Prediction of the future landuse and land cover changes in the Parangtritis sand dune : a spatio temporal analysis using QGIS MOLUSCE

This study was aimed to analyze the changes and predict the future land use in the Parangtritis Sand Dunes, located in Bantul Regency, Yogyakarta Special Region, Indonesia. Parangtritis Sand Dunes face the threat of damage due to various human factors such as settlements, tourism, agriculture, plantations, etc. The Cellular Automata method, coupled with the Modules for Land Use Change Evaluation (MOLUSCE) plug-in, was employed to assess the dynamics of land cover/land use and make projections for the future. The analysis revealed significant changes in several land cover classes between 2019 and 2021, while others remained relatively stable. From 2019 to 2021, the change in area amounted to 5.6 hectares, while from 2021 to 2031 the change was only slight, amounting to 0.7 hectares. So it can be concluded that the change in the area of sand dunes in the future, namely in 2031, has not changed much, only increased by 0.7 Ha. So that more effort is needed to curb the core zone intended for the preservation of the Parangtritis sand dunes. Further investigation into the underlying causes of these shifts could provide valuable insights for effective land management and conservation efforts.


Introduction
Coasts play a vital role as transition zones between terrestrial and marine ecosystems.Globally, coastal systems stand among the most valuable and vulnerable ecosystems [1].One of Indonesia's most valuable and vulnerable coastal areas, however the sand dunes are threat by human-induced degradation, is the Parangtritis sand dunes in Bantul Regency, Yogyakarta.The sand dunes have a main function as a conservation area, a natural tsunami barrier, an air catchment area, and a habitat for sand dune flora and fauna.However, the existence of sand dunes is currently endangered due to their reduced area, which is caused by changes in land use.Every year, land use in the Parangtritis sand dunes changes, which eventually causes the area of the sand dunes to decrease every year [2].Research on the Parangtritis sand dunes area has been conducted by Dartoyo (2013) through aerial photography.His research was in the form of a comparative analysis of the area of the sand dune area in 1974 and 2013, which found that in 1974 the area of the sand dune area was 456 hectares, but until 2013 the sand dune area that was not covered by vegetation or buildings was 41 hectares [3].The decrease in the area of the sand dunes shows that the management of this area has not been maximized.This area seems to be less important because it does not provide many benefits that can be taken, whereas in fact this area is very important because it is a valuable asset and the only one in Yogyakarta [4].1313 (2024) 012014 IOP Publishing doi:10.1088/1755-1315/1313/1/012014 2 Indications of damage to the Parangtritis sand dune area begin with the emergence of an imbalance in natural processes.The sand dune area is formed by three natural processes that work, namely fluvial processes, marine processes, and aeolin processes.Until now, these processes have been disrupted along with engineering and human intervention.Sand mining carried out by the community in the river and the policy of check dam construction in the upstream indirectly have an impact on the reduction of sediment material in the form of sand in the downstream [5].The development of tourism activities around it can have a big impact on the process of sand transportation.The deflation process was greater in daytime observations than at night, averaging 2.42 gm-1s -1 during the day and 0.03 gm-1s -1 at night.Each sample location had different deflation characteristics for the material being transported.The grain diameter ranged from 0.318 mm to 0.395 mm with the dominance of medium sand texture.The roundness and sphericity of the sediment material are on the scale of 0.5 and 0.7 [6].

Figure 1. Research Location Parangtritis Sand Dune
Several anthropogenic activities were identified in the study area such as settlement, agriculture, ponds, and coastal forest cultivation.Among these anthropogenic activities, beach forest cultivation is the most threatening activity to the aeolian process in Parangtritis sand dunes.Besides vegetation density, sand transportation is also the main factor controlling the formation of sand dunes [7].Vegetation and buildings also play a role as a barrier to the deflation process, vegetation is abundant in the western part or in the support zone while buildings are abundant in the eastern part, namely in the limited use zone.This is supported by population data, it is known that the population in Parangtritis Village is 7,894 people [8].Based on the background and problems described, it is necessary to conduct research on the impact of dynamic land use change in Parangtritis Sand Dunes, Bantul Regency, Yogyakarta Special Region Province.The impact of the dynamics of land use change disrupts the aeolin process of the sand dunes, in the form of vegetation and buildings that become obstacles to the movement of sand by the wind.The purpose of this research is to analyze the dynamics of land cover and predict the land cover of Parangtritis Sand Dunes using the Cellular Automata method in the MOLUSCE plug-in.The types of data, measurement methods, and sources used in this research were supporting tools for data processing for the dynamics of land cover change and prediction using Quantum GIS (QGIS) Software with the Modules for Land Use Change Simulation (MOLUSCE) Plug-In.The data and variables are presented in Table 1 below The data taken to analyze the dynamics of land change and land cover prediction of Parangtritis Sand Dunes include land cover data for three periods: sand dune land cover in 2019, 2020, and 2020.These data were generated from aerial photographs that have been processed by digitizing on-screen and classified according to the Standard Nasional Indonesia (SNI) Land Cover Classification Scale 1:25,000.Land cover data were obtained from the Parangtritis Geomaritime Science Park (PGSP) agency under the Badan Informasi Geospasial (BIG).This land cover data was obtained in March 2023 through a research data request letter to BIG and was forwarded to PGSP to submit the data.Additionally, Parangtritis Sand Dune boundary data was also obtained.
This land cover data was then processed using QGIS 3.16.16software to obtain driving factors such as distance from roads, settlements, buildings, tourism, forests, and agriculture.Then, all the data was standardized in terms of data format.The format included the same projection system, namely WGS 84 / UTM Zone 49 S, the same length and width (width of 6166m and length of 2304m), the same data type (Float32), the same pixel size of 1m x 1m, and the data no-must also be the same (-9999).After all the data is formatted uniformly, it is ready for processing in QGIS 2.18.18Software with the MOLUSCE plug-in.

Data Analysis Technique
Land cover projections were conducted using a cellular automata (CA) model with an Artificial Neural Network (ANN) architecture.Land cover in 2031 was obtained by comparing changes between 2019 and 2020.The model simulation was run with the cellular automata simulation model.Land cover change is based on land suitability, previous land cover, and neighboring land cover [9].The pattern of land cover change that occurred between 2019 and 2020 became a variable to predict land cover in 2031.Projections were made by assuming that future changes will have similar patterns and opportunities as the patterns of change that occurred during the time period used.Land cover in 2021 was obtained by comparing changes between 2019 and 2020.
The results of the analysis were inputs for determining land cover in the projection year.The transition matrix, which was a transition matrix of changes from the previous year to the projection year, was used to determine the chances of changes in each land cover.Next, learning was performed on each information (input) obtained from land cover changes by running an artificial neural network system.The artificial neural network learns patterns from the given inputs (land cover in 2019 and 2020) and then generates solutions (possibilities) of changes that will occur.The next step was to run the cellular automata (CA) model to obtain land cover projections for 2021.The results of the 2021 land cover prediction that have been tested for accuracy were then overlaid with the actual 2021 land cover.After that, the change in the land cover area was observed.The modeling and projection stages of land cover in this study were carried out using the MOLUSCE plugin in the Quantum GIS (QGIS) software.The following are the steps taken for projection in the QGIS software with the MOLUSCE plugin in Figure 2. The interpretation was carried out on aerial photographs in 2019 and 2021 at Parangtritis Sand Dunes.Each land cover class that has been interpreted changes over time.These changes are caused by natural and human factors, such as human activities that can alter land functions.The changes that occurred between 2019 and 2020 are listed in Table 2.  1 shows that each land cover class in Parangtritis Sand Dunes experienced significant changes in several land cover classes, while the rest experienced little or no change.The most significant changes occurred in land cover that increased in the form of sand dunes by 26.6%, footpaths by 70.37%, and road borders by 22.32%.Meanwhile, the land cover that decreased significantly included livestock by 21.56% and irrigation canals by 23.86%.The land cover of shrubs and open land transformed into sand dunes, footpaths, and road borders.The map of land cover in 2019 and 2021 is depiced in Figure 3 and 4.  Land cover in 2019 and 2020 was used as data to predict land cover in 2031.The prediction involves a simple rule of land cover change by comparing two preceding land covers to determine the next period of land cover, which will be applied to cellular automata modelling [11].The accuracy of the results of this land cover prediction modeling is 92% so it can be said that the modeling results are very good, then for a kappa value of 0.98 which means it shows a very good agreement.The results of the interpretation of land cover in 2019 and 2031 are presented in Table 3.

Conclusion
Based on the results of data analysis, it is evident that each class of land cover in Parangtritis Sand Dunes in 2019 and 2021 experienced significant changes in several classes of land cover, while the rest experienced little or no change.The projections for land cover in 2031 tend to indicate that all land cover has decreased in area, except for sand dunes, footpaths, airstrips, markets, offices, agriculture, shrubs, and roadsides.From 2019 to 2021, the change in area amounted to 5.6 hectares, while from 2021 to 2031 the change was only slight, amounting to 0.7 hectares.So it can be concluded that the change in the area of sand dunes in the future, namely in 2031, has not changed much, only increased by 0.7 Ha.So that more effort is needed to curb the core zone intended for the preservation of the Parangtritis sand dunes.Especially from vegetation and buildings, so that the aeolin process can run well and the sand dunes can be formed properly.so that the barchan type sand dunes will be formed again which is a unique landscape in the parangtritis sand dunes.

Figure 2 . 5 3. Result and Discussion 3 . 1
Figure 2. Steps in executing the molusce process Explanation: 1. Input: The Input stage involved entering data into the MOLUSCE plugin.For the Initial column, initial land cover data (2019) was entered, and for the Final column, final land cover data (2020) was entered to then project land cover in 2021.Meanwhile, for the Spatial variables column, data on driving factors such as distance from roads, rivers, settlements, and population density were entered.A geometry check was then performed to ensure that all data entered has the same geometry to proceed to the next stage.All data entered was in raster format because the system will analyze each pixel in the data.2. Evaluating Correlation: Evaluating Correlation measures the relationship between a variable and each land cover with a value range of 0 to 1, where 0 indicates no relationship, while 1 indicates a strong relationship between the variable and the land cover.3. Transition Potential Modeling: At this stage, possibilities of changing one land cover to another will be generated.The principle works by studying changes that have occurred and repeating the pattern of change until a suitable model is reached.At this stage, several methods were provided that can be used for modeling, one of which is used in this research, namely the Artificial Neural Network (Multi-layer Perceptron) or ANN-MLP model.In this model, several network simulations were carried out by setting parameter values.The parameters used in the simulation were to find the best RMS value.The model will stop when it has reached the specified conditions.4. Cellular Automata Simulations: At this stage, the future land cover projection process was carried out.In Mollusce, the length of the prediction time (automatic) is t1+(t1-t0), where t1 is the final year (Final), and t0 is the initial year (Initial), so the predictions generated in this study were for 2019 and 2020 using iteration 1, and to predict land cover in 2021. 5. Validation: At this stage, a validation process was carried out on the land cover projected results in 2021, which were compared with the land cover interpreted (actual) results in 2021.If the results are close to each other, then the validation can be accepted for further projection, namely land cover in 2031.Validation itself is calculated in Mollusce by considering the Kappa value.Kappa value of 0.81-1.00indicates excellent agreement, 0.61-0.80 is good, 0.41-0.60 is moderate, 0.21-0.40 is less than moderate, and a value of <0.21 is considered poor [10].

Figure 4 . 7 3. 2 Future
Figure 4. 2021 Land Cover/Land Use MapIn Figures3 and 4, it can be seen that the Parangtritis sand dunes are symbolized in red to make the changes more visible.The dynamics of the sand dunes from 2019 to 2021 occurred an area change of 5.5 hectares, this occurred due to the curbing of the core area which functioned as the preservation of the parangtritis sand dunes.

Table 1 :
: Data and Research Variables

Table 2 .
Land Use and Land Use Change in 2019 and 2021

Table 3 .
Land cover area in 2019 and 2031The data in Table3are the prediction data of land cover in 2031, revealing that each land cover class in Parangtritis Sand Dunes has experienced further area reduction, except for sand dunes, footpaths, airstrips, markets, offices, agriculture, shrubs, and roadsides.The map of land cover in 2031 presented in Figure4.