Materials and Methods

In this study, the driving factors that affecting environmental quality index (EQI) in Java Island were
identifying using GWR model. GWR is a statistical method to identify local spatial variations. This
model works based on the “Tobler’s First Law of Geography”, which mentioned that everything is
usually related with everything else, but those which are near to each other are more related when
compared to those that are further away. In GWR model, spatial non-stationarity was assumed and
tested. This model addresses the non-stationarity and allows the local spatial variations to vary over
space. The result of this analysis is a regression model whose parameter values apply only to each
location of observation, and different from other locations. In GWR, we use the weighted matrix
element W(i) which the amount depends on the proximity between locations. The weighting
function to be used for GWR model in this study is Gaussian Kernel function. The GWR is an
expansion of the global regression model. However, unlike global regression that is applied in general
at every observation location, GWR produces local model parameter estimators for each observation
location using the Weighted Least Square (WLS) method. The formulas and variables used in the
model are as follows (Fotheringham et al., 2002):

where: Yj = Dependent variable for observation j; Xij = Independent variable Xi at location j; uj,vj =
Coordinate point for location of observation j; C0 (uj,vj) = Intercept for observation j; Ci (uj,vj) =
Regression coefficient or local parameter estimate for independent variable Xi at location j. The
optimal bandwidth of GWR analysis in this study was determined by minimizing the corrected Akaike
Information Criterion (AIC’s value) with a correction for finite sample sizes, as described in
Fotheringham, Brunsdon, and Chalton. The dependent variable (Y) used in this model is
Environmental Quality Index (EQI), this index is designed to provide an overview of the extent to
which the living environment of a region in good or bad condition. EQI usually affecting various
indicators such as air quality, water quality, land quality, and other factors that influence
environmental quality.Eleven (11) independent variables (X) included in this model are listed in Table
1.

Code Variables
X1 Percentage of paddy field area to the total land area (%)
X2 Percentage of built-up area to the total land area (%)
X3 Percentage of forest area to the total land area (%)
X4 Percentage of bareland to the total land area (%)
X5 Percentage of agricultural area to the total land area (%)
X6 Population density (person/km2)
X7 Number of flood events in the last 10 years
X8 Number of forest fire incidents in the last 10 years
X9 Human development index
X10 Water pollution
X11 Air pollution

Results and Discussion

Identifying Driving Factors which Affecting Environmental Development Index in Java Island

We employed 11 variables as driving factors of environmental quality index (EQI) in Java Island.
Figure in below illustrates the comparison between the Y variables of the observed data shown on
the left-hand map with the variable Y of the model predicted results shown on the right-hand map.
From both maps it can be seen that the spatial distribution pattern of environmental quality index of
both observation and prediction of the model are almost the same. From the pattern of distribution
can be seen that most areas have a low to medium environmental quality index value.

Then, we also analysed another important analysis results from GWR model that is the value of local
R2. The local R2 of the GWR model was found to range from 0.45 to 0.68. The highest local R2 value
is located in the eastern part of the study area which is covered by red colour area especially located
in East Java. Those area with a higher local R2 value areas are more likely to increase EQI values
compared with the other area with smaller value of local R2. The further the location from east part,
the value of local R2 tends to be lower.

Figure a. The local R2 values based on GWR analysis

The GWR analysis shows that the coefficient value or parameter estimate (C) for each X variable is
used in the model. The spatial distribution map of the estimated parameters for each independent
variable of the GWR model is shown in each figure. Based on the analysis results, the variables that
significantly positively affect the environmental quality index in most areas are the Percentage of
forest area to the total land area (C3 X3) and the Percentage of agricultural area to the total land
area (C5 X5). The variables that have a significant adverse effect on the environmental quality index
in most areas are the Percentage of paddy field area to the total land area (C1 X1) Percentage of
built-up area to the total land area (C2 X2) Percentage of forest area to the total land area (C4 X4)
Percentage of bare land to the total land area (C6 X6) Number of flood events in the last ten years
(C7 X7) Number of forest fire incidents in the last ten years (C8 X8) Human development index (C9
X9) and Water pollution (C10 X10), also air pollution (C11 X11).
A positive effect means that the variable has a positive effect or is directly proportional to the
environmental quality index. Therefore, it is understandable that increasing the forest and
agricultural land percentage will increase the environmental quality index. The dark green areas in
Figure D and Figure F (C3 X3) and (C5 X5). Each variable shows a different parameter estimation
pattern. Thus, EQI will quickly increase on the island of Java if the percentage of forest area to total
land area (C3 X3) and agricultural area to total land area (C5 X5) increases. Meanwhile, in the
eastern part of Java Island, the EQI will quickly increase if there is an increasing percentage of forest
area in the total land area. The eastern part of Java Island is a nature reserve and national park area
with excellent potential to continue to be protected so that it can impact increasing EQI.
A negative effect means the variable has a negative effect or is inversely proportional to the
environmental quality index. Therefore, it is understandable that the Percentage of paddy field area
to the total land area (C1 X1) Percentage of built-up area to the total land area (C2 X2) Percentage of
forest area to the total land area (C4 X4) Percentage of bare land to the total land area (C6 X6)
Number of flood events in the last ten years (C7 X7) Number of forest fire incidents in the last ten
years (C8 X8) Human development index (C9 X9) and Water pollution (C10 X10), also air pollution
(C11 X11) to EQI.