Land Use Change Modeling in Siak District, Riau Province, Indonesia Using Multinomial Logistic Regression

Abstract

Siak District is an enlargement from some parts of Bengkalis District that was established in 1999. The development, that has been conducted so far, has altered land uses which involve land conversion from a type of use to other uses. In this study, land use change modeling has been developed in Siak District that may facilitate the understanding of the process of land use change and its relevant factors in the research site after the enlargement process done completely. Land use change modeling in Siak District has been conducted in order to analyze the land use change during 2002 – 2005 and 2005 – 2008, to develop the land use change scheme of Siak District, to identify the driving factors of land use change and develop the land use change model of Siak District, and to examine the performance of Multinomial Logistic Regression (MLR) model in modeling the land use change. During 2002 – 2008, Siak District was dominated by Forest land, Cropland and Grassland, whereas Wetlands, Settlements, and Other lands occupied smaller area. Based on the transition probability matrices, Forest land, Cropland, Grassland, Wetland, and Settlement tend to be in stable condition, and only Other lands which changes dynamically. However, the probabilities of each land use to transform into other land uses are also quite significant. In this research, the land use change model has been developed in two scenarios: (1) using all significant variables determined by the MLR model analysis and (2) using observed variables determined by the observation of existing condition in the field. Both scenarios indicate that the final models of land use change in Siak District which were developed using MLR model were good models. The models could explain most of the variability of land use change which happen in the research site. However, the model validations have been conducted spatially for both scenarios indicate that the final model could not fit the actual spatial data layers into the actual condition of land use change 2005 – 2008. Based on the result of model validations, there are some research findings which can be concluded. Land use change model that has been developed by using MLR model is a generalized model of logistic regression. The MLR model forces every land use transitions to be driven by all significant parameters, while in the real condition each land use transition probably has unique combination of parameters which drive its land use transition.

Kabupaten Siak merupakan pemekaran dari beberapa kecamatan di Kabupaten Bengkalis pada tahun 1999. Pembangunan yang telah dilakukan selama ini, telah mengubah penggunaan-penggunaan lahan dari satu tipe lahan menjadi penggunaan lain. Pada penelitian ini, pemodelan perubahan penggunaan lahan dilakukan di Kabupaten Siak dalam rangka memahami proses perubahan penggunaan lahan yang terjadi setelah proses pemekaran kabupaten. Pemodelan perubahan penggunaan lahan di Kabupaten Siak ini bertujuan untuk menganalisis perubahan penggunaan lahan yang terjadi selama periode 2002 – 2005 dan 2005 – 2008, membangun skema perubahan penggunaan lahan, mengidentifikasi faktorfaktor yang mempengaruhi perubahan penggunaan lahan dan membangun model perubahan penggunaan lahan, dan menguji performa model Multinomial Logistic Regression (MLR) dalam memodelkan perubahan penggunaan lahan. Selama tahun 2002 – 2008, Kabupaten Siak didominasi oleh Lahan Hutan, Lahan Perkebunan, Lahan Rerumputan, sedangkan Lahan Basah, Pemukiman, dan Lahan Lain menempati lahan yang lebih sempit. Berdasarkan matrik-matrik kemungkinan transisi (transition probability matrices), Lahan Hutan, Lahan Perkebunan, Lahan Rerumputan, Lahan Basah, dan Pemukiman cenderung berada dalam kondisi stabil, dan hanya Lahan Lain yang berubah secara dinamis. Namun demikian, kemungkinan masing-masing penggunaan lahan berubah menjadi lahan-lahan lain juga cukup signifikan. Pada penelitian ini, model perubahan penggunaan lahan dibangun menggunakan dua skenario: (1) menggunakan seluruh variabel-variabel penting hasil analisis model MLR dan (2) menggunakan variabel-variabel hasil obeservasi lapangan. Kedua skenario tersebut menunjukkan bahwa model-model perubahan penggunaan lahan yang telah dibangun merupakan model yang baik. Model-model tersebut mampu menjelaskan sebagian besar variasi dari perubahan penggunaan lahan di lokasi penelitian. Akan tetapi, validasi-validasi model yang telah dilakukan secara spasial menunjukkan model-model tersebut tidak dapat mencocokkan data spasial aktual dengan kondisi perubahan penggunaan lahan yang terjadi pada tahun 2005 – 2008.