Page 4 - NGTU_paper_withoutVideo

Modern Geomatics Technologies and Applications

Knowledge Based Method for Land Surface Emissivity and Temperature Retrieval of the Remote
Sensing Data

Hassan Emami , Seyyed Qasem Rostami 2
1*

1 Department of Geomatics, School of Marand Engineering, University of Tabriz, Tabriz-Iran, h_emami@tabrizu.ac.ir,
h_emami@ut.ac.ir, ORCID: 0000-0002-0171-6487.

2 Department of Surveying Engineering, Faculty of Engineering, University of Bojnourd, Bojnourd-Iran gh.rostami@ub.ac.ir
* corresponding author: h_emami@ut.ac.ir

Abstract
In this work, a knowledge based approach is proposed to overcome the errors and uncertainties in land surface
emissivity (LSE) estimation and consequently land surface temperature (LST) retrieval. The Knowledge Based Methods
(KBMs) which including two LSE estimation methods. The effectiveness of KBMs proposed is empirically tested over two
scenes of Landsat-8 (known as Landsat Data Continuity Mission, LDCM) data sets and the obtained LSEs by conventional
and proposed methods were compared to the LSE product of the ASTER by image-based cross-comparison. In both
scenes, the NDVI-based emissivity method (NBEM) provide appropriate results among five conventional methods. In
contrast, Validity Average (VAvg) achieves superior results among proposed methods for both scenes. Moreover, the error
ranges and RMSE of cross-comparison for the obtained LSE in proposed methods were remarkably decreased. Also, in
this research, for LST cross-comparison, an alternative scaling method based on LST products of MODIS was
proposed .The LST validation results demonstrated that proposed methods provide better estimates in terms of three
accuracy measures in both examined datasets. Furthermore, the obtained LST of Knowledge Based LSE estimation
method, show that the proposed methods provide better estimates in both examined datasets in terms of the three
2
2
statistical R (improved 8.16%), the adjusted R (improved 5%), MD (Bias) (improved 1.03K), and RMSE (improved 0.6K)
measures rather than LST retrieval using conventional LSEs method.

Keywords: Remote sensing, Knowledge based method, Land surface emissivity, Land surface temperature, LDCM.

1. Introduction
Land surface emissivity (LSE) is an important intrinsic property of materials and knowledge of the LSE is essential to
derive the land surface temperature (LST) that can be obtained from the emitted radiance measured from space. LSE provides
useful information for geological and environmental studies, mineral mapping and is one of the important input parameters for
climate, hydrological, ecological and biological models [1, 2].Several methods exist to estimate spectral emissivity from satellite
data, which apply the visible and near-infrared (VNIR) or thermal infrared (TIR) spectral regions or both of them. According to
the way by which the LSE is determined along with LST, the emissivity estimation methods from optical remote sensing data
can be categorized into three distinct types [3].The first group is a stepwise retrieval method that determines the LSE and the
LST separately. Representative methods of this group include the NDVI-based emissivity method (NBEM) [4, 5], the
classification-based emissivity method (CBEM) [6, 7], and so forth. The second group of algorithms retrieve both LSE and LST
from at-surface radiance, based on some assumptions or constraints. The representative methods of this group consist of the
two-temperature Method [8], the physics-based day/night operational method [9],temperature/emissivity separation (TES)
method [10], and so on. The third group simultaneously retrieves the atmospheric profiles along with both LST and LSE. The
representatives of this group are the artificial neural network (ANN) method [11] and the two-step physical retrieval method [12]
which are mostly used for thermal hyperspectral imaging. The methodologies for estimating LSEs from satellite data were briefly
reviewed in[13], with their advantages and limitations. Since, various LSE estimation methods have been proposed with the
same aims but under different conditions, with different assumptions for various applications, and with their advantages and
limitations[3].These methods have different accuracies and are applicable for various sensors and applications.

1