Garlic Land Suitability System based on Spatial Decision Tree

Imas Sukaesih Sitanggang

, Andi Nurkholis

, Annisa

and Muhammad Asyhar Agmalaro

Department of Computer Science, IPB University, Darmaga,Jl. Meranti Wing 20, Level V,Bogor 16680, Indonesia

Keywords:

classiﬁcation, decision tree, garlic, land suitability, spatial ID3 algorithm

Abstract:

Domestic garlic needs are not equal to the production level that leads to high import of garlic to meet the

domestic consumption. Land suitability identiﬁcation for garlic is required in order to increase domestic

garlic production. The spatial classiﬁcation method can be used to determine garlic land suitability classes

based on the garlic planting criteria. This study aims to develop a web-based application to visualize the garlic

land suitability classes based on the spatial decision tree which was created using the spatial ID3 algorithm.

The application has four main functions namely visualization of garlic land suitability in map, user proﬁle

management, land suitability information, and garlic varieties information.

1 INTRODUCTION

Garlic is one of the horticultural crops commodities

that are needed by most of Indonesian, especially for

consumption as a cooking spice and food ﬂavoring.

Local garlic production in 2016 is 21,150 thousand

tons however the need of garlic reach 470,031 thou-

sand tons and it increases 8.78% in average per year

(BPS, 2017). Domestic garlic needs are not equal to

the production level, therefore import is done in order

to meet the domestic garlic consumption. The low

garlic production is the basis for the government to

launch a projection of garlic self-sufﬁciency. In 2020-

2029 the government will develop an area for growing

garlic. The plan aims to increase the level of domes-

tic garlic production (Holtikultura, 2017). There is

a problem in preparing the area for planting garlic,

namely the lack of land for garlic, so identifying the

land suitability of garlic id required (Statistik, 2015).

Land suitability identiﬁcation is conducted using land

suitability analysis techniques.

Land suitability evaluation is the process of esti-

mating land suitability classes and potential land uses

for agriculture. Land suitability class describes the

level of land suitability for a particular use. Land eval-

uation is conducted to assess the potential of land for

a certain crop by providing the growth requirements

for the crop.

Studies on land suitability have been done by

many researchers in several countries including In-

donesia.

A spatial model for land suitability evaluation for

wheat crop integrated with Geographical Information

System (GIS) was proposed (El Baroudy, 2016). The

proposed model showed that about 71.44% of the to-

tal area fall within the highly suitable class and the

moderately suitable class for wheat crop. (Chairani

et al., 2017) determines the physical land suitabil-

ity for civet Arabica coffee in Bandung and Bandung

Barat, Indonesia based on the criteria temperature,

rainfall, humidity, duration of dry season, slope, al-

titude, type of soil, soil texture, and erosion potential.

Land suitability using GIS for annual crops: Thanh

Tra pomelo, rubber, and stAcacia mangium in Thua

Thien Hue province Vietnam was evaluated based on

soil type, slope, terrain elevation, soil layer thickness,

mechanical composition, humus content, bio-climatic

conditions and irrigation (Dan, Ping and Lang, 2018).

The level of suitability to shallot and lemon in Harian

District, Samosir Regency Indonesia was determined

in the previous study (Tampubolon et al., 2018). (Se-

tyowati et al., 2018) estimates both production and

productivity of rice, maize, and cassava using land

suitability approach in Karangasem Regency, Bali In-

donesia. This study applied remote sensing and GIS

techniques. An assessment on a spatial basis was

conducted using GIS for agricultural land suitability

evaluation of rice (irrigated paddy ﬁeld, rainfed rice)

and corn (Ramlan et al., 2018). Spatial data used in

this study include digital topographic map, soil sur-

vey, soil characteristics, as well as climate data.

GIS and remote sensing technology have been

widely used in studies of land suitability evaluation.

However, the land suitability evaluation requires large

206

Sitanggang, I., Nurkholis, A., Annisa, . and Agmalaro, M.

Garlic Land Suitability System based on Spatial Decision Tree.

DOI: 10.5220/0009908002060210

In Proceedings of the International Conferences on Information System and Technology (CONRIST 2019), pages 206-210

ISBN: 978-989-758-453-4

spatial data of crop planting conditions especially for

large study area. In order to obtain interesting patterns

from the large spatial data in evaluating land suitabil-

ity, spatial data mining can be applied. One of spatial

algorithms that can be used in modelling land suitabil-

ity for crop is spatial decision tree. A spatial decision

tree is a classiﬁcation model for determining the class

label of objects based on their characteristics. Our

previous study has been successfully a spatial deci-

sion tree for garlic land suitability in Magetan district,

East Java province and Solok district, West Sumatra

province Indonesia. However, the model could not

easily be interpreted by users. Therefore, this study

aims to develop a visualization module of spatial de-

cision tree for garlic land suitability. The following

discussion is divided as follows. Section 2 discusses

data, study area, and brieﬂy explain the methods ap-

plied in this study. Result and discussions are brieﬂy

discussed in Section 3. We summarize our work and

future work in Section 4.

2 METHODS

2.1 Data

The study area is Magetan district, East Java province

Indonesia with an area of 70,143 ha (dan Pengemban-

gan Sumberdaya Lahan Pertanian (BBSDLP), 2017)

and Solok district, West Sumatra province Indonesia

with an area of 335,086.53 ha (dan Pengembangan

Sumberdaya Lahan Pertanian (BBSDLP), 2017). The

two districts are predicted to be the center for pro-

ducing garlic for Indonesia in the future (Holtikultura,

2017). The data used in this study are provided in Fig-

ure 1.

Figure 1: Data and Its Source.

2.2 Research Step

This study was conducted in several steps includ-

ing data preprocessing, spatial decision tree model

and developing a visualization module for garlic land

suitability. In the data pre-processing stage, we per-

formed spatial interpolation for weather data using the

ordinary co-kriging (Adhikary et al., 2017). In addi-

tion, spatial operations were applied on Digital Eleva-

tion Model (DEM) data to produce an elevation layer

in vector format. Explanatory and target layers for

spatial task relevant data were prepared after spatial

feature validity test was performed.

Spatial decision tree was created on the garlic

planting criteria dataset using the algorithm the spa-

tial ID3 algorithm (Sitanggang et al., 2013). The use

of spatial decision tree algorithms is based on the im-

portance of considering spatial data relations (i.e., po-

sition, distance, orientation, etc.) in the analysis of

geographically referenced data carried out in this case

of land suitability (Rinzivillo and Turini, 2004). The

tree was developed by splitting the spatial dataset into

smaller datasets by selected the best feature for data

partition. The measurement used to select the split-

ting feature is spatial information gain in which the

features with the highest spatial information gain will

be selected as the label of root of sub tree in each it-

eration.

The visualization module of spatial decision tree

was developed in order to provide the classiﬁcation

model for garlic land suitability. There are four main

functions available in the system namely 1) visualiza-

tion of garlic land suitability in map, 2) user proﬁle

management, 3) Land suitability information, and 4)

Garlic varieties information. The system was devel-

oped using the following software:

• Laravel as the backend framework to integrate

database and the user interface

• Bootstrap as the front end to provide webbased

interface

• Leaﬂet.js is used as the API frontend of map at

web-based interface

• PostgreSQL with the PostGIS extension as spatial

database management system

3 RESULTS AND DISCUSSION

The spatial ID3 algorithm (Sitanggang et al., 2013)

was implemented on the garlic spatial dataset to result

a spatial decision tree for garlic land suitability. The

dataset contains ten explanatory layers and one target

layer which is Land suitability. The attributes of each

layer for two study area namely Magetan dan Solok

are provided in Figure 2.

Garlic Land Suitability System based on Spatial Decision Tree

207

space

Figure 2: Spatial Data of The Garlic Planting Criteria.

The rules in land suitability evaluation can be ex-

tracted from garlic planting criteria dataset using the

classiﬁcation method. The class labels of the dataset

in this classiﬁcation task represent garlic land suitabil-

ity classes namely S1 (highly suitable), S2 (moder-

ately suitable), and S3 (marginally suitable).

The implementation of spatial ID3 algorithm on

Magetan dataset result as many 33 rules whereas on

the Solok dataset result as many 66 rules. The exam-

ples of rules generated from Magetan dataset are as

follows:

• IF relief = steep AND elevation = high AND tem-

perature = 24

◦

c AND rainfall = high AND depth

of soil mineral = deep AND Soil pH = slightly

acid AND soil texture = medium AND cation ex-

change capacity = low AND base saturation =

medium THEN land suitability = S1 (highly suit-

able)

• IF relief = steep AND elevation = low THEN land

suitability = S2 (moderately suitable)

• F relief = ﬂat AND rainfall = rather low AND

depth of soil mineral = very deep THEN land suit-

ability = S3 (marginally suitable)

The examples of rules generated from Solok

dataset are as follows:

• IF soil texture = smooth AND cation exchange ca-

pacity = height AND depth of soil mineral = very

deep THEN land suitability = S1 (highly suitable)

• IF soil texture = smooth AND cation exchange

capacity = very low AND relief = slightly steep

AND elevation = rather high AND temperature =

◦

c AND rainfall = high THEN land suitability

= S2 (moderately suitable)

• IF soil texture = slightly coarse AND Soil pH =

acid and base saturation = height THEN land suit-

ability = S3 (marginally suitable)

Figure 3 shows the Garlic land suitability in Solok

dan Magetan district. The area of garlic land suitabil-

ity for each class is illustrated in Figure 4, Figure 5

and Figure 6. While land suitability description and

garlic varieties recommended by the Indonesian gov-

ernment are shown in Figure 7 and Figure 8.

Figure 3: Garlic land suitability in (a) Solok dan (b) Mage-

tan

Figure 4: The area of garlic land suitability class S1 (highly

suitable) in Solok

The garlic varieties menu in Figure 8 provides

information on garlic varieties recommended by the

Ministry of Agriculture, which produces reliable seed

varieties with a yield of at least 9 tons/ha. This feature

can be used to obtain recommendation of garlic com-

modities to be cultivated in Indonesia, for the achieve-

ment of selfsufﬁciency in garlic by 2033.

Figure 5: The area of garlic land suitability class S2 (mod-

erately suitable) in Magetan

CONRIST 2019 - International Conferences on Information System and Technology

208

space

Figure 6: The area of garlic land suitability class S3

(marginally suitable) in Solok

Figure 7: Land suitability description

The land suitability description menu in Figure 7

provides information on land suitability classes and

their area in Solok and Magetan district. With this

feature, it is expected to provide information to rel-

evant parties (farmers, private sector, and Indonesian

government) in knowing the capabilities of Solok and

Magetan district for garlic farming.

Figure 8: Garlic varieties recommended by the Indonesian

government

Although land suitability analysis produces suit-

able class for garlic plant, it does not give a guarantee

that this area can be used for planting garlic. This

is because not all existing land use is used to grow

garlic. Figure 3 to Figure 6 show that there are la-

bels other than land suitability class (highly suitable,

moderately suitable, marginally suitable), namely set-

tlement area, water body, and unclassiﬁed. Therefore,

we need to decide the exiting land uses that can be

used for garlic plants. Figure 9 provides area of land

suitability class for garlic in Magetan and Solok dis-

trict.

space

Figure 9: Area of Land Suitability Class for Garlic.

In Figure 9 it can be seen that there are area that

cannot be planted with garlic because it has a func-

tion as settlement area and body of water. Settlement

area information indicates that cultivation/planting of

garlic in the area cannot be carried out due to social

conditions (the area is a community settlement), while

water body is an area that cannot be planted with gar-

lic because the terrain/location is water. The unclassi-

ﬁed area label is an area where the rule based model

can not predict the land suitability classes. Unclas-

siﬁed cases can be caused by the non-representative

dataset in creating classiﬁcation models, so that when

the resulting rules are applied to data for spatial vi-

sualization, the data cannot be classiﬁed into a land

suitability class.

A desktop application for Indonesian crop land

suitability has been developed based on the fuzzy in-

ference system (Insani et al., 2015). In addition, an

expert system based on fuzzy genetic was built for

land suitability of paddy and corn (Hartati and Sitang-

gang, 2010). Both studies do not include the spatial

relations that appear in the data of growing require-

ment of crop. The advantage of this proposed system

is its ability to include the spatial relations as the im-

portant aspect in the land suitability evaluation.

The development of geographic information sys-

tem of garlic land suitability in this study is expected

to be used as an interactive mapping visualization that

can be directly accessed by the community. This

application may support the target of the Indone-

sian government in achieving garlic self-sufﬁciency

in 2033 which one way to reach this target is to ex-

pand the existing garlic farmland. The application de-

veloped in this study is expected to facilitate the de-

termination of suitable land areas for garlic farming,

especially in Magetan and Solok district. Informa-

tion on land suitability class is expected to reduce er-

rors in determining the area to be used as garlic farm-

ing which is very important, because it will affect the

amount of garlic production.

The main advantage of this system is that the sys-

Garlic Land Suitability System based on Spatial Decision Tree

209

tem provides more speciﬁc information about land

suitability, namely to the village level. The result of

this study also shows that a village can consist of more

than one land suitability class. The system has been

able to accommodate this situation by providing in-

formation on the area of a land suitability class in a

village. Thus, users are expected to know the area of

land suitability S1 / S2 / S3 in a village in detail. An

example can be seen in Figure 2., which shows the

Garabak Data village has two land suitability classes,

namely S1 (blue legend) and S2 (green legend), where

S1 has an area of 3,774.49 ha. It is also made easier by

the availability of longitude and latitude information

that will show users to get to a location more accu-

rately.

4 CONCLUSIONS

This study has been successfully implemented the vi-

sualization module for spatial decision tree represent-

ing garlic land suitability. The spatial ID3 algorithm

results 33 rules on Magetan dataset and 66 rules on

Solok dataset. Those rules were implemented in a

web-based application that represent the area of garlic

land suitability classes in Magetan and Solok Indone-

sia. System testing will be conducted as the future

work by comparing the output of the application to

the knowledge from experts and real garlic land suit-

ability in the study area.

ACKNOWLEDGEMENTS

The authors would like to thank to Directorate of

Research and Community Service, Ministry of Re-

search, Technology and Higher Education, Republic

of Indonesia for the research grant.

REFERENCES

Adhikary, S. K., Muttil, N., and Yilmaz, A. G. (2017). Cok-

riging for enhanced spatial interpolation of rainfall in

two australian catchments. Hydrological processes,

31(12):2143–2161.

Chairani, E., Supriatna, J., Koestoer, R., and Moeliono, M.

(2017). Physical land suitability for civet arabica cof-

fee: Case study of bandung and west bandung regen-

cies, indonesia. In IOP Conference Series: Earth and

Environmental Science, volume 98, page 012029. IOP

Publishing Ltd.

dan Pengembangan Sumberdaya Lahan Pertanian (BBS-

DLP), B. B. P. (2017). Peta tanah semidetail kabu-

paten solok besar skala 1: 50.000. bogor (id).

Hartati, S. and Sitanggang, I. S. (2010). A fuzzy based

decision support system for evaluating land suitabil-

ity and selecting crops. Journal of Computer Science,

6(4):417.

Holtikultura, D. J. (2017). Pengembangan bawang putih

nasional.

Insani, F., Sitanggang, I. S., et al. (2015). Expert system

modeling for land suitability based on fuzzy genetic

for cereal commodities: Case study wetland paddy

and corn. Telkomnika, 13(3).

Ramlan, A., Baja, S., Arif, S., and Neswati, R. (2018).

Gis-based agroecological assessment of land suitabil-

ity for food crop development at a regional scale: A

study case of buton island. In IOP Conference Series:

Earth and Environmental Science, volume 157, page

012024. IOP Publishing.

Rinzivillo, S. and Turini, F. (2004). Classiﬁcation in ge-

ographical information systems. In European Con-

ference on Principles of Data Mining and Knowledge

Discovery, pages 374–385. Springer.

Setyowati, H., Murti, S., Nurani, R., Susanti, E., Puspi-

tasari, S., Ariyani, R., Firdausi, A., et al. (2018). Ap-

plied geo-eye imagery and gis for estimating crop pro-

duction based on land suitability in karangasem re-

gency, bali. In IOP Conference Series: Earth and En-

vironmental Science, volume 165, page 012025. IOP

Publishing.

Sitanggang, I. S., Yaakob, R., Mustapha, N., and Ainud-

din, A. (2013). Classiﬁcation model for hotspot occur-

rences using spatial decision tree algorithm. Journal

of computer science, 9(2):244.

Statistik, B. P. (2015). Produksi, luas panen dan produktiv-

itas sayuran di indonesia. Badan Pusat Statistik dan

Direktorat Jenderal Hortikultura, Jakarta.

Tampubolon, E. et al. (2018). Evaluation of land suitability

for shallot (a. ascalonicum l.) and orange (orange sp.)

at harian district of samosir regency. In IOP Confer-

ence Series: Earth and Environmental Science, vol-

ume 122, page 012035. IOP Publishing.

CONRIST 2019 - International Conferences on Information System and Technology

210