Kajian Model Penduga Suhu Permukaan Lahan Berbasis Citra Resolusi Sedang di Kota Medan

Abstract

Kajian ini mengulas tentang pembangunan model penduga suhu permukaan lahan dengan pemanfaatan teknologi penginderaan jarak jauh. Suhu permukaan lahan diturunkan dari saluran termal citra Landsat 8 rekaman tahun 2021. Model penduga dibangun menggunakan beberapa model regresi linier, eksponensial, dan regresi linier berganda dengan peubah bebas geo-sosio biofisik dan indeks vegetasi. Penentuan model terbaik didasarkan pada skoring hasil analisis statistik simpangan rata-rata (SR), simpangan agregat (SA), bias (e), dan root mean square error (RMSE). Kajian ini menemukan bahwa tiga model penduga suhu permukaan lahan terbaik di Kota Medan diperoleh dengan bentuk persamaan regresi linier berganda LST = 30,502 + 16,60 NBBI – 1,42 NDVI sebagai peringkat pertama, serta persamaan regresi eksponensial LST = 30,3242e0,7107 NBBI dan regresi linier berganda LST = 29,081 + 0,000735 Kepadatan Pemukiman + 15,492 NBBI sebagai peringkat kedua dan ketiga. Adapun model dengan peringkat pertama memiliki SR sebesar 0,075%, SA sebesar 0,0014, bias sebesar 0,262%, dan RMSE sebesar 0,492%.

This study uses remote sensing technology to examine models for estimating land surface temperature. Land surface temperature is derived from the thermal band of Landsat 8 images recorded in 2021. Several regression models, including linear, exponential, and multiple linear regression models with geo-socio-biophysical and vegetation indices variables, were used to build the estimation model. The optimal model is determined by scoring of mean deviation (SA), aggregate deviation (SR), bias (e), and root mean square error (RMSE) values obtained from statistical analysis. This study found that the three best Land Surface Temperature (LST) estimation models in Medan City were obtained with the multiple linear regression equation LST = 30.502 + 16.60 NBBI – 1.42 NDVI as the first rank, followed by exponential regression equation LST = 30,3242e0,7107 NBBI and multiple linear regression equation LST = 29,081 + 0,000735 Settlement Density + 15,492 NBBI as the second and third rank. The model with the first rank exhibits a SR of 0,075%, SA of 0,0014, bias of 0,262%, and RMSE of 0,492%.