


































 | 109  
 

         Geoplanning 
Vol 4, No. 2, 2017, 109-120                                                                                                                                                        Journal of Geomatics and Planning 

                                                                                                 E-ISSN: 2355-6544 
http://ejournal.undip.ac.id/index.php/geoplanning 

doi: 10.14710/geoplanning.4.2.109-120 

LAND USE ANALYSIS USING TIME SERIES OF VEGETATION INDEX DERIVED FROM 
SATELLITE REMOTE SENSING IN BRANTAS RIVER WATERSHED, EAST JAVA, 
INDONESIA 

K. Yoshino a, Y. Setiawan b,d, E. Shima c 

a Faculty of Engineering, Information and Systems, University of Tsukuba, 1-1-1 Tennoudai, Tsukuba, Ibaraki, 305-8573, Japan 
b Faculty of Forestry, Bogor Agricultural University, Kampus IPB Darmaga, Bogor 16680, Indonesia 
c School of Veterinary Medicine, Kitasato University,23-35-1 Higashi, Towada, Aomori, 034-0005, Japan 
d Center for Environmental Research, Bogor Agricultural University, Kampus IPB Darmaga, Bogor, 16680, Indonesia  

Abstract: In this study, time series datasets of MODIS EVI (Enhanced Vegetation Index) 
data from 2002 and 2011 in the Brantas River watershed located in eastern Java, 
Indonesia were analyzed and classified to make ten land use maps for each year, in 
order to support watershed land use planning which takes into account local land use 
and trends in land use change. These land use maps with eight types of main land use 
categories were examined. During the 10 years period, forested area has expanded, 
while upland, paddy rice field, mixed garden and plantation have decreased. One of the 
reasons for this land use change is ascribed to tree planting under the joint forest 
management system by local people and the state forest corporation. 
 
  

 
Copyright © 2017 GJGP-UNDIP  

This open access article is distributed under a  
Creative Commons Attribution (CC-BY-NC-SA) 4.0 International license. 

 

 
How to cite (APA 6th Style): 
Yoshino, K., Setiawan, Y., & Shima, E. (2017). Land Use Analysis using Time Series of Vegetation Index Derived from Satellite Remote Sensing in 
Brantas River Watershed, East Java, Indonesia. Geoplanning: Journal of Geomatics and Planning, 4(2), 109-120. doi:10.14710/geoplanning.4.2.109-
120 

 

1. INTRODUCTION 

The sustainability of regional environment and society has become an important issue due to the effects 
of global warming being recognized throughout the world (Parry et al., 2007). Regional land use is easily 
affected not only by global warming but also socio-economic development and population growth (Kaneko 
et al., 1998). On the other hands, land use itself affects global environment and regional-global climate 
(Foley et al., 2005). Examination of regional land use over a long period of time is useful to determine any 
changes in the natural environment and society and to understand the trends in these changes, causes and 
processes (Himiyama & Okamoto, 1992). 

In regions where large-scale irrigation development projects such as dams, irrigation canals and so on all 
over the world (JICA, 2011) have been carried out, the projects have affected both regional society and the 
natural environment, since infrastructure development induces socio-economic development, then 
urbanization begins in the region. Generally, urbanization affects both regional society and environment. 
Monitoring regional land use over a long period is helpful to study the sustainability of natural environment, 
society and development projects (Ramankutty & Foley, 1999). 

Water infrastructure in Indonesia had been developed since the Dutch Colonization period, in order to 
avoid floods and to build irrigation system. Many of the infrastructures for irrigation and drainage 
constructed by these projects have remained and are still in use and providing social services to the region 
(Pasandaran, 2007). These large-scale developing projects under some urban policies on regional 
development could actually have affected regional environment (Amato et al., 2016). Indonesia is a suitable 
area to study sustainable agricultural development and future land use against global climate change. That 

Article Info: 
Received: 29 December 2016 
in revised form: 7 April 2017 
Accepted: 9 May 2017 
Available Online:  30 October 2017 
 

Keywords:  
Time series dataset, Land use 
classification, MODIS vegetation 
index, Brantas watershed 
 

Corresponding Author: 
Kunihiko Yoshino 
University of Tsukuba, 1-1-1 
Tennoudai, Tsukuba, Ibaraki, 305-
8573, Japan 
Email: sky@sk.tsukuba.ac.jp 

OPEN ACCESS 

https://doi.org/10.14710/geoplanning.4.2.109-120


 
Yoshino, Setiawan, Shima / Geoplanning: Journal of Geomatics and Planning, Vol 4, No 2, 2017, 109-120 
doi: 10.14710/geoplanning.4.2.109-120 

110 | 
 

is how they can modify their regional development plans in future, because they have a very long history of 
regional development for longer than one hundred years, especially in the areas or watersheds where 
regional socio-economies have been rapidly growing. These areas are thought to be dramatically changing 
due to the expansion of urban areas or agricultural land development by increasing population. In these 
areas, the newly developed lands have been expanding following the land use plans of the local 
government and also by illegal development. In order to control this illegal development, the 
characteristics of regional land use patterns and their changing trends should be understood in order to 
propose some applicable land use plans for watersheds and also to have better sustainability of the global 
ecosystems and human beings (Foley et al., 2011). 

Brantas watershed is the biggest watershed in East Java Province (BBWS Brantas, 2011). However, the 
local environmental agency indicated that vegetated lands in this watershed are under pressure, especially 
in upstream areas. Many existing land use allocations are inconsistent with spatial land use planning, even 
though the government rules, namely the Law No. 26/2007 on spatial planning and Law No. 41/2000 on 
basic forestry law, are stating that at least 30% of the area has to be allocated to vegetated land, such as 
garden and forest (BLH Prov. Jawa Timur, 2009). Based on these reports, it could be estimated that the 
agricultural lands or residential lands have disorderly invaded into the natural land use such as forest land. 
As a result, the forest environment has been degraded. The land use in this area could be intently changing 
to agricultural lands. 

The objectives of this research were, 1) to create annual land use maps of the Brantas River watershed 
in eastern Java, Indonesia from 2002 to 2011 using time series MODIS EVI data applying a wavelet de-noise 
filter, and 2) to study the characteristics of regional land use patterns and trends in land use change. This 
paper contains five chapters. After mentioning the backgrounds of this research, data and methods used in 
this research are described in the chapter 2. In the chapter 3, the results of land use mapping for 10 years 
in Brantas watershed using time series MODIS EVI data are given. In the chapter 3, we also discuss about 
land use patterns in this area and land use change during these 10 years. Finally, in the chapter 4, we 
conclude our findings resulted in this research. 

 

2. DATA AND METHODS 

2.1. Study site 

Brantas River watershed located in the eastern part of Java, Indonesia (Figure 1) and has a tropical 
monsoon climate. The annual average air temperature from 1996-2000 and the annual average 
precipitation are 25.12 degrees Celsius and 1,876 mm/year, respectively (Widianto et al., 2010; WMO, 
2013). Its area comprises approximately 12,000 km2 and Surabaya city is situated at the mouth of Brantas 
River, the second largest city in Indonesia with population of about 3 million. It is a famous agricultural area. 
Most of the land use types are: forest, plantation, mixed garden, rice paddy field, and uplands. Triple-rice 
cropping is undertaken in this watershed (Haruyama, Ooya, & Mizuhara, 1992). Recently, the socio-
economic situation in this watershed has rapidly risen along with an increasing population (Bhat, Ramu, & 
Kemper, 2005). Therefore, rational land use and sustainable environmental planning are necessary (JBIC 
Institute, 2008). 

The downstream area of Brantas River has frequently suffered from flooding such as the large flood 
occurred in 2010. The flooding caused damage to crops, inability to cultivate land due to water logging of 
soils, disruption of settlement, transportation and loss of property. To counter these problems, many flood 
control projects have been carried out by the Indonesian government since the 1950’s as well as other 
large scale flood control projects such as construction of big dams and irrigation-drainage projects under 
Japanese-aid (JICA, 2011). However, in spite of these projects and the forest management policy of 
Indonesia’s government, floods have frequently occurred (Hidayat, 2009). In addition, dynamics changes of 
forest cover are found in forested areas (Setiawan, Yoshino, & Prasetyo, 2014), and sedimentation on the 
river floor has become severe (Adi, Jänen, & Jennerjahn, 2013; Widianto et al., 2010). It is therefore 
assumed that there are other reasons for frequent floods. The characteristics of land use patterns and the 
trends in land use change in this watershed should be understood in detail to facilitate rational watershed 
management and sustainable agricultural development. 

https://doi.org/10.14710/geoplanning.4.2.109-120


 
Yoshino, Setiawan, Shima / Geoplanning: Journal of Geomatics and Planning, Vol 4, No 2, 2017, 109-120      

doi: 10.14710/geoplanning.4.2.109-120 
                                                              

 | 111  
 

 
 

 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 1. Outline map of Brantas River watershed 

 
2.2. Remotely sensed imagery in this study 

Monitoring land use precisely changes in large regions over a long period of time and to carry it out in 
detail is generally difficult. Recently, satellite optical remotely sensed images are used to study land use and 
land cover both on a global scale and in regional scale (Di Palma et al., 2016). However, there are some 
shortcomings in applying this method to tropical regions where the cloud coverage rate is high (Setiawan, 
Yoshino, & Philpot, 2013). Time series MODIS (MODerate resolution Imaging Spectroradiometer) data 
recorded by two earth observation satellites, Terra and Aqua, are valid to obtain land use and land cover 
change over a global scale or regional scale (Friedl et al., 2002; Sakamoto et al., 2006). These two satellites 
observe the whole surface of the earth twice a day. Secondary MODIS datasets that were processed from 
observed data for specific purposes are open to the public and can be downloaded via the internet 
(LPDAAC, 2013a). These datasets provide information on the condition and dynamics of the earth surface 
such as land use/land cover in short time intervals. 

The time series MODIS EVI (Enhanced Vegetation Index) produced from the observed data were used 
in this study. The datasets are referenced as H29V9 of MODIS Sinusoidal Tiling System (LPDAAC, 2013b) and 
cover the study site. The 230 scenes of MODIS EVI datasets, MYD13Q1 (LPDAAC, 2013d) observed from July 
2002 to June 2012, a period of 10 years. The 16-days composite data with the spatial resolution of 250m 
were downloaded from the NASA’s web-site (NASA, 2013). These datasets comprising point data of MODIS 
original scenes for 16 days as the probability for cloudless observation of the ground is considered optimal 
during this time period.  

https://doi.org/10.14710/geoplanning.4.2.109-120


 
Yoshino, Setiawan, Shima / Geoplanning: Journal of Geomatics and Planning, Vol 4, No 2, 2017, 109-120 
doi: 10.14710/geoplanning.4.2.109-120 

112 | 
 

EVI is one of the vegetation indices developed by Huete et al. (2002). It was computed using band data 
from the visible blue band, visible red band and infra-red band. It reduced the effects of variation in the sun 
irradiance, thin cloud or cloud shadow, topographic effects, sun elevation and view-angle effects. Moreover, 
atmospheric effects such as scattering and absorption are removed as it uses visible blue band data. 
However, additional data processing is necessary, since datasets from tropical regions are affected by cloud 
noise (Solano, Didan, & Jacobson, 2013). 

 
2.3. Procedures for data processing 

The 230 scenes of MODIS EVI 16-days composite datasets for the 10 years were aggregated and 
analyzed as follows. Every processing algorithm is standard and popular in order to analyze satellite remote 
sensing data.  
[1]  Reprojection of each image using MRT (LPDAAC, 2013c): an application was used for reprojection of 

datasets from the MODIS original projection system to WGS-84 geographical projection system. For 
computation of area of land use, images were reprojected to UTM coordinates system zone 49S and 
resampled at every 250m. At the same time, the QA (Quality assurance) of each dataset was validated. 

[2]  De-noising of the aggregated dataset: In order to remove the spike-like noises per pixel from the time 
series data in the aggregated dataset, a wavelet transformation filter was applied to the time 
sequential EVI of every pixel in the aggregated dataset using MATLAB wavelet tool (Mathworks, 2009),  
then a smooth time sequential EVI dataset was obtained. The wavelet model was the Coiflet model, 
the same model used by Sakamoto et al. (2006). The order 2 for the wavelet was applied in order to 
retain the original characteristics of the data. This de-noised time sequential data was divided into 10 
datasets corresponding to each year. 

[3]  Supervised classification for land use mapping: The de-noised image of the base year was classified 
with a supervised classification algorithm using reference training data. 

[4]  Computation of correlation coefficients (Morita, 1985) between the base year dataset and the target 
year dataset: A pixel of each one-year dataset is a 23 dimensional vector. The correlation coefficients 
between pixels of the base year dataset and those of other one-year datasets were computed. The 
dataset from 2007 was retained as the base year dataset. Later, pixels which had higher correlation 
coefficients than 0.8 were chosen as training samples for supervised classification. 

[5]  Statistics of signature of each land use category were computed using the training samples chosen in 
previous step [4]. 

[6]  Supervised classification for land use mapping: The algorithm of MLH (Maximum Likelihood) 
supervised classification was applied to map land use, and a land use map of each year was drawn.  

[7]  Characterization of trends: Using a land use map of each year, the characteristics of land use and the 
change in this watershed were analyzed and discussed. 

 
 

3. RESULTS AND DISCUSSION 

3.1. Land use categories for classification and land use mapping of the base year 

Nineteen land use categories were determined (Table 1) by the previous work (Setiawan, Yoshino, & 
Philpot, 2013). These 19 land use categories were those that could be selected as training data in this 
watershed. There were four categories of rice paddy fields, four of upland fields, one mixed garden, five of 
forest lands, three of plantations, one urban area, and one water surface. They are appropriate land use 
categories to study land use and land use change in this watershed. Totally 13,723 pixels were selected as 
the training samples. The de-noised image of the base year, a dataset from 2007, was classified with a 
supervised classification algorithm using these training data. This land use map in 2007 is used as a base or 
reference land use map for analysis of datasets of other years.  The reason of 2007’s dataset selection for 
creating a base land use map in this study is that a reliable land use map was drawn using satellite images 
taken in this year (Department of Forestry, 2008). 

 
 

 

https://doi.org/10.14710/geoplanning.4.2.109-120


 
Yoshino, Setiawan, Shima / Geoplanning: Journal of Geomatics and Planning, Vol 4, No 2, 2017, 109-120      

doi: 10.14710/geoplanning.4.2.109-120 
                                                              

 | 113  
 

Table 1. Reclassification of land use categories in this study (analysis, 2015) 

Land use category 
(Setiawan et al., 2013) 

Re-classed land use category 

1. Triple irrigated paddy field 1. Triple crop paddy  

2. Double irrigated paddy field I 2. Double crop paddy  

3. Double irrigated paddy field II 

4. Double irrigated paddy field III 

5. Upland 3. Upland 

6. Upland with intensive agricultural land 

7. Upland with mixed agricultural land 

8. Irrigated fields 

9. Mixed garden with bush 4. Mixed garden 

10. Forest mixed with bush 5. Forest 

11. Bush mixed grass 

12. Dryland forest 

13. Heterogeneous mangrove forest 

14. Mangrove 

15. Rubber plantation 6. Plantation 

16. Oil palm plantation 

17. Timber forest plantation 

18. Builtup 7. Urban area 

19. Pond 8. Water surface 

 
In order to provide a consistent definition for pixel based analysis, we need to resolve the differences 

among the nineteen land use categories in the dataset. The process of reassigning land use categories 
based on the knowledge of the each land use characteristic is shown in Table 1. Nineteen land use 
categories were re-classed into eight categories: triple cropping rice paddy field, double cropping rice 
paddy field, upland field, mixed garden, forest, plantation, urban area, and water surface. When tabulating 
an error matrix for evaluation of classification (Richards, 2006), it was found the overall classification 
accuracy was 71 % for these 8 categories. In consideration of the spatial resolution (250 m) used on this 
dataset, this land use map is assumed to show a true land use pattern in this watershed around 2007. 

3.2. Sampling training data based on the correlation coefficients 

In general, strictly selected reference data should be used for supervised classification. Usually, 
samples are selected by designating several pixels which are thought to be proper training areas of each 
land use category with several ways such as ground survey, referring reliable land use maps in order to 
obtain sufficient number of training samples (Oobayasi & Kojima, 2002). However, it is important to collect 
fine training data for every image to detect land use change in short periods from the results of land use 
classification in every year.  

In this study, training samples were collected by selecting pixels which have higher correlation 
coefficients than 0.8 between the base year dataset of 2007. The other year datasets where those pixels 
have the same land use categories as the pixels in the base-year dataset were equally used, because only in 
2007, we have a reliable land use map authorized by the Ministry of Forestry of Indonesia.  For accessing 
the similarity of two spectral signatures which are recognized as two vectors of multiple variables, the angle 
between two vectors is observed as the similarity of two vectors. The smaller angle means that they have 
high similarity.  

The value of COSIN of that angle is equal to the correlation coefficients in terms of the fundamental 
relationship between geometry and statistics. So, in this research, the correlation coefficients are used to 
measure the similarity of two vectors of the same pixels of two years. Figure 2 shows the accumulated 
frequency in percent of the correlation coefficients between each year and the base year dataset of 2007. 
Although the correlation coefficients of 2010 looks different from those of the other years, over 30 to 45% 

https://doi.org/10.14710/geoplanning.4.2.109-120


 
Yoshino, Setiawan, Shima / Geoplanning: Journal of Geomatics and Planning, Vol 4, No 2, 2017, 109-120 
doi: 10.14710/geoplanning.4.2.109-120 

114 | 
 

of pixels in the other year datasets have  correlation coefficients higher than 0.8. Thus, at least over 100 
pixels were selected as training samples for each land use category in each year. 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 2. Accumulated histogram of correlation coefficients  

(the first 2 digit number in the legend shows the year of interest) 
 
 
 

3.3. Supervised classification result 

Supervised classification of land use for each year dataset except the dataset of 2007 was conducted 
with MLH (Maximum Likelihood method) using training samples selected in the former section. As example, 
the land use classification maps for 2002 and 2011 are shown in Figure 3.  In these two land use maps, it is 
clear that land use in this area has been changed in many places.  

In general, rice paddy fields widely spread downstream from Kediri and the mountains are covered 
with forest. Most lands except rice paddy fields and forest are upland field, mixed garden and plantation. In 
the coastal area of the eastern part of Surabaya, large fish ponds are expanding. However, they were 
classified into water surface. 

 
3.4. Land use change from 2002 to 2011 

The acreages of each land use category in each year from 2002 to 2011 are shown in km2, to study the 
trends of land use change in Brantas River watershed (Figure 4). Variations of acreages in each year are 
rather large for many land use categories. One of the causes of these large variations of estimated acreages 
of land use classes is assumed to be attributed to the intrinsic drawback of MODIS composite products. A 
pixel value is affected by the viewing geometry of MODIS sensor of observation dates which changes the 
dimension of the pixel on the ground, so that a pixel consists of spectral signals from ground cover 
conditions or land use of surrounding pixels. This intrinsic drawback leads a fairly large number of 
misclassification of MODIS composite data (Tan et al., 2006). As this drawback cannot be compensated, we 
will ignore the uncertainty of classification results in this paper.  

Moreover, the previous study in East Java (Muhammad et al., 2016) also mentioned that the mixed 
pixel issue was an important consideration of land use classification using MODIS data since the 
classification result of specific land use classes revealed the overall accuracy to be 57.7 %. 

https://doi.org/10.14710/geoplanning.4.2.109-120


 
Yoshino, Setiawan, Shima / Geoplanning: Journal of Geomatics and Planning, Vol 4, No 2, 2017, 109-120      

doi: 10.14710/geoplanning.4.2.109-120 
                                                              

 | 115  
 

Apparently, the forest increased during the past 10 years, while other land use categories such as 
double cropping rice paddy, upland, mixed garden, and plantation seem to decrease. Water surface, triple 
cropping rice paddy field and urban area maintained their acreages. 

The five years average acreage for each land use category from 2002 to 2006 was compared to those 
from 2007 to 2011 to summarize the trends of land use changes during the ten years period. As for forest, 
upland field and single cropping rice paddy field, they have increased to 718, 48 and 5 km2 in acreage, 
respectively. By contrast, double cropping rice paddy field, triple cropping paddy field, mixed garden, 
plantation and urban area have decreased to 345, 130, 124, 112, and 59 km2, respectively. 

Land use categories such as forest, upland field and double cropping rice paddy field showed a rapid 
change in their acreage on a scale of several tens thousands hectares since 2009 compared to other years.  
Forest area has increased, on the other hand, upland field and double cropping rice paddy field have 
decreased. The total annual precipitation of 2,483 mm and 2,274 mm during the period from July, 2009 to 
June, 2010 and during the period from July 2010 to June, 2011 in Surabaya area, respectively, was the 
reason of this change. These heavy rainfalls were recorded at the Surabaya meteorological observation 
station in this period. They were calculated based on precipitation records. These precipitation exceeded 
the 30 year average annual precipitation of 1,876 mm (WMO, 2013). Due to the natural physiological 
response of tropical plants against sufficient rainfall, vegetation in this area consequently grew further 
between 2009 and 2010 than in other periods, so that the annual EVI patterns in this period are understood 
to show different patterns from those of other periods. 

 

 

 

 

Figure 3. Eight categories land use maps in 2002 (left) and 2011 (right) 

 

 

 

https://doi.org/10.14710/geoplanning.4.2.109-120


 
Yoshino, Setiawan, Shima / Geoplanning: Journal of Geomatics and Planning, Vol 4, No 2, 2017, 109-120 
doi: 10.14710/geoplanning.4.2.109-120 

116 | 
 

 

 

 
 

Figure 4. Land use change in Brantas River watershed 2002-2011 
 

 
3.5. Characteristics of land use change in the watershed 

3.5.1. Land use change matrix 

Table 2 tabulated the land use change matrix between 2002 and 2011. The rows are the acreage for 
each land use in 2002, while the columns are those from 2011 in km2. The values in the diagonal items are 
unchanged acreages of the same land use categories both in 2002 and 2011. The other items indicate the 
land use changes from 2002 to 2011. In this table, the conversion of land use from 2002 to 2011 focused on 
changes more than one hundred km2; 110 km2 of triple cropping rice paddy field changed to forest, 490 km2 
of double cropping rice paddy field changed to upland, 457 km2 to forest, 151 km2 to plantation and 107 
km2 to urban area.  

About 498 km2 of upland have changed to forest, 321 km2 to plantation, 236 km2 to double cropping 
rice paddy, and 132 km2 to urban area. Meanwhile, 189 km2 of mixed garden have been converted into 
forest, 142 km2 into plantation, 306 km2 into upland, respectively. Then, 478 km2 of forest have been 
converted into upland, 311 km2 into plantation, 306 km2 into double cropping rice paddy field, 577 km2 of 
plantation into forest, 391 km2 into upland, 124 km2 into mixed garden, 94 km2 into double cropping rice 
paddy field, 85 km2 into forest. Moreover, 17 km2 of water surface have changed to upland, 10 km2 to 
double cropping rice paddy field, 12 km2 to urban area. 

3.5.2. Long term land use change 

In order to assess the long term stability of land use from 2002 to 2011, the unchanged land use of 
every land use category was examined by computing the acreages that did not change between 2002 and 
2011 for each land use category (See the last line of Table 2).  The total acreage of unchanged lands in 8 
land use categories was about 872 km2. Other lands have temporally or almost permanently changed to 
other land use. About 219 km2 of rice paddy field, 477 km2 of forest, 54 km2 of mixed garden, 84 km2 of 
water surface, 9 km2 of upland field, 13.1 km2 of plantation and 15.8 km2 of urban area respectively have 
not changed.  

https://doi.org/10.14710/geoplanning.4.2.109-120


 
Yoshino, Setiawan, Shima / Geoplanning: Journal of Geomatics and Planning, Vol 4, No 2, 2017, 109-120      

doi: 10.14710/geoplanning.4.2.109-120 
                                                              

 | 117  
 

Forest distributed on the hillside of Mt. Kelud was unchanged. Unchanged rice paddy fields are seen 
in flat plain downstream of the Brantas River. Unchanged mixed garden is expanding in the suburban area 
near Kediri town. Unchanged water surface remains in the tidal area in the river mouth of the Brantas River. 
Unchanged triple cropping rice paddy fields are recognized upstream of the Brantas River. The unchanged 
urban area and upland fields are scattered. They are not spatially accumulated. 

 
Table 2. Land use change matrix between 2002 and 2011 

Land use in 
2002 

Land use in 2011 

Triple crop 
paddy 

Double crop 
paddy 

Upland 
field 

Mixed 
garden 

Forest Plantation Urban 
area 

Water 
surface 

Total 

Triple crop 
paddy 

139.7 62.1 40.2 21.3 110.7 15.3 21.6 0.1 411.0 

Double 
crop paddy 

46.9 1153.4 490.8 96.1 456.6 150.9 106.5 6.4 2507.6 

Upland 
field 

21.1 235.8 721.6 69.0 498.1 320.5 131.9 91.4 2089.3 

Mixed 
garden 

14.3 61.3 101.7 417.0 188.6 142.1 10.9 0.1 936.0 

Forest 57.6 306.3 477.6 88.9 2017.3 311.1 53.4 20.4 3332.6 

Plantation 9.1 107.1 391.2 123.8 577.1 601.4 26.9 0.3 1836.9 

Urban area 7.2 93.3 116.8 4.2 85.3 19.4 251.3 37.8 615.3 

Water 
surface 

0.1 9.8 16.7 0.0 6.0 0.2 12.0 184.9 229.6 

Total 296.1 2028.9 2356.4 820.3 3939.6 1561.0 614.6 341.4 11958.3 

Unchanged 30.8 188.0 9.1 54.3 477.1 13.1 15.8 83.8 871.9 

         (Unit km2) 

 
 

3.6. Driving forces of land use change 

According to the results of land use classification in this study, forest has increased its acreage 
between the first 5 years and the last 5 years. Furthermore, rice paddy field, upland field, mixed garden and 
plantation have decreased their acreages 100 km2 to several 100s km2. These results are contrary to 
expectation which was described in the introduction. As for the increase of forest, it is assumed that 
afforestation projects conducted by community based forest management system (CBFM) under the forest 
management system promoted by Perhutani Indonesia are working well to conserve the forest 
environment (Asia Forest Network, 2004). The fact that forest in this area has been steadily increasing 
indicates that the regional environmental policies in this watershed have begun to show their effects to 
regulate land use pattern and to mitigate the imprudent land development (Amato et al., 2016). Since the 
increase of forest in the hillside reduces the surface soil loss rates from the sparsely vegetated lands to the 
river, soil suspension of the river water could get lower, then, the sedimentation on the river floor in the 
river streams would decline (Yoshino & Ishioka, 2005). Consequently, the frequency and the severalty of 
flooding of Brantas River will decrease in the near future. 

Additionally, contrary to our expectation that the forest lands have intently decreased, the urban area 
did not significantly expand during the 10 years. One possible reason for this is that the spatial resolution of 
MODIS EVI, which is 250 m x 250 m, is too large to detect urban change. The small changes in urban area 
could be buried in this large spatial resolution of MODIS data. In order to detect area and location of such 
kinds of small scale land use changes in nearly real time, it is better to use the satellite images with higher 
spatial resolution. Plus, land use classification accuracies in this study are problematic. The highest overall 
accuracy was about 71% in 2007, while in other years, they were at most 50%. Setiawan, Yoshino, & Philpot 
(2013) reported that over 40% of pixels from MODIS EVI data are mixed pixels that contain several land 
uses. The low spatial resolution of MODIS EVI dataset and temporal composite dataset of MODIS data 

https://doi.org/10.14710/geoplanning.4.2.109-120


 
Yoshino, Setiawan, Shima / Geoplanning: Journal of Geomatics and Planning, Vol 4, No 2, 2017, 109-120 
doi: 10.14710/geoplanning.4.2.109-120 

118 | 
 

products mainly resulted in the low land use classification accuracy (Campagnolo et al., 2016; Tan et al., 
2006), although the satellites mounting MODIS sensors daily and observing the terrestrial surface and the 
remote sensed images are supposed to be useful to monitor temporal environmental changes of the earth. 
For detecting tiny land use changes over a wide extent of over several 100’s km, datasets of high spatial 
and low temporal resolution are more suitable for analysis, although some troublesome problems remain 
to be solved. Actually, the answers to these problems depend on the research objectives. 

 

4. CONCLUSION 

Comprehensive land use planning based on long term monitoring of watershed is necessary in 
developing countries. This research selected the Brantas River watershed in eastern Java, Indonesia as a 
study area and analyzed land use change from 2002 to 2011 using a time sequential MODIS EVI dataset.  

The results of this study were as follows: 1) during the 10 year period, forest had a tendency to increase 
acreage, while other land use categories, especially arable lands have decreased, 2) only 872 km2 of land 
remained unchanged in this watershed during these ten years, 3) examination of a land use change matrix 
between 2002 and 2011 clarified many types of land conversion, 4) afforestation projects conducted by 
local people promoted under the policy of Perhutani Indonesia could be one of the causes of land use 
changes in this watershed.  

Lastly, the overall accuracies of classification were not highly adequate to deal with land use change in 
detail. Congalton (1991) reported that over 85% to 90% of overall accuracy of land use classification was 
necessary to study land use change. A much higher classification accuracy to detect tiny land use changes 
was indispensable. The following future works are essential to obtain more detailed results on land use and 
land use changes in this area: a) The determination of timing of land use changes, locations and driving 
forces for these changes, b) Analysis of the temporally trajectories of land use change and cycles of land use 
which seen in rice paddy fields or in traditional shifting cultivation, c) Research on optimal watershed 
management policy for the sustainable watershed ecosystem. 

 

5. ACKNOWLEDGMENTS 

This research was financially supported by research grand-aid of JSPS (Overseas research project 
22402033) from 2010 to 2012. We very much appreciate anonymous reviewers for their very helpful 
comments and suggestions to improve our manuscript. 

 

6. REFERENCES 

Adi, S., Jänen, I., & Jennerjahn, T. C. (2013). History of development and attendant environmental changes 
in the Brantas River Basin, Java, Indonesia, since 1970. Asian Journal of Water, Environment and 
Pollution, 10(1), 5–15. 

Amato, F., et. al. (2016). The effects of urban policies on the development of urban areas. Sustainability, 
8(4), 297. [Crossref] 

Asia Forest Network. (2004). Communities Transforming forestland, Java, Indonesia. 
BBWS Brantas. (2011). Balai Besar Wilayah Sungai Brantas Edisi 2011. Directorate General of Water 

Resources. Ministry of Public Works and Housing, Republic of Indonesia. 
Bhat, A., Ramu, K., & Kemper, K. (2005). Institutional and policy analysis of river basin management -The 

Brantas River Basin, East Java, Indonesia. World Bank Policy Research Working Paper 3611, May 2005. 
BLH Prov. Jawa Timur. (2009). Land cover change in Brantas watershed for year 2000-2008. 
Campagnolo, M. L., Sun, Q., Liu, Y., Schaaf, C., Wang, Z., & Román, M. O. (2016). Estimating the effective 

spatial resolution of the operational BRDF, albedo, and nadir reflectance products from MODIS and 
VIIRS. Remote Sensing of Environment, 175, 52–64. [Crossref] 

Congalton, R. G. (1991). A review of assessing the accuracy of classifications of remotely sensed data. 
Remote Sensing of Environment, 37(1), 35–46. [Crossref] 

 

https://doi.org/10.14710/geoplanning.4.2.109-120
https://doi.org/10.3390/su8040297
https://doi.org/10.1016/j.rse.2015.12.033
https://doi.org/10.1016/0034-4257(91)90048-b


 
Yoshino, Setiawan, Shima / Geoplanning: Journal of Geomatics and Planning, Vol 4, No 2, 2017, 109-120      

doi: 10.14710/geoplanning.4.2.109-120 
                                                              

 | 119  
 

Department of Forestry. (2008). Recalculation of Land Cover in Indonesia 2008. The Agency of Forestry 
Planning, Department of Forestry, Republic of Indonesia. 

Di Palma, F., et. al. (2016). A SMAP Supervised Classification of Landsat Images for Urban Sprawl Evaluation. 
ISPRS International Journal of Geo-Information, 5(7), 109. [Crossref] 

Foley, J. A., et. al. (2005). Global consequences of land use. Science, 309(5734), 570–574. [Crossref] 
Foley, J. A., et. al. (2011). Solutions for a cultivated planet. Nature, 478(7369), 337–342. [Crossref] 
Friedl, M. A., et. al. (2002). Global land cover mapping from MODIS: algorithms and early results. Remote 

Sensing of Environment, 83(1), 287–302. [Crossref] 
Haruyama, S., Ooya, M., & Mizuhara, Y. (1992). Development and conservation of rivers in active volcane 

zones -case study of Brantas watershed in the eastern Java, Indonesia. Journal of Geography, 101(2), 
89–106. 

Hidayat, F. (2009). Floods and climate change--Observations from Java. CRBOM Small Publicafions. 
Himiyama, Y., & Okamoto, J. (1992). Land use change and causes. Committee of agricultural policy survey, 

Tokyo. 
Huete, A., et. al. (2002). Overview of the radiometric and biophysical performance of the MODIS vegetation 

indices. Remote Sensing of Environment, 83(1–2), 195–213. [Crossref] 
JBIC Institute. (2008). Aid Effectiveness to Infrastructure: A Comparative Study of East Asia and Sub-Saharan 

Africa: Case Studies of East Asia. JBICI Research Paper No.36-2. 
JICA. (2011). History-Goods for the future generation-The engineers spirits cultivated by one river. JICA’s 

World, Special topic. 
Kaneko, K., et al. (1998). Iwanami Kouza, Global environment. Sustainable Social Systems, 10. 
LPDAAC. (2013a). MODIS. Retrieved from https://lpdaac.usgs.gov/products/modis_products_table.htm  
LPDAAC. (2013b). MODIS Overview. Retrieved from https://lpdaac.usgs.gov/products/modis_overview 
LPDAAC. (2013c). MODIS Reprojection Tool. https://lpdaac.usgs.gov/tools/modis_reprojection_tool.htm 
LPDAAC. (2013d). Vegetation Indices 16-Day L3 Global. Retrieved from 

https://lpdaac.usgs.gov/products/modis_products_ table/myd13q1 
Mathworks. (2009). Matlab Online Documentation. Tech. Rep. The Mathemowrks, Inc. 
Morita, Y. (1985). Introduction to new statistics. Nihon-hyouronn Inc., Tokyo. 
Muhammad, M., et al. (2016). Analysis of the dynamics pattern of paddy field utilization using MODIS 

image in East Java. Procedia Environmental Sciences, 33, 44–53. [Crossref] 
NASA. (2013). Reverb|ECHO. Retrieved from http://reverb.echo.nasa.gov/reverb/ 
Oobayasi, N., & Kojima, N. (2002). Remote Sensing for strategists. Tokyo: Fuji Tecnosis System. 
Parry, M. L., et al. (2007). Contribution of Working Group II to the Fourth Assessment Report of the 

Intergovernmental Panel on Climate Change. Cambridge University Press. 
Pasandaran, E. (2007). Management of Infrastructure irrigation in national food security framework. 

Analisis Kebijakan Pertanian, 5(2), 126–149. 
Ramankutty, N., & Foley, J. A. (1999). Estimating historical changes in global land cover: Croplands from 

1700 to 1992. Global Biogeochemical Cycles, 13(4), 997–1027. [Crossref] 
Richards, J. (2006). Remote Sensing Digital Image Analysis. https://doi.org/10.1007/978-3-642-30062-2 
Sakamoto, T., et al. (2006). Spatio--temporal distribution of rice phenology and cropping systems in the 

Mekong Delta with special reference to the seasonal water flow of the Mekong and Bassac rivers. 
Remote Sensing of Environment, 100(1), 1–16. [Crossref] 

Setiawan, Y., Yoshino, K., & Philpot, W. D. (2013). Characterizing temporal vegetation dynamics of land use 
in regional scale of Java Island, Indonesia. Journal of Land Use Science, 8(1), 1–30. [Crossref] 

Setiawan, Y., Yoshino, K., & Prasetyo, L. B. (2014). Characterizing the dynamics change of vegetation cover 
on tropical forestlands using 250m multi-temporal MODIS EVI. International Journal of Applied Earth 
Observation and Geoinformation, 26, 132–144. [Crossref] 

Solano, R., Didan, K., & Jacobson, A. (2013). MODIS Vegetation Index user’s Guide version 2.00. Retrieved 
from http://vip.arizona.edu/documents/MODIS/ MODIS_VI_Users Guide_ 01_2012.pdf   

Tan, B., et al. (2006). The impact of gridding artifacts on the local spatial properties of MODIS data: 
Implications for validation, compositing, and band-to-band registration across resolutions. Remote 
Sensing of Environment, 105(2), 98–114. [Crossref] 

Widianto, W., Suprayogo, D., Sudarto, S., & Lestariningsih, I. D. (2010). Implementasi Kaji Cepat Hidrologi 
(RHA) di Hulu DAS Brantas, Jawa Timur. Working Paper World Agroforestry Centre, 121. 

https://doi.org/10.14710/geoplanning.4.2.109-120
https://doi.org/10.3390/ijgi5070109
https://doi.org/10.1126/science.1111772
https://doi.org/10.1038/nature10452
https://doi.org/10.1016/s0034-4257(02)00078-0
https://doi.org/10.1016/s0034-4257(02)00096-2
https://doi.org/10.1016/j.proenv.2016.03.055
https://doi.org/10.1029/1999gb900046
https://doi.org/10.1016/j.rse.2005.09.007
https://doi.org/10.1080/1747423x.2011.605178
https://doi.org/10.1016/j.jag.2013.06.008
https://doi.org/10.1016/j.rse.2006.06.008


 
Yoshino, Setiawan, Shima / Geoplanning: Journal of Geomatics and Planning, Vol 4, No 2, 2017, 109-120 
doi: 10.14710/geoplanning.4.2.109-120 

120 | 
 

WMO. (2013). World Weather Information Services. Retrieved from http://www.worldweather.org/043 
/c00648.htm  

Yoshino, K., & Ishioka, Y. (2005). Guidelines for soil conservation towards integrated basin management for 
sustainable development: a new approach based on the assessment of soil loss risk using remote 
sensing and GIS. Paddy and Water Environment, 3(4), 235–247. [Crossref] 

https://doi.org/10.14710/geoplanning.4.2.109-120
http://www.worldweather.org/043%20/c00648.htm
http://www.worldweather.org/043%20/c00648.htm
https://doi.org/10.1007/s10333-005-0023-5

	doi: 10.14710/geoplanning.4.2.109-120
	How to cite (APA 6th Style):
	Yoshino, K., Setiawan, Y., & Shima, E. (2017). Land Use Analysis using Time Series of Vegetation Index Derived from Satellite Remote Sensing in Brantas River Watershed, East Java, Indonesia. Geoplanning: Journal of Geomatics and Planning, 4(2), 109-12...
	1. INTRODUCTION
	The sustainability of regional environment and society has become an important issue due to the effects of global warming being recognized throughout the world (Parry et al., 2007). Regional land use is easily affected not only by global warming but a...
	In regions where large-scale irrigation development projects such as dams, irrigation canals and so on all over the world (JICA, 2011) have been carried out, the projects have affected both regional society and the natural environment, since infrastru...
	Water infrastructure in Indonesia had been developed since the Dutch Colonization period, in order to avoid floods and to build irrigation system. Many of the infrastructures for irrigation and drainage constructed by these projects have remained and ...
	Keywords:
	Time series dataset, Land use classification, MODIS vegetation index, Brantas watershed
	Corresponding Author:
	Kunihiko Yoshino
	University of Tsukuba, 1-1-1 Tennoudai, Tsukuba, Ibaraki, 305-8573, Japan
	Email: sky@sk.tsukuba.ac.jp
	Brantas watershed is the biggest watershed in East Java Province (BBWS Brantas, 2011). However, the local environmental agency indicated that vegetated lands in this watershed are under pressure, especially in upstream areas. Many existing land use al...
	The objectives of this research were, 1) to create annual land use maps of the Brantas River watershed in eastern Java, Indonesia from 2002 to 2011 using time series MODIS EVI data applying a wavelet de-noise filter, and 2) to study the characteristic...
	2. DATA AND METHODS
	3. RESULTS AND DISCUSSION
	4. CONCLUSION
	5. ACKNOWLEDGMENTS
	6. REFERENCES

