 'dOq SINVISNOO Iq NDTISSV - W* ISNO~ddl %

£(,I:ArIn3 'PH) UTmU = PH !(O 'IUSAqflD-PH)xPul = PO Paip Apppd jed unu o; P~q~aAUOD % !dV/IT = 3

!= PH !2zv/I-2V/PVT PH+T aH = JIH

*PY/LlZ0tTPO = PH !T JH = 21H o =< qDTPO JT !I- PV*T PH = LJD4TPO {ATddns'PUPUIOPjUTUI UOT4P5Ta2T 90 eumTOA % !(IrwT 3H+PV*T PH 'dj)UTIu =I

!= PH o => PH iT (d+TPH.,TdS) -pV/dv*[ do+pv/JV*X+AE-U'd+T PH = PH

!Aa-pV/aV4XUE = PH

f,;suooddq4 una saqoq~p IApppd UT @3P~ aaqpm% (T J~H 'dI 'X 'T do 'AH 'UE~ 'T PH '4)qO4TP = [II 'PO 'ZH 'PH] UOTqoufl

!'(o '3H) x~uI = EH

!(HJdECCQNOd ' H) uTUI = a

(0 'H~dSGGNOd -. H) XeU = X PUG

!0 =>21H JT I (Zds+Z T JH*Tds)-a/q* qC+aV/PV*T P+A9-U+T J~H = 21H

!(0'AE-UU) UTUI = rH T==q JT qsuooddq una 2TOAJG9ea UIP2-UO UP UT a~Pp 2[9Wm % (T LlO 'TIPO 'T 'H 'A luEj ';)puod = [X 'aH] uoT;Dunq Ap = .95*RICEPERC*rai; % rice paddy Ad = RICEPERC*rai - Ap; % ditches around paddy Ah = HOUSPERC*rai; % housing & road

% surface area and capacity of each reservoir in sq.m. and cu.m. Ar = PONDPERC * rai;

% land elevation in mm CULVERT = 300; % drain from paddy ditches to pond PONDDEPTH = 4000; % reservoir depth RICEROOT = 400; % root zone for rice, top layer

K = Ar*PONDDEPTH; % pond capacity

% Seepage coefficient in reservoir/ditches, Sp(t) = Spl*depth^2 + Sp2 [mm/day] % constant throughout the year due to assumption that there is water. Spl = 0.49e-6; Sp2 = 0.005e-8;

% embankment for paddy fields varies on different stage of rice (mm) MINDIKE = 45; MAXDIKE = 120;

88 Appendix B

CROP - DSSAT Code

*EXP.DETAILS: DTLP1004SN N-FERT. MEASURED WTHER

*TREATMENTS ------F ACTOR LEVELS------@N R O C TNAME...... CU FL SA IC MP MI MF MR MC MT ME MH SM 1 1 0 0 IRRIGATED FERTI 1 1 0 1 1 1 1 0 0 0 0 1 1 3 1 0 0 NON-IRRIG. 1 1 0 1 1 2 1 0 0 0 0 1 1

*CULTIVARS @C CR INGENO CNAME 1 RI IB0301 RD7, MANKEB

*FIELDS @L IDFIELD WSTA.... FLSA FLOB FLDT FLDD FLDS FLST SLTX SLDP ID SOIL 1 DTSPOO01 DTLP -99.0 0 DROOG 0 0 00000 -99 51 IBRI910024 @L ...... XCRD...... YCRD .... .ELEV ...... AREA .Sl EN .FLWR .SLAS 1 0.00000 0.00000 0.00 0.0 0 0.0 0.0

*INITIAL CONDITIONS @C PCR ICDAT ICRT ICND ICRN ICRE ICWD ICRES ICREN ICREP ICRIP ICRID 1 RI 61100 100 -99 1.00 1.00 1.0 0 0.80 0.00 100 15 @C ICBL SH20 SNH4 SN03 1 5 0.332 1.0 15.7 1 8 0.332 1.0 15.7 1 19 0.341 0.5 6.5 1 28 0.369 0.5 4.7 1 38 0.369 0.5 4.7 1 51 0.344 0.5 2.9

*PLANTING DETAILS @P PDATE EDATE PPOP PPOE PLME PLDS PLRS PLRD PLDP PLWT PAGE PENV PLPH SPRL 1 61247 -99 75.0 25.0 T H 20 0 5.0 0 25 25.0 3.0 0.0

*IRRIGATION AND WATER MANAGEMENT @I EFIR IDEP ITHR IEPT IOFF IAME IAMT 1 1.00 -99 -99 -99 -99 IR001 1 IDATE IROP IRVAL IIRV 1 61247 IR009 120 8 1 61247 IR006 45 0 @I EFIR IDEP ITHR IEPT IOFF IAME IAMT 2 1.00 -99 -99 -99 -99 IR001 1 @I IDATE IROP IRVAL IIRV 2 61247 IR009 120 0 2 61247 IR003 0 0

*FERTILIZERS (INORGANIC) @F FDATE FMCD FACD FDEP FAMN FAMP FAMK FAMC FAMO FOCD 1 61246 FE002 15 100 0 0 0 0 1 61273 FE002 15 60 0 0 0 0 1 61307 FE002 15 40 0 0 0 0

89 *HARVEST DETAILS @H HDATE HSTG HCOM HSIZE HPC HBPC 1 100 100.0 0.0

*SIMULATION CONTROLS @N GENERAL NYERS NREPS START SDATE RSEED SNAME ...... 1 GE 35 1 I 61100 2480 RICE MIT @N OPTIONS WATER NITRO SYMBI PHOS P POTAS DISES CHEM TILL 1OP Y Y Y N N N N N @N METHODS WTHER INCON LIGHT EVAPO INFIL PHOTO HYDRO 1ME M M E P R C R @N MANAGEMENT PLANT IRRIG FERTI RESID HARVS 1 MA R R R N D @N OUTPUTS FNAME OVVEW SUMRY FROPT GROUT CAOUT WAOUT NIOUT MIOUT DIOUT LONG CHOUT OPOUT 1OU Y Y A 1 Y N Y Y N N N N N

@N AUTOMATIC MANAGEMENT @N PLANTING PFRST PLAST PH2OL PH2OU PH2OD PSTMX PSTMN 1 PL 97159 97173 40 100 30 40 10 @N IRRIGATION IMDEP ITHRL ITHRU IROFF IMETH IRAMT IREFF 1 IR 30 50 100 IB001 IB001 10 0.75 NITROGEN NMDEP NMTHR NAMNT NCODE NAOFF 1 NI 30 50 25 IB001 IB001 @N RESIDUES RIPCN RTIME RIDEP 1 RE 100 1 20 @N HARVEST HFRST HLAST HPCNP HPCNR 1 HA 0 97365 100 0

90 Bibliography

[1] Apland, J., McCarl, B. A. and Miller, W. L. 1980. Risk and the demand for supplemental

irrigation: a case study in the Corn Belt. American Journal of Agricultural Economics, 62, 142-145.

[2] Arunmas, Phusadee. "New standards established for varieties of fragant rice." Bangkok Post, 19 July 2001.

[3] Bhuiyan, S. I. ed. 1994. On-farm reservoir systems for rainfed ricelands. International Rice Research Institute, Manila, Philippines.

[4] Boers, Th. M., and Ben-Asher, J. 1982. A review of rainwater harvesting. Agricultural Water Management, 5, 145-158.

[5] Bouman B. A. M. and Thong, T. P. 2001. Field water management to save water and

increase its productivity in irrigated lowland rice. Agricultural Water Management, 49, 11-30.

[6] Buapeng, Somkid. 1999. personal interview. Chief, Ground Water Technical and Techni- cal Section, Ground Water Division Department of Mineral Resources, Bangkok, Thailand.

[7] Burt, 0. R. and Stauber, M. S. 1971. Economic analysis of irrigation in subhumid climates. American Journal of Agricultural Economics, 53, 33-46.

[8] Center for Agricultural Information. 1998. 'Forecasting of Rice Production. In Journal of Forecasting Agricultural Products Crop Year 1997/1998 and 1998/1999. Office of Agri- cuitural Economics, Ministry of Agriculture and Cooperatives, Vol. 13, No. 2., June 1998. (ii Thai.) [9] Chaipattana Foundation. The New Theory. Retrieved November 15, 1999 from the World Wide Web: http://www.chaipat.or.th/ntheory/index e.html.

[10] Cohen, I. S.,Lopes, V. L., Slack, D. C. and Yanez, C. H. 1995. Assessing risk for water harvesting systems in arid environment. Journal Soil and Water Conservation, 50, 446- 449.

[11] Critchley, W. and Siegert, K. 1991. Water harvesting. FAO, Rome, Italy.

[12] Deaton, A. 1992. Understanding Consumption. Clarendon Press, New York.

[13] Deaton, A. 1997. The Analysis of Household Surveys : A Microeconometric Approach to Development Policy. John Hopkins University Press, Baltimore.

[14] Dent, J. B. and Blackie, M. J. 1979. Systems Simulations in Agriculture. Applied Science Publisher, London. 180p.

[15] Dudley, N. J., Howell, D. T., and Musgrave, W. F. 1971a. Optimal intraseasonal irrigation water allocation. Water Resources Research, 7, 770-788.

[16] Dudley, N. J., Howell, D. T., and Musgrave, W. F. 1971b. Irrigation planning, 2, choosing optimal acreages within season. Water Resources Research, 7, 1051-1063.

[17] Dudley, N. J., Howell, D. T., and Musgrave, W. F. 1972. Irrigation planning, 3, the best size of irrigation area for a reservoir. Water Resources Research, 8, 7-17.

[18] Featherstone, A. M. and Goodwin, B. K. 1993. Factors influencing a farmers decision to invest in long-term conservation improvements. Land Economics, 69, 67-81.

[19] Frasier, G. W., and Myers, L. E. 1983. Handbook of water harvesting. Agricultural Hand- book No.600, United States Department of Agriculture, Agricultural Research Service.

[20] Godwin, D. C. and Singh, U. 1989. Nitrogen dynamics in IBSNAT crop models. In Agronomy Abstracts. American Society of Agronomy, Madison, Wisconsin.

[21] Graf, B., Dingkuhn, M., Schnier F., Colonel, V., and Akita, S. 1991. A simulation- model for the dynamics of rice growth and development: III. Validation of the model with high-yielding varieties. Agricultural Systems, 36, 329-349.

92 [22] Hansen, J. W. and Jones, J. W. 2000. Short survey: scaling-up crop models for climate variability applications. Agricultural Systems, 65, 43-72.

[23] Hutasing, Onnucha. "Rice price increase too late for farmers." Bangkok Post, 8 May 2000.

[24] International Benchmark Sites Network for Agrotechnology Transfer (IBSNAT). 1990. Technical Report 5, Documentation for IBSNAT crop model input & output files, v. 1.1: for the Decision Support System for Agrotechonology Transfer (DSSAT v.2.1). Dept. of Agronomy and Soil Science, University of Hawaii, Honolulu.

[25] International Rice Research Institute. 1998. IRRI rice facts. IRRI, Los Bafios, Phillipines.

[26] International Rice Research Institute. 1999. "Rice gets Thailand through crisis," Rice: Hunger or Hope? IRRI Annual Report 1998-1999.

[27] Japan International Cooperation Agency (JICA). 1987. Integrated small-scale rural devel- opment by a ponds-system in Northeast Thailand in comparison with Toban Area in Japan: final report. The Joint Research Team (Chulalongkorn University, Kyoto University, and Royal Irrigation Department), March 1987, 515p.

[28] Jearakongman, S., Rajatasereekul, S., Naklang, Romyen, P., Fukai, S., Skulkhu, E., Jumpaket, B. and Nathabutr, K. 1995. Growth and grain yield of contrasting rice cultivars grown under different conditions of water availability. Field Crop Research, 44, 139-150.

[29] Jintrawet, A., Namuang, C., Vehara, G. and Tsuji, G. Y. 1990. Ex-ante screening of rice production strategies with the CERES-Rice model. In Papers presented at the second conference on "The Impact of Weather on Agricultural Production in the Pacific Rim Countries." University of Melbourne, Australia, 22-28 September 1990.

[30] Krishna, J. H., Arkin, G. F. and Martin, J. R. 1987. Runoff impoundment for supplemental irrigation in Texas. Water Resources Bulletin, 23, 1057-1061.

[31] Loucks, D. P., Stedginger, J. R. and Haith, D. A. 1981. Water resources systems planning and analysis. Prentice Hall, Englewood Cliffs, NJ.

93 [32] Mainudding, M., Das Gupta, A. and Onta, P. R. 1997. Optimal crop planning model for

an existing groundwater irrigation project in Thailand. Agricultural Water Management, 33, 43-62.

[33] Maneerungsee, Woranuj. "Weighing will be standardized." Bangkok Post, 7 December 2000.

[34] Mankeb, P. 1993. Calibration of genetic coefficients of paddy rice (Oryza sativa L.) for validation of the CERES-Rice model in Northern Thailand. M.S. Thesis. Chiangmai University, Chiangmai, Thailand.

[35] McLaughlin, D. B., Amornvivat, S. and Kiang, J. 1999. A Hydrologic Analysis of New Theory Concepts: On-farm Water Harvesting and Storage. Presented to His Majesty

the King Bhumibol Adulyadej on the occasion of the 72 n Royal Birthday Anniversary, 5 December 1999.

[36] Mongkolsmai, D. and Rosegrant, M. W. 1989. The effect of irrigation on seasonal rice prices, farm income, and labor demand in Thailand. In D. E. Sahn (ed), Seasonal Vari- ability in Third World Agriculture: The Consequences of Food Security, International Food Policy Research Institute, John Hopkins University Press, Baltimore, pp. 246-263.

[37] Naklang, K., Fukai, S. and Nathabut, K. 1996. Growth of rice cultivars by direct seedling and transplanting under upland and lowland conditions. Field Crop Research, 48, 115-123.

[38] Oswal, M. C. 1994. Water conservation and dryland crop production in arid and semi-arid regions. Annals of Arid Zone, 33, 95-104.

[39] Oweis, T. and Taimeh, A. T. 1996. Evaluation of a small basin water-harvesting system in the arid region of Jordan. Water Resources Management, 10, 21-34.

[40] Oweis, T., Hachum, A. and Kijne, J. 1999. Water harvesting and supplementary irrigation for improved water use efficiency in dry areas. SWIM Paper 7. Columbo, Sri Lanka: International Water Management Insititute.

94 [41] Palmer, W. L., Barfield, B. J. and Haan, C. T. 1982. Sizing farm reservoirs for supple- mental irrigation for corn, I, Modeling reservoir size-yield relationships. Transaction of the ASAE, 25, 372-376.

[42] Passioura, J. B. 1996. Simulation models: science, snake oil, education, or engineering? Agronomy Journal, 88, 690-694.

[43] Paulson, A. L. and Townsend, R. 2001. The nature of financial constraints: distinguishing the micro underpinnings of macro models. Manuscript. University of Chicago.

[44] Pender, J. L. 1992. Credit rationing and farmers' irrigation investments in South India: theory and evidence. Food Research Institute, Stanford University, Standford, August 1992. Ph.D. thesis.

[45] Pinnschmidt, H. 0., de la Pena, F., Suriya A. D., Don, L. D. and Teng, P.S. 1996. Approaches for the characterization of cropping situations and yield constraints in rainfed lowland rice agroecosystems of some Southeast Asian countries and consequences for crop protection. Journal Of Plant Diseases And Protection, 103, 620-643.

[46] Poomklom, M. 1999 personal interview. Chief, Phra Phutabart Field Crop Experiment Station, Department of Agricultural Extension, Lopburi, Thailand.

[47] Ritchie, J. T., Alocilja, E. C., Singh, U. and Uehara, G. 1986. IBSNAT/CERES Rice Model. Agrotechnology Transfer, 3, 1-5.

[48] Rodriguez, A. 1997. Sustainable agricultural groundwater management in Syria: Utopia or reality? In: Water: Economics, management, and demand, eds. Melvin Kay and the ICID. E. and F. N. Spon, London.

[49] Salam, M. U., Jones, J. W., and Kobayashi, K. 2001. Predicting nursery growth and transplanting shock in rice. Experimental Agriculture, 37, 65-81.

[50] Siamwalla, Ammar. 1991. Changing Comparative Advantage in Thai Agriculture. OECD Technical Paper No. 35, March 1991. Retrieved September 7, 2000 from the World Wide Web: http://www.oecd.org/dev/PUBLICATION/tp/tp35.pdf

95 [51] Siegert, K. 1994. Introduction to water harvesting: some basic principles for planning, design and monitoring. In. Water harvesting for improved agricultural production. Pre- ceedings of the FAO Expert Consultation, Cairo, Egypt, November 1993, Rome, Italy: FAO.

[52] Somrith, B. 1995. Thai rice industry: progress through research. A paper presented at

the International Conference on Asia Grain Markets'95, 28-29 September 1995, Bangkok, Thailand.

[53] Somrith, B. 1996. Curtivar improvement for rainfed lowland rice in Thailand. In S. Fukai, M. Cooper and J. Salisbury (eds), Breeding Strategies for Rainfed Lowland Rice in Drought-prone Environments, Preceedings of an International Workshop, 5-8 Novem- ber 1996, Ubon Ratchathani, Thailand, Australian Centre for International Agricultural Research (ACIAR) Proceedings No. 77, pp. 36-42.

[54] Somrith, B. 1998. Rice production in Thailand. A country report presented in the 19th Session of the International Rice Commission, 7-9 September 1998, Cairo, Egypt.

[55] Suleman, S., M. K. Wood, B. H. Shah, and L. Murray. 1995. Rainwater harvesting for

increasing livestock forage on arid rangelands of Pakistan. Journal Range Management, 48, 523-527.

[56] Timsina, J., Singh, U., Badaruddin, M. and Meisner, C. 1998. Cultivar, nitrogen, and moisture effects on a rice-wheat sequence: experimentation and simulation. Agronomy Journal,90, 119-130.

[57] Tsuji, G.Y., Uehara, G., Balas, S., 1994. Decision Support System for Agrotechnology

Transfer, Version 3. International Benchmark Sites Network for Agrotechnology Transfer, University of Hawaii, Honolulu, Hawaii.

[58] United Nations Environment Program (UNEP). 1983. Rain and storm water harvesting in rural areas. Tycooly International Publishing Limited, Dublin, Ireland.

[59] Wiens, T. B. 1976. Peasant risk aversion and allocative behavior: a quadratic programming experiment. American Journal of Agricultural Economics, 58, 629-635.

96 [60] Willmoltt, C. J. 1982. Some comments on the evaluation of model performance. The Bulletin of the American Meteorological Society, 63, 1309-1313.

[61] Wopereis, M. C. S., Kropff, M. J., Maligaya, A. R., and Tuong, T. P. 1996. Drought-stress reponses of two lowland rice cultivars to soil water status. Field Crops Research, 46, 21-39.

[62] Ziari H. A., Mccarl B. A. and Stockle, C. 1995. A nonlinear mixed-integer program model for evaluating runoff impoundments for supplemental irrigation. Water Resources Research, 31, 1585-1594.

97