Land Use and Land Cover (LULC) Change in Eastern Areas of East Java From 1972 To 2021: Learning From Landsat Image

ABSTRACT


INTRODUCTION
Changes in land use and cover (LULC) are becoming a primary global concern in our environmental issues.The accomplishment of the agenda items for the Millennium Development Goals (MDGs) may be hampered by these complex problems.At the local the area of ponds, plants, and towns (Wahyuni et al., 2021).This change in land use significantly impacts the area's plant cover and airflow processes, both directly and indirectly (Priyadarshini et al., 2019;Astuti et al., 2019;Sujarwo et al., 2021).For the sake of future planning and environmental preservation, these changes must be investigated.

Study Site and Input Data
The study was conducted in the Eastern part of East Java Province (Figure 1), covering an area of 47,075.35km 2 .East Java is ranked number one as Indonesia's province with the highest population growth and urban development.The Primary input data for the research were Landsat 1 MSS, Landsat 5 ETM, and Landsat-8 OLI/TIRS images, which were chosen based on the amount of cloud cover present.Free cloud cover images were downloaded from the Google Earth Engine (GEE) Platform.Table 1 shows the metadata related to the raw images used in the study.
The Landsat-8 imagery was captured dated from 1972 to 2021.In reality, acquiring Landsat imagery of this area with small cloud cover (>5%) is difficult.Clouds frequently taint (cover) the optical picture data acquired in tropical regions like East Java (Tseng et al., 2008).In cloudy regions, the actual ground information is obscured by clouds, limiting the applicability of optical images (Pouliot & Latifotic, 2018).Furthermore, for multi-temporal remote sensing applications (such as LUCL), cloudcovered data will produce irregular time intervals, increasing difficulty in further time series analysis (Wu et al., 2018).Minimizing data loss in optical satellite imagery due to cloud cover will impact data availability and multi-time analysis (Zhou et al., 2022).Using the GEE could anticipate this obstacle.Finally, the best quality images obtained by GEE are in 1972GEE are in , 1997GEE are in , 2013GEE are in -2014GEE are in , and 2020GEE are in -2021.

Procedure
This research procedure is divided into two stages: image treatment and LULC classification.The image treatment included post-processing, generating a land cover map, clipping with a polygon border, constructing a training area and supervised classification, atmospheric correction, pan-sharpening, composite, clip, and comparing.MultiSpec is open-source software used for image processing jobs (Landgrebe & Biehl, 2018).There are two pathways in the picture treatment process (Figure 2).

Figure 2. Flowchart of This Study
Cloud-free Landsat images were downloaded using the Google Earth Engine (GEE) platform.This platform is used because the mosaic image is smoother (Figure 3a.) compared to manual mosaics using the QGIS application (Figure 3b).Post-processing was used in this study to improve the classification findings' ability to reflect the conditions in the field to lower classification errors, lessen the impact of salt and paper, and increase accuracy.Most of the QGIS software's filter and sieve capabilities are used for this step.The majority filter is used to reduce the "salt and paper effect" phenomenon, in which isolated pixels appear in the bulk of pixels.In parallel, raster polygons lower than the predetermined threshold (in pixels) are removed using the sieve and replaced with pixel values from the closest neighboring polygon (QGIS Development Team, 2019).
The accuracy test was conducted to obtain information on the classification results' accuracy level.The method used is to use an error matrix (confusion matrix) to obtain the overall and Kappa accuracy values.Overall accuracy is the percentage of the number of pixels classified correctly (located on the matrix diagonal) divided by the total number of pixels.Kappa accuracy is a measure of the difference between the number of pixels classified correctly (diagonal matrix) and the number of pixels expected to be classified completely correctly from matching the map classification results and the reference map.Field survey data obtained is used as a reference in this accuracy test.
The algorithm used is the maximum likelihood for the image classification process.Two hundred training areas are used as input for the guided classification sample.The subset area is used to determine the changes in LULC and discuss the importance of the changes.Subset A shows changes in LULC in three regencies and one city, which includes Ngawi Regency, Magetan Regency, Madiun Regency, and Madiun City.Subset B shows big cities in East Java Province: Surabaya City, Mojokerto City, Mojokerto Regency, and Sidoarjo Regency.Cities and regencies in Subset B are the areas with the largest economic, trade, and industrial activities in East Java Province.At the same time, Subset C shows the Jember region.The selection of subset areas is based on the focus of regional development (Regional Planning Agency, 2018).
The atmospheric correction was processed using the Pan-sharpening technique and DOS (dark object subtraction) in the Semi-automatic classification plugin (SCP) (Congedo, 2016), which is accessible in QGIS (QGIS Development Team, 2019).A composite image was created using six Landsat-8 bands: bands 2, 3, 4, 5, 6, and 7. Three bands were then used to visualize the pictures (6, 5, and 2).The national standard, SNI 7645:2014, was followed in creating the number of LULC courses (BSN, 2014).The categorization procedure was carried out using Multispec's conventional image processing (Landgrebe, 2015).Here, we categorize pixels using maximum likelihood techniques.Ninety ( 90) training areas helped with the processing of supervised categorization.

Classification result
LULC types in this study were divided into eight (8) classes, namely (1) PUA, (2) HAL, (3) BS, (4) PF, (5) OWB, (6) VG, (7) SL, and (8) WL.Each type of LULC represents a different shape of the earth's surface (Table 3).Figure 4 illustrates a comparison of satellite images with actual land conditions.3. It shows each land cover class's individual, overall, and kappa accuracy.The smallest kappa accuracy value was obtained in 1972 at 76.41%.The kappa accuracy value is included in the substantial agreement category, where the value of the kappa accuracy suitability category for the substantial agreement category ranges from 61% to 80% (Viera & Garrett, 2005).The kappa accuracy value from the classification results in 2013 was 85.20%.In 2021 it was 87.31%, and the highest in 1997 was 93.65%.Table 3 also shows that the accuracy values for each class (Kappa and overall accuracy) have met the minimum USDG threshold with a value above 75% (Foody, 2004;2008).

LULC Changes in the Overall Area
The Landsat can identify and separate the LULC features into eight (8) classes, i.e., (1) PUA, (2) HAL, (3) BS, (4) PF, (5) OWB, (6) VG, (7) SL, and (8) WL. Figure 5 illustrates LULC that has occurred over the last 50 years in the study area.Table 4 shows the LULC change from Landsat imagery from 1972 to 2021.Residential areas showed an increase; in 1972, the residential area was 2.05%, and in 1997, it was 4.15%.In 2013, it was 7.91%, and in 2021, it was 9.65%.The increase in the residential area is caused by the increasing population of the East Java Province, so the need for housing is also increasing.In addition, the construction of toll roads, new road networks, and changes in paddy land cover to residential areas have occurred almost completely in East Java Province.The need for land for housing and urban service areas has grown along with the population.As a result, land resources formerly used for diverse agriculture and paddy fields have been transformed to meet demand.Agricultural regions are being transformed throughout the region into paved or urban areas.The OWB class has increased from 1972 to 2021.This increase is due to the construction many new dams in East Java Province.The construction of reservoirs in East Java is more devoted to irrigation and raw water.

. LULC change in the Eastern part of East Java
There is a significant difference between the classification data and actual land use data from BPS in 1972(BPS, 1972).For example, according to BPS, HAL's land area was 21,972 km 2 , while the classification results showed data of 16,132 km 2 .Misclassification occurred in the 1972 Landsat image due to the limited spectral information in the Landsat 1 image.Landsat 1 satellite imagery has a spatial resolution of 80 meters (Tsuchiya & Oguro, 2007), which makes the visualization of the earth's surface features less clear.In addition, the number of bands that were composited for classification only amounted to 4 bands, causing a lack of information to visualize the appearance of East Java Province that year.As a result, we may see more mixed zones with several class regions in Landsat.The study's LULC modifications also show how built-up regions have developed and become necessary for the services provided to urban residents as a result of population growth.Due to the growing population, there must be more builtup areas for homes, public spaces, and city services.

LULC Change in Subset A (Pasuruan)
Subset A is an area of 3239.62 km 2 located in Pasuruan Regency. Figure 6 and Table 5 show land cover changes from 1972, 1997, 2013, and 2021.In general, it can be seen that there have been significant changes in land use in Pasuruan from 1972 to 2021.This is related to the policy of the East Java Provincial Government regarding the development of industrial estates in several districts such as Sidoarjo, Surabaya, Jombang, and Pasuruan in 1989 (Amanda et al., 2021).The industrial area development policy will certainly affect changes in land use in Pasuruan.This can be seen in Table 5, where the urban area (PUA) increased significantly from 60.77 km 2 (1.88%) to 408 km2 (12.60%).The largest industrial area in Pasuruan Regency is the Rembang Industrial Estate (PIER) in Pasuruan, the largest area in East Java with an initial area of 254 hectares (Agustini & Winarni, 2014) industrial estates is also increasing every year; currently, the PIER area has an area of 510 hectares.In the PIER, eighty-five (80) industries are divided into three categories: food and beverage, manufacturing, and chemicals (Nainggolan et al., 2021).There was an increase in built-up and industry in Pasuruan in 2002-2018.It is projected that the built-up land will increase by 130 ha (13.62%), and industry will increase by 68 ha (20.7%) in 2026 (Prayitno et al., 2020).

LULC Change in Subset B (Banyuwangi)
Subset-B covers an area of 2036.57km 2 in Banyuwangi Regency.Banyuwangi Regency is located in the eastern region of East Java province.Figure 7 and Table 6 show the results of land cover change classification from 1972 to 2021.From 1997 to 2013, population growth experienced a very significant increase.Furthermore, Banyuwangi Regency is a developing area for Super Priority Tourism Destinations in Indonesia.In the last five years, Banyuwangi district has aggressively developed the tourism sector (Alfiyan et al., 2023).Therefore, changes in the natural landscape are very visible, especially in coastal areas.Economic activity and trade are concentrated in this city.There was an increase in the wetland area from 156 ha to 214 ha (Table 6).It can be explained that the conversion of land and forest production plantations causes the growth of wetlands in Banyuwangi Regency.There is a tendency for changes in built-up land of 1,193 ha/year trade and service areas of 28.4 ha/year in Banyuwangi Regency until 2036 (Firmansyah et al., 2018).

LULC Change in Subset C
Subset C is Jember Regency, which covers an area of 3309.60 km 2 .Figure 8 and Table 7 show the development of land cover change from 1972 to 2021.In the Jember Regency, many vegetative areas have been turned into agricultural land (BPS, 2021).It can be seen that the land cover of the pavement area (PUA) and paddy field (PF) classes is increasing.A significant decrease occurred in the shrubland (SL) class.
Table 7 shows the LULC that has occurred in the Jember Regency.In 1972, the land cover in the form of shrubland was 524.07 km 2, which then dropped drastically to 43.37 km 2 .Many areas of scrubland have been converted into rice fields.As shown in Table 7, the paddy field cover (PF) in 1972 was 817.67 km 2 and then continued to increase until 2021 with an area of 1153.06 km 2 .8).The increase in paddy field area can be caused by efforts to improve agricultural management, such as improvement of facilities and infrastructure for support the agricultural sector, starting from the aid of superior seeds to farmers, farmer cooperation for carrying out pest control many construction and requirement for irrigation facilities such as waterworks, water gates, and designated water channel to sufficiently irrigate paddy fields as well as improving management institutional.Furthermore, (Sunartomo, 2015) shows that in the coming period (2014)(2015)(2016)(2017)(2018)(2019)(2020)(2021)(2022)(2023)(2024)(2025)(2026)(2027)(2028), the rice cultivation area shows an average growth rate of 6.83% per year.

CONCLUSION
Applying a series of Landsat images from 1972 to 2021 can significantly demonstrate the change in land use and land cover (LULC) in the eastern part of East Java.The development of the region from 1972 to 2021 propagated and changed the LULC, which tended to increase significantly: the pavement and urban area (PUA), paddy field (PF), vegetation (VG ), open water body (OWB), and wetland (WL).Those classes have increased more than 100% during the last fifty years.The PUA increased from 2,05% in 1972 to 9,65% in 2021 (about 370% increase) in the total areas, while during the same period, PF increased from 16,1% to 35,8% (about 122% increase ) in the total areas.Consequently, the occupation of space for the heterogeneous agricultural land (HAL), bare soil (BS), and shrubland (SL) is significantly decreased to balance this LLC change.This also means that more and more HAL, BS, and SL are converted to PUA, PF, VG, OWB, and WL.

Figure 4 .
Figure 4. Reference photo and image visualization of training class

Figure 6 .
Figure 6.LULC changes in a large agglomeration

Figure 7 .
Figure 7. LULC changes in a large accumulation.

Table 2 .
LULC types It covers all surface features such as grass, mixed grass, dry areas with less vegetation, and abandoned agricultural land.8WetlandWLVisualizingwet areas refers to areas dominated by water and vegetation.Wetlands are usually present along the sea borders.This feature dominated the northern part of the main island in East Java.

Table 6 .
LULC change in subset B (Banyuwangi) As a result of this rapid population growth rate, the area of agricultural land has decreased, and the build-up area (PUA) has increased.Banyuwangi is an area that has the highest population density in East Java Province (BPS Jawa Timur, 2022).

Table 8 .
Rate of Development of Rice Commodity in Jember Regency 2005 ± 2013 Based on BPS data for 2006-2014, paddy fields in Jember Regency increased by 1.82% per year (Table