Preparation of Saffron Cultivation Suitability Map Based on the Comparison of Different Weighting Methods in GIS Environment

Manuchehr Farajzadeh
Reza Mirza Bayati
Mohammad Rahimi
Tarbiat Modares
University, Department of remote sensing and GIS, Tehran, Iran P.O.Box 14155-4838
Reza Amiri
Pyam Nour Universuty
Hashtrod, Iran
[email protected]
[email protected]
[email protected]
[email protected]

Abstract
Saffron is the most expensive agricultural and medical crops in the world which have an important role in the economic development regions. Site selection for the cultivation of various agricultural crops in regard to environmental conditions of different geographic regions is an important aspect of agricultural studies and abilities of GIS systems. Different weighting methods in multi-criteria analysis were used in GIS to determine the best environment for various agricultural crops. The main object of this paper is to compare the four weighting methods including the boolean, ranking, rating and pair-wise comparison and introduce the best model in GIS environment. To do that, in the first step determined the necessary maps according to required environmental conditions including elevation, slope, land use, soil depth, availability of surface water, groundwater and temperature threshold maps were prepared in the GIS. Provided data layers were analysed using GIS multi-criteria evaluation analysis performed by boolean, ranking, rating and pairwise comparison methods and the related maps were produced for each of the applied methods in GIS and the results were compared. The result of this study indicates that the pairwise comparison method shows a better result than the other and presents GIS potential ability to select suitable lands for saffron cultivation. Key words: saffron cultivation, multicriteria analysis, climate, GIS, Nishabour plain.

1. Introduction
The problem of selecting the correct land for the cultivation of a certain agriculture crops is a long-standing and mainly empirical issue (Kalogirou, 2002). Therefore many researchers have tried to prepare a standard framework for suitable and optimum agriculture land use. FAO presented a procedure to evaluation of lands potential in the cultivation of crops based on soil and environmental characteristics into five classes including highly suitable, moderately suitable, marginally suitable, currently not suitable and permanently not suitable (FAO, 1976, 1984, 1985). Considering the ability of GIS in the integration and analysis of spatial and non-spatial data, it is a very useful tool for the selection of suitable areas in the cultivation of specific crops into to different environmental conditions. So, recognition and identification of the sites which include required parameters such as topography, soil, climate, water resources and other parameters can be attained using the capabilities of GIS. The recognized zones by this approach are called Agro-climatic Zones (ACZ).

Malczewski classified GIS-based land use suitability analysis into 3 main groups: 1)computer-assisted overlay mapping, 2)multicriteria decision making methods including multi-objective methods and multi-attribute methods and 3)artificial intelligence methods including fuzzy logic techniques, neural networks, evolutionary (genetic) algorithms and cellular automata (Malczewski, 2004). Using multi-criteria methods, Ceballos-Silva and Lopez-Blanco delineated suitable areas for the production of maize and potato crops in central Mexico (Ceballos- Silva and Lopez-Blanco, 2003). They used 3 main climate, soil and elevation data and have weighted these data using AHP procedure. The use of fuzzy method also has absorbed attentions of some researchers, for example Nisar et al. (2000) presented a GIS-based fuzzy membership model for crop-land suitability analysis. Liu and Samal (2002) used a fuzzy clustering approach to delineate agroecozones. Malczewski (2006) presented ordered weighted averaging with fuzzy quantifiers. Additionally, Zewen et al. (2003) using remote sensing data and GIS techniques made a possibility study for flax cultivation in some regions of Henan province of China. Caldiz et al. (2001) made an ecological agricultural classification and studied capabilities of agricultural lands in cultivation of potato by using GIS in Argentina.

In Iran, Farajzadeh and Bighash (2000) determined Agroclimatological regions for increasing wheat cultivation in Hamadan province, Iran. Yazdzdanpanah et al. (2003) determined agroclimatological regions of East Azerbaijan province of Iran from the point of view of almond requirements by using GIS and data of 10 synoptic stations.

Literature review indicates that the study of crop-land suitability evaluation for saffron cultivation is not considered until now. Saffron is a plant which its active period and effective growth occurs in fall and winter. The most important climatic and environmental requirements and limitations are as follow:

1. The maximum resistance to the cold air is -18 to -20 °C,
2. Frosts at the time of flowering, destroy the flowers and decreases the yield strongly,
3. The minimum temperature threshold of 10 °C and less at night and maximum threshold of 22 °C in daytime are the required conditions for flowering,
4. The soil with moderate, somehow soft and good infiltration is suitable for saffron. So, in clay sand limy soils this plant has a fine growth,
5. This plant needs 416 degree-days from the time of sowing to flowering phase.
6. The main framework of this paper is to analyse the collected data based on different weighting methods. The main object of this study is the determining of the best methods of weighting methods in GIS environment to select of the best suitable areas to development of saffron cultivation.

2. Material and methods
Nishabour plain is located in Khorasan province of Iran. This plain is a part of Kalshoor Nishabour basin that is located in the side of Binalood Mountains in Northeast of central desert in Iran (figure 1). This plain is located between latitude 35? 40? N and 36? 39? N and longitude 58? 17?? and 59? 30? E. The total area of the basin is 7300 square km which an area of 4100 square km is plain and the rest is mountainous area. According to statistics of Agricultural ministry of Iran more than 60% of saffron is being produced in Iran and 95% of this amount belongs to Khorasan province (Jihad Keshavarzy Ministry, 2000). Average annual precipitation of the region within the period of 1991-2002 was 256 mm. Annual average temperature in Nishabour station is 14.2 °C.

In this study, different groups of data layers including topography, soil resources, water resources, climatological setting and land use were collected and analyzed. All the base and thematic maps was in the scale of 1:250000 which was assembled from different organizations in the country which their characteristics are explained as follow:

• Topographic data: In this study, two maps, namely elevation and slope gradient maps, were prepared and analyzed. These maps were prepared based on general topographic maps in the scale of 1:250000 produced by National Geographic Organization of Iran. The produced elevation map indicates that prevailing altitude in the area is between 1330 to 2300 meters.
• Climatic data: To carry out possibility study of cultivation and development of saffron in the region regarding to its phonological phases, daily meteorological data of Nishabour, Gaen, Gonabad, Torbat Heydarieh, Kashmar, and Birjand weather stations within the period of 1991-2002 belonging to Iranian Meteorological Organization (IMO) were collected. Based on the statistics of weather stations, 5 different factors including precipitation, probability of fall frost, probability of 5°C temperature and lower, probability of occurrence of 10°C and 22°C temperature and degree-day maps as the frameworks below were extracted. The preparation of precipitation map was carried out based on the calculation of linear regression between precipitation data and elevation of the study area.


Figure 1. Geographical location of Nishabour Plain

Information about fall frosts is an important issue, because fall frosts occurs at the time of saffron flowering and has harmful affects on the yield. On the other hand, each field crop needs a special minimum temperature to initiate germination. Moreover plant growth decreases or stops on or less than this point. This temperature is called “biological Zero” or “base temperature” and for saffron it is equal 5°C. One of the suitable conditions for saffron flowering is that the temperature should be more than 10°C at night and less than 22°C during day at the flowering stage. In order to calculate the probability of the above mentioned temperature thresholds occurrences, the probability of occurrences in effective months on saffron cultivation were calculated by using daily data of weather stations. Furthermore probability percent was determined by the interpolation of occurrences probability for those stations.

Among climatologically parameters, thermal regime has the greatest effect on plant growth and development. So, for each phonological phase, crop requires given amount of heat as thermal unit or degree-days. For saffron this amount is 416 degree days from sowing time till flowering stage. Since the sowing time of saffron is in September, therefore the degree-days of synoptic weather stations data were calculated and the map of degree-day’s distribution was created by interpolation function of GIS.

• Hydrologic data: Based on their effects on saffron cultivation, 3 main hydrologic factors including water quality, accessibility to surface waters and accessibility to groundwater were prepared. The water quality map of the region was classified to 5 categories according to the importance by considering high-quality water distribution and also different characteristics of plain. As the supplier of a part of required water for irrigation and soil conservation, surface water has an important role in crop growth. In order to provide the map of accessibility to river network, GIS buffering function was used to prepare the map. In a similar method, the accessibility to groundwater resources including wells, springs and Qantas were prepared.
• Land use and soil resources data: Land use map of the study area was prepared by the classification of ETM+ 2002 satellite images using supervised method. In this map, main land uses classes including agricultural lands, water resources, residential areas, rangelands and forests were determined. Also the land form maps and soil depth were acquired from Soil and Water Research Institute of Iran.

To consider required conditions for saffron cultivation, different weights were assigned to 11 provided data layers using four different methods including boolean, ranking, rating and pairwise comparison methods. Table 1 indicates the considered data layers and assigned weights according to applied weighting methods. The main framework to assign a weight was expert opinion.

In the boolean method, the minimum condition of saffron cultivation that mentioned in the introduction of this paper was considered and a given number between 0 and 1 was assigned to each of the effective data layers in saffron cultivation. Number 1 indicates existence of the required condition for cultivation and number 0 indicates absence of the required condition. Therefore in this method different classes were categorized into only two groups of 0s and 1s.

Based on the mentioned explanation, all the related conditions considered as following: altitude 2300 m, slope gradient 15%, landforms of slope plains and traces, deep and semi deep soil depth, water quality from medium to very good, accessibility to groundwater with more than 25 litre per second, precipitation with more than 20 mm in sowing periods, the probability of autumn frost occurrence less than 15%, the probability of occurrence of 5°C and lower with less than 16%, the probability of occurrence of 10°C to 22°C with more than 27% and degree-day with more than 475.

In the ranking method, considered parameters were prioritized according to their importance for saffron cultivation and were ranked. Based on experts suggestions, the probability of fall frost factor and precipitation factor, due to irrigation system required for saffron cultivation, are the most and least important factors respectively (table 1). There are various methods for weighting after ranking such as rank sum, reverse ranking, and exponentional ranking, which rank sum equation used in this research as formula 1 (Malczewski, 1999).

in which w; is the normalized weight for j th criterion, n is the number of criteria under consideration (k = 1, 2,…, n) and rj is the rank position of the criterion. Each criterion is weighted (n-rj+1) and then normalized by the sum of all weights. The result of this calculation and assigned weights are presented in table 1.

In rating method, all data layers were weighted with a percent scale from 0 to 100. The number 0 was assigned to the parameter that has the lowest importance and number 100 belongs to the factor which has the highest importance. In improved rating method, parameters were graded based on their importance and number 100 was assigned to the parameters with the highest importance and number 0 to the parameter with the least importance (table 1). The primary weight was calculated by dividing assigned number of each parameter to the number of the parameter which has the least importance and then primary weights were standardized.

Pairwise comparison method (PCM) was developed by Saaty in 1980 in the context of analytic hierarchy process (AHP). This method involves pairwise comparisons to create a ratio matrix. It takes as an input the pairwise comparison and produces the relative weights as output (Malczewski, 1999). This method was introduced for the first time that moreover applied to GIS application by Matt in 1991(Thirumalaivasan et al. 2003). The concept of weighting in this method is based on the comparison of the parameters item by item. In this method the importance of parameters in a consistent comparison divides to 9 categories. In order to weighting by this method, factors were rated according to the opinion of crop experts as other mentioned methods. In the weight calculation step, the pairwise comparison matrix and factor maps were applied using weight module in the IDRISI environment. Then the principle eigenvector of the pairwise comparison matrix was computed to produce a best fit to the weight set. In a MCE using a linear weight combination, it is necessary that the weights sum to 1. The consistency ratio of the matrix was calculated as well. This value indicates the probability that rating was randomly assigned. A consistency ratio of 0.01 or less is considered acceptable (Malczewski, 1999). The computed weights was shown in table 1. This procedure indicates that the most important factors for saffron are: land use types (0.65), water quality (0.625) and slop gradient (0.6).

After weighting the affective layers in saffron cultivation, the distribution map of the suitable sites for saffron was created based on each weighting methods.

Using GIS capabilities, the combination and overlapping of assigned weights was carried out. Finally, in order to show the suitable sites and also to compare the maps resulted from various methods more accurately, the normalisation of maps was done at the range of 0 to 1. Then resulted maps, were classified to 4 categories of importance (suitable, medium, low, and not suitable) from the viewpoint of saffron cultivation possibility.

3. Result and discussion
Figures 2, 3, 4 and 5 show the results of applying 4 methods of Boolean, ranking, rating and Pairwise Comparison Method respectively. The distribution of suitable sites for saffron cultivation, developed by various methods, was shown in Table 2.

Table 2. Distribution of suitable places for saffron cultivation based on different weighting methods Table 2. Distribution of suitable places for saffron cultivation based on different weighting methods Table 2. Distribution of suitable places for saffron cultivation based on different weighting methods

Figure 2, based on Boolean method, shows that suitable areas for saffron cultivation are very limited and mainly concentrate in the southern part of the study area depicting a scattered pattern. These areas are regions that have the minimum threshold for cultivation of saffron without any priorities. Therefore in this method conditions are very definite and only 4.27 percent of the area under investigation is suitable for saffron cultivation. The results of applying the ranking method presented in figure 4 illustrates that 21.4 percent of the study area has a good possibility for saffron cultivation which are mainly located in southeast and western parts. But other classes with medium and low possibility exist in the area with a percentage of 32.1 and 20.02 respectively. Also the area without possibility in this method is less than Boolean method. Additionally, this method can show the degree of possibility of saffron cultivation unlike Boolean method. In rating method, as figure 4 shows, about 13.56 percent of area was suitable for saffron cultivation which were located in the south and southeast and also some minor patches in the northwest. In this method similar to ranking method, different classes are observable. The most obvious distinction between the ranking and rating method is the difference in the extent of not-suitable class. While the extent in the ranking
Figure 2. Possibility study of saffron cultivation (Boolean method)

Figure 3. Possibility study of saffron cultivation (ranking method)

Figure 4. Possibility study of saffron cultivation (rating method)

Figure 5. Possibility study of saffron cultivation (pairwise comparison method)

method is about 26 percent, in the rating method it decreases to only 5 percent. In other word unlike the ranking method, the not-suitable class converts to low possibility. This class could be used as the area for saffron cultivation in certain circumstances.

And finally, the figure 5 which is prepared based on Pairwise comparison method indicates that about 6.02 percent of the study area is suitable for saffron cultivation that is located in southeast, and some minor patches in south, west and northwest of the basin. The extent of the area of good suitable class of this method is very similar to Boolean method, but in the former we can see the possibility potential in other categories. In this method the highest extent with about 55 percent belongs to the moderate class which is similar to the rating method result.

The calculation of correlation coefficient between resulted extents in table 3 shows that these coefficients are 0.7938 between ranking and rating methods, 0.6199 between ranking and Pairwise comparison method and 0.8753 between rating and pairwise comparison methods. All of these values are statistically significant in 0.95 levels. According to this calculation, it is clear that the results of rating and Pairwise comparison methods are very similar and both of them could be used to classify the suitability of areas for saffron cultivation.

If we consider the areas identified by the boolean method as a the best areas for cultivation of saffron, the pairwise comparison weighting method presents an acceptable result because the extent of the suitable area in both of them is very close (4.27 in boolean model unlike 6.02 in pairwise comparison method). Also, the suitable areas in two maps resulted from the above methods match spatially to a large extent.

Table 3 indicates cross-tabulation of different degrees of saffron cultivation based on the pairwise comparison method with existing land use types extracted from satellite data. This table shows that about 40 percent of the area with low dense rangelands land use has good suitability for saffron cultivation. In other word, these land uses can be used for saffron cultivation. Also about 33 and 25 percent of low dense rangelands have medium and weak suitability respectively which can be converted to saffron cultivation with some amendments. These changes in land use types will lead to extended and increased value of lands and finally to economical development of the area. Other land use classes in the area can conserve existing conditions and required changes take places only in specific conditions.

Table 3. Cross tabulation of different degree of saffron cultivation based on pairwise comparison method with exciting land use types extracted from satellite data

4. Conclusion
The review of mentioned results indicates that in different methods, there is only a little difference between the results and all have the same scheme except from the Boolean method. Since in the Boolean method limited choices and range of thresholds are limited, there is not a suitable flexibility required in the process of site selection. The Ranking method can not be considered as an accurate method to prepare final map because the weight of each parameter is being defined only by considering the number of grades determined by its importance. The rating method, due to its simplicity in weighting, can be used with confidence. The pairwise comparison method which is carried out by comparing parameters two by two, in spite of high accuracy, because of weaknesses such as high number of comparisons in matrix calculation can be used only in computerized methods.

Therefore it seems that there is not an accurate and clear method for the weighting of the parameters in all situations. So, it is better to calculate all of the methods for a given area and then select one of them.

References:

• Caldiz Daniel O., Gsapari Fernanda J., Haverkort Anton J. , Struik Paul C.,2001. Agro-ecological zoning and potential yield of single or double cropping of potato in Argentina, Journal of agricultural and forest meteorology, ELSEVIER, No. 109, 311-320.
• Ceballos- Silva Alejandro, Lopez-Blanco Jorge, 2003. Delineation of suitable areas for crops using a Multi-Criteria Evaluation approach and land use/cover mapping: a case study in Central Mexico, Agriculture system, 77, 117-136.
• Food and Agriculture Organization of the United Nations (FAO), 1976. A framework for land evaluation, Soils bulletin, 32, Rome: FAO.
• Food and Agriculture Organization of the United Nations (FAO), 1984. Guidelines for land evaluation for rainfed agriculture, Soils bulletin, 52, Rome: FAO.
• Food and Agriculture Organization of the United Nations (FAO), 1985. Guidelines for land evaluation for irrigated agriculture, Soils bulletin, 55, Rome: FAO.
• Farajzadeh M., Byghash, A., 2000. Agro climatic classification of Hamadan province of Iran using GIS with emphasis to wheat cultivation, Journal geographic research, No 41, P 93- 105. (In Persian) Jihad Keshavarzy Ministry, 2000. Agricultural statistics (In Persian)
• Kalogirou S., 2002. Experts systems and GIS: an application of land suitability evaluation, Computer, Environment and Urban Systems, 26, P 89-112.
• Liu Mingqin, Samal Ashok, 2002. A fuzzy clustering approach to delineate agroecozones, Ecological Modelling, 149, P 215-228.
• Malczewski, J., 1999. GIS and multi criteria analysis, John Wiley & Son, New York, P458.
• Malczewski, Jacek, 2004. GIS-based land use suitability analysis: a critical overview, Progress in Planning, 62, P 3-65.
• Malczewski, Jacek, 2006. Ordered weighted averaging with fuzzy quantifiers: GIS-based multicriteria evaluation for land-use suitability analysis, International Journal of Applied Earth Observation and Geoinformation, Article in Press.
• Nisar Ahmad T. R., Gopal Rao K. Murthy J. S. R., 1998. GIS-Based fuzzy membership model for crop-land suitability analysis, Agriculture system, 63, P 75-95.
• Thirumalaivasan, D.., Karmegam, M., Venugopal, K., 2003. AHP-DRASTIC: Software for specific aquifer vulnerability assessment using drastic model and GIS, Environmental Modelling & software, V. 18, pp. 645-656.
• Yazdanpanah, Hojjat, Khalili Ali, Hajam S., Kamali G., Vezvai, 2003. Agro climatic zoning of Azarbayjan-Sharghi province for rain fed almond using GIS, GISdevelopment.net/application/agriculture.
• Zewen Li, Dong Jiang, Denghuai Lu, Cuizhi Zhang, 2003. The estimation of cotton-growing areas by remote sensing, isdevelopment.net/aars/acrs1990/posterpresentation.