Estimasi Risiko Kerugian Bencana Banjir Berbasis Frequency-Severity dengan Dependensi Spasial Menggunakan Bivariate Inseparable Conditional Autoregressive di Kabupaten Aceh Barat Daya

Putri, Yuliana Cinta Damayanti (2026) Estimasi Risiko Kerugian Bencana Banjir Berbasis Frequency-Severity dengan Dependensi Spasial Menggunakan Bivariate Inseparable Conditional Autoregressive di Kabupaten Aceh Barat Daya. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5006221022_Undergraduate-Thesis.pdf - Accepted Version Restricted to Repository staff only Download (5MB) | Request a copy

Abstract

Banjir merupakan salah satu bencana hidrometeorologi yang dapat menimbulkan kerugian ekonomi dengan tingkat risiko berbeda antarwilayah. Penelitian ini bertujuan untuk memodelkan risiko kerugian banjir pada tingkat desa di Kabupaten Aceh Barat Daya dengan mempertimbangkan frekuensi kejadian, potensi kerugian (severity), dan dependensi spasial antarwilayah. Data yang digunakan merupakan data agregat pada 152 desa selama periode 2019–2023. Variabel severity dalam penelitian ini merepresentasikan potensi kerugian berdasarkan Kajian Risiko Bencana (KRB), bukan kerugian aktual per kejadian banjir. Pemodelan dilakukan menggunakan pendekatan frequency-severity dalam kerangka Hierarchical Bayesian. Komponen frekuensi dimodelkan menggunakan kandidat distribusi Poisson dan negative binomial, sedangkan komponen severity dimodelkan menggunakan kandidat distribusi gamma, lognormal, Weibull, dan inverse gaussian. Dependensi spasial dimodelkan melalui struktur Bivariate Inseparable Conditional Autoregressive (CAR) dengan matriks bobot spasial berbasis queen contiguity. Estimasi posterior dilakukan menggunakan metode Markov Chain Monte Carlo (MCMC), sedangkan pemilihan model terbaik menggunakan Deviance Information Criterion (DIC). Hasil pengujian Moran’s I menunjukkan adanya autokorelasi spasial positif dan signifikan pada komponen frekuensi maupun severity. Berdasarkan nilai DIC terkecil sebesar 3.485,05, model terbaik adalah kombinasi negative binomial–Weibull. Pada komponen frekuensi, curah hujan menunjukkan pengaruh positif yang cukup kuat secara posterior. Pada komponen severity, luas bahaya dan jumlah penduduk terpapar menunjukkan pengaruh positif yang cukup kuat. Korelasi spasial antara frekuensi dan severity memiliki kecenderungan positif, tetapi belum cukup kuat secara posterior karena interval kredibel 95% masih mencakup nol. Risiko kerugian diestimasi menggunakan Expected Loss (EL), varians kerugian, Value at Risk (VaR) 95%, dan Tail Value at Risk (TVaR) 95%. Desa Padang Baru dan Geulima Jaya di Kecamatan Susoh secara konsisten memiliki risiko tertinggi.
======================================================================================================================================
Flooding is one of the hydrometeorological disasters that can cause economic losses with varying levels of risk across regions. This study aims to model flood loss risk at the village level in Southwest Aceh Regency by considering the frequency of flood events, potential losses (severity), and spatial dependence between regions. The data consist of aggregated observations from 152 villages during the 2019–2023 period. In this study, severity represents potential losses based on the Disaster Risk Assessment (Kajian Risiko Bencana or KRB), rather than actual losses for each flood event. The modeling was conducted using a frequency-severity approach within a Hierarchical Bayesian framework. The frequency component was modeled using Poisson and Negative Binomial distributions, while the severity component was modeled using Gamma, Lognormal, Weibull, and Inverse Gaussian distributions. Spatial dependence was modeled using a Bivariate Inseparable Conditional Autoregressive (CAR) structure with a queen contiguity-based spatial weight matrix. Posterior estimation was performed using the Markov Chain Monte Carlo (MCMC) method, while the best model was selected using the Deviance Information Criterion (DIC). The Moran’s I test results indicate significant positive spatial autocorrelation in both the frequency and severity components. Based on the smallest DIC value of 3,485.05, the best model is the Negative Binomial–Weibull combination. In the frequency component, rainfall shows a sufficiently strong positive posterior effect. In the severity component, hazard area and the number of exposed residents show sufficiently strong positive posterior effects. The spatial correlation between frequency and severity tends to be positive, but the posterior evidence is not sufficiently strong because the 95% credible interval still includes zero. Flood loss risk was estimated using Expected Loss (EL), loss variance, Value at Risk (VaR) at the 95% confidence level, and Tail Value at Risk (TVaR) at the 95%
confidence level. Padang Baru and Geulima Jaya Villages in Susoh District consistently have the highest risk levels.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Banjir, Bivariate Inseparable Conditional Autoregressive (CAR), Frequency-Severity, Hierarchical Bayesian, Risiko Kerugian, Flood, Loss Risk
Subjects:	H Social Sciences > HD Industries. Land use. Labor > HD61 Risk Management
Divisions:	Faculty of Science and Data Analytics (SCIENTICS) > Actuaria > 94203-(S1) Undergraduate Thesis
Depositing User:	Yuliana Cinta Damayanti Putri
Date Deposited:	18 Jul 2026 08:15
Last Modified:	18 Jul 2026 08:15
URI:	http://repository.its.ac.id/id/eprint/135411

Actions (login required)

View Item