Factors affecting agricultural land use for vegetables production-a case study of the state of Selangor , Malaysia

Economic transformation and growth had resulted into increased population and change in taste of the people of Selangor, Malaysia. Evidence shows that the people have graduated from the consumption of highly starchy food to the consumption of more proteinous food, fruits and vegetables. This changing taste will hitherto produce effects on the sizes of agricultural land use for production of different agricultural crops. A case of drivers of agricultural land use for vegetables production (ALUVP) is assessed in this study. ALUVP was compared with the potential driving variables at three different scales using Spatial Analyst 3.2 in an arcGIS 9.2 environments. Findings indicated that vegetables are found cultivated in mixed cropping systems with coconut, orchard, paddy, rubber, idle grass, whereas there were observed competition between vegetables production and oil palm, swamp/forest for land use. Factors such as maximum temperature, average temperature, slope, population density and soil series have inverse relations with ALUVP while distance to lake, major river, minor roads, road density, number of raining days, percentage urban residents have direct relationships with it (ALUVP). Drivers of ALUVP differ at different scales of analysis with the effects of accessibility becoming more pronounced at lower scales of analysis than at higher scale. With reference to competition for land use within and between agricultural sector and other more profitable non agricultural sectors, it is unlikely that the government policy initiative of achieving self sufficiency in vegetables production by the year 2010 will be achieved.


INTRODUCTION
The state of Selangor covers an area of about 8000 square kilometers and is located between latitudes 2.58 and 3.83°N and longitudes 101.17 and 102.00°E (Figures 1a and b).The climate of Selangor is governed by two *Corresponding author.E-mail: amakmom@env.upm.edu.my.Tel: +60102933451.monsoonal wind which originates from the north east between October to February and the south west from May to September.Climatically, the state has no distinct dry season.Average daily temperature ranges from 21 to 32°C and mean annual temperature is 26°C and the daily humidity level exceeds 80%.At the North -Eastern part of the state is a range of granite intrusion, having narrow valleys and steep crest and on the fringe of this range is a metamorphic zone with hilly to steep terrain and also a   sedimentary zone with undulating topography and plain quaternary marine alluvia towards the state's boundary with a state called Melacca.The state of Selangor is one of the most developed andin fact the richest state in Malaysia (Alias et al., 2010).For instance, the Malaysian Department of Statistics in 2010 found that the state single handedly contributed 22% to the national GDP.Selangor economy is primarily driven by the manufacturing and the services sector as these two sectors only contributed 33.2 and 61.3% to the state economy in 2009.UNDP asserted that the state has an adult literacy level of 95.1%.
This thus underscores the importance of the state in labour supply.The state houses the country's largest port (Port Klang) and the nation's largest industrial catalyst of economic growth especially the Kuala Lumpur International Airport and Subang Airport are found in the state (Alias et al., 2010).
Agricultural sector has contributed significantly to the economic development of the country as it used to be the major source of income, employment and foreign exchange earner (Alam et al., 2010).
However, from 1970 till date the contribution of agriculture to the economic development of the country dropped significantly from 22.9 to 18.7% and later to 13.6% in 1995.
Also the contributions of agriculture to employment fall from 39.7% in 1980 to 27.8% in 1990.
Comparatively, agricultural contribution to the GDP decreased from 8.8% in 2000 to 8.2% in 2005.Currently, agriculture is the third engine of growth next to manufacturing and service sectors (Alam et al., 2010).
However, agriculture still remains important in the production of food, vegetables and fruits for the teeming population (Table 1 and Figure 2).Currently, Malaysia farmers produce fruits and vegetable crops such as, banana, coconuts, durian, pineapple, rambutan and others (EoN, 2010).

Major vegetables and growing areas
Malaysia characteristics tropical climate is favorable for Olaniyi et al. 5941 the production of fruits and vegetables EoN (2010).
Vegetables productions have been found to be smallholder crops in Malaysia, with an average farm size of between 0.5 and 1.0 ha under agro forestry/mixed cropping system and low input technology.Important vegetable producing states in Peninsula Malaysia are Johor, Perak, Kelantan, and Pahang, and recently Selangor accounting for about 75% of the total vegetable production in Malaysia (Table 2).

Production
At the Peninsular Malaysia level, the agricultural areas gazette for vegetables had been increasing (Table 3) from 2000 till date and subsequently leading to production increase from 512,000tons to 717,000tons within that period worth of about MYR574 million or US$220 million.Fruit vegetables recorded the highest increase in both cultivated area and production with average yield between 15 and 18 t/ha (Table 4).
In Peninsula Malaysia, the annual per capita consumption of vegetables is 63.6 kg against the Food and Agriculture Organization (FAO, 1998) recommended 73 kg per capita per year (Table 5 and Figure 3).The growing demand for vegetables has been due to the rising population, higher per capita income, migration to urban areas, a greater awareness of health issues and an increase in tourism.In this context, this study concludes that demand for these products will continue to grow in the next few years (Table 6).
To take care of the shortfall between demand and supply, the country resulted to importation of vegetables.In 2005, the value of fresh vegetable imports reached EUR 324.37 million, 10.5% more than the previous year with major import partner being China, Australia, USA and Thailand.Despite the fact that Malaysia's vegetable industry has doubled its production over 10 years, its capacity is yet unable to meet the demand (Chong, 2007).However, availability of market has encouraged farmers in other states to embark on vegetable production.
The objective of this study is to identify in quantitative terms the factors affecting agricultural land use for vegetable production (ALUVP) with aims of finding whether land use constitutes potential impediment to the achievement of sustainable production of vegetables in the study area using logistic regression analysis.

Logistic models in land use and land use change analysis
The relationship between land use and its driving factors is a complex, dynamic and nonlinear process that is usually accomplished by comparing land use and the       location specific characteristics at a polygon or pixel level (Karimi et al., 2010).These relationships have been successfully measured with the use of logistic models (Lesschen et al., 2005;Zhu et al., 2005;Zhang et al., 2010).This logistic regression (Equation 1) is used to predict the probability of the presence or absence of a particular (in our case ALUVP) land use type in relation with the driving factors of the land use.
The coefficients (β's) of the logit model are the values for the respective driving factor for a particular land use type.
Values lower than 1 mean that the probability will decrease upon an increase in the value of the independent variables whereas, the values higher than 1 indicates an increase in the probability upon an increase in the value of the independent variables (Menard, 2001).While R 2 is usually used to describe the goodness of fit in ordinal least squares regression, the measure of goodness of fit in logistic regression can only be evaluated with the ROC method (Pontius and Schneider, 2001).Relative Operation Curve (ROC) is a measure of how well the model/independent variables correctly predict the value of the dependent variable.A ROC value higher than 0 .7 is considered acceptable in land use and land cover change modelling (Pontius and Schneider, 2001) while a value beyond 0.8 is considered excellent and value above 0.9 is considered outstanding (Homer and Lemeshow 2000).

Data
The 2006 state land use vector map used in this study was collected from the Malaysian Department of Agriculture (MDoA).Relevant biophysical data (soil, elevation, slope, minimum, maximum and average temperature) were obtained from the relevant government agencies while the socio-economic data were collected from the Malaysian Department of Statistics (MDoS, 2006).Elevation was included as a proxy for drainage that increases the cost of land development (Monmonier, 1982).Slope data as obtained from Malaysian Department of Survey was used as a proxy for suitability of a piece of land for agricultural purposes given the Malaysian edict that prohibits the use of land of slope greater than 15° for agriculture (MDoA, 1999).Several potential spatially explicit independent variables hypothesized to affect the dependent variables were also developed within a GIS.These include Euclidean distances to minor roads, lakes, water bodies, and major roads, population density, elevation, slope,aspect, and curvature.All variables were mapped at a resolution of 150 m; 200 m and 250 m to assess the effects of scales on factors affecting agricultural land use for vegetables production within arcGIS 9.2 software.

Derivation of GDP at district level
Some socio economic data (population and housing density) obtained from the Malaysian Department of Statistics (MDoS, 2006) were available at the district level whereas others such as (GDP were supplied at the state level.Therefore, to derive GDP at district level, some assumptions were made in order to simplify our study.For instance, the total age group between 15 to 64 per location (district) was assumed as the workforce of a particular location and the subsequent GDP were calculated based on the available workforce and the status (rural or urban) of that location.Therefore, manufacturing and services sectoral GDP were assumed to be contributed by labour in an urban areas and primary GDP was assumed to be contributed by workforce in the rural areas implying that mobility of labour were assumed to be restricted.

Derivation of population data
Population data for this study area were obtained from the Malaysian Department of Statistics.Data for the missing periods were derived using the method documented by (Braimoh and Onishi, 2007).This approach was used to derive population data of rural and urban residents of this study area in the year 2006.

Derivation of agro climatic variables
The main agro-climatic variables used in this study Table 7 are precipitation (number of raining day per annum) and temperature (minimum, average and maximum) (in °C) and these data were obtained from the Malaysian Meteorological Department and were converted to vector format using Spatial Analyst 3.2 in an arcGIS 9.2 environment.Per mapping unit pH and CEC were derived from global data set of derived soil properties (Batjes, 2000).pH and CEC is a measure of the soil acidity or alkalinity of a soil otherwise called the soil "water" pH is a measure of the soil pH in soil solution, which is considered the active pH that affects plant growth.CEC is a value that is an estimate of the soils ability to attract, retain, and exchange cation elements.It is expressed in millequivalents per 100 g of soil (meq/100 g).Soil pH and CEC were selected because they (soil pH and CEC) have been considered the most functional soil property that is important in defining agricultural possibilities and constrains of a soil (Staal et al., 2002).Therefore, access to market and agricultural inputs were taken into account by the distance to transportation facilities (minor and major road, rail, and river) (Staal et al., 2002).Standard of living is determined by GDP per capita and this is used as a proxy for availability of market for vegetables.

METHODOLOGY
The methodology employed in this study involve the comparison of the land use type Agricultural land use for vegetables production (ALUVP) with the potential explanatory variables on a cell by cell basis (Karimi et al., 2010).Raster data of ALUVP and all the potential drivers were prepared using Spatial Analyst 3.2 in an arcGIS 9.2 environment (Figure 4).Land use map of this study area for year 2006 was obtained from the Malaysian Department of Agriculture.This map was reclassified to single out the area used for vegetable production.Data of slope, elevation and soil map Vector map of the potential socio-economic drivers such as population, population density, rural and urban work force, percentage rural and urban residents, primary, secondary and tertiary GDP maps were generated from the raw data obtained from the Malaysian Department of Statistics using arcGIS 9.2 and were thereafter rasterised.
Raw climatic data for the year 2006 were equally mapped into vector data and were subsequently rasterized.Rasterization of all vector data were done at the same spatial extent, geographical coordinates and three different scales in order to account for the effects of scales on drivers of agricultural land use for vegetables production ALUVP.The resulting raster data were all converted into ASCII files.A computer program was written to convert matrixbased m x n ASCII files into p x 1 vector data where p = m × n.The vector data were then imported into SPSS version 18 were logistic regression analysis was performed to identify and quantify the drivers of agricultural land use for vegetables production (ALUVP).

RESULTS AND DISCUSSION
The analysis were carried out at three scale such as 150; 200 and 250 m 2 so as to investigate the effects of spatial scale of analysis on drivers of agricultural land use for vegetable production (ALUVP).
These scales were chosen because it is related to the farm size maintained by every group of farmers in vegetables production (2.25; 4.0 and 6.25 ha).Since, vegetables are cultivated in mixed cropping systems with some crops such as paddy, coconut.The above spatial scale is considered as appropriate to represent farm size maintained.
Results in Table 8 clearly show the effect of three categories of driving factors on ALUVP viz: availability of suitable agricultural lands, biophysical factors and socioeconomic factors.

Availability of suitable agricultural lands
ALUVP is observed to be competing for land use with other agricultural land uses such as oil palm cultivation, idle grass and swamp forest land uses (Olaniyi et al., 2011), whereas a level of mixed cropping is observed between ALUVP and other crops such as orchard, rubber, paddy and coconut this observation is in agreement with the report of (EoN, 2010).The practise of mixed cropping by farmers has both advantages of income security and higher productivity.However, where it is not properly managed, it could results to environmental degradation of the land resources and pollution of the aquatic environment as a result of excessive application of fertilizer on paddy fields (Chong, 2007).
It is noteworthy to observe from the result in Table 8 that at lower scale swamp forest showed a positive relationships with ALUVP, this is an indication that some vegetable are produced under agro forestry system while at a larger scale, ALUVP has negative relationships with swamp forest indicating total replacements of swamp forest by ALUVP and this has a serious environmental impact.

Biophysical variables
These include the effects of climatic and soil factors on ALUVP.From Table 8, it is observed that the following biophysical factor have negative effects on ALUVP, that is, maximum and average temperatures, slope and soil series whereas, elevation (a proxy for drainage), number of raining days and relative humidity have positive effects

Socio-economic variables
Factors such as dependency ratio (a proxy for availability of rural work force) distance to major river (a proxy for drainage); distance to minor road; minor road density (a proxy for access to farm inputs and information as well as market) and percentage urban population (a proxy for market access) have positive relationships with ALUVP while build up areas (a proxy for reduction in available land for vegetables cultivation), distance to lake (a proxy for access to irrigation water) and population density (a

Figure 1b .
Figure 1b.Showing map of Peninsula Malaysia and the study area.

Figure 2 .
Figure 2. Change in agricultural land use for different crops.

Figure 3 .
Figure 3. Showing per capita consumption of food commodities.

Table 1 .
Percentage change in Agricultural land use in Malaysia from 1985 to 2010.

Table 3 .
Showing the vegetable farm per states in Peninsula Malaysia.

Table 5 .
Per capita consumption of food commodities (kg/year).
Source: Department of Agriculture, Malaysia, 1999 estimates

Table 7 .
Lists of potential drivers of ALUVP used in this study.