Can biophysical parameters derived from Sentinel-2 space-borne sensor improve land cover characterisation in semi-arid regions?

被引：11

作者：

Mudereri, Bester Tawona ^{[1
,2
]}

Chitata, Tavengwa ^{[3
]}

Mukanga, Concilia ^{[1
]}

Mupfiga, Elvis Tawanda ^{[3
]}

Gwatirisa, Calisto ^{[4
]}

Dube, Timothy ^{[5
]}

机构：

[1] Midlands State Univ, Dept Anim & Wildlife Sci, Gweru, Zimbabwe

[2] ICIPE, Nairobi, Kenya

[3] Midlands State Univ, Dept Land & Water Management, Gweru, Zimbabwe

[4] Makoholi Expt Stn, Dept Res & Specialist Serv, Masvingo, Zimbabwe

[5] Univ Western Cape, Dept Earth Sci, Bellville, South Africa

来源：

GEOCARTO INTERNATIONAL | 2021年 / 36卷 / 19期

关键词：

Bayesian; FAPAR; LAI; naive Bayes; random forest; SNAP (R); rural Zimbabwe; RANDOM FOREST; SPECTRAL RESOLUTION; GROUND BIOMASS; NAIVE BAYES; CLASSIFICATION; REGRESSION; UTILITY; INDEX;

D O I：

10.1080/10106049.2019.1695956

中图分类号：

X [环境科学、安全科学];

学科分类号：

08 ; 0830 ;

摘要：

We explore the potential contribution of Sentinel-2 (S2) wavebands and biophysical parameters, i.e. Leaf Area Index (LAI), Chlorophyll content (Cab), Fraction of Absorbed Photosynthetically Active Radiation (FAPAR), Fraction of Vegetation Cover (FVC) and Canopy Water Content (CWC) in mapping land use and land cover (LULC) in Zimbabwe. Random forest (RF) and naive Bayes (NB) were used to classify S2 imagery. S2 biophysical variables resulted in LULC overall accuracy (OA) of 96% and 86% for RF and NB respectively, whereas S2 wavebands produced slightly higher accuracies of 97% and 88% for RF and NB respectively. Combining wavebands and biophysical variables enhanced classification results (OA = 98%: RF and 91%: NB). Variable importance analysis showed that FAPAR, red-edge 2, green, red-edge 3, FVC and band 8a, are the most relevant in the classification. Our work shows the strength and capability of biophysical variables in discerning different LULC attributes in semi-arid environments.

引用

页码：2204 / 2223

页数：20

共 60 条

[1] Detecting Sirex noctilio grey-attacked and lightning-struck pine trees using airborne hyperspectral data, random forest and support vector machines classifiers
Abdel-Rahman, Elfatih M.
Mutanga, Onisimo
Adam, Elhadi
Ismail, Riyad
[J]. ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING, 2014, 88 : 48 - 59
[2] Random forest regression and spectral band selection for estimating sugarcane leaf nitrogen concentration using EO-1 Hyperion hyperspectral data
Abdel-Rahman, Elfatih M.
Ahmed, Fethi B.
Ismail, Riyad
[J]. INTERNATIONAL JOURNAL OF REMOTE SENSING, 2013, 34 (02) : 712 - 728
[3] Ahmed N., 2018, GEOL ECOL LANDSCAPES, P1
[4] Baloloy A., 2018, ISPRS Annals of Photogrammetry, Remote Sensing Spatial Information Sciences, V4
[5] Barakat A., 2018, Geology Ecology and Landscapes, V2, P203, DOI DOI 10.1080/24749508.2018.1452478
[6] Sentinel-2 cropland mapping using pixel-based and object-based time-weighted dynamic time warping analysis
Belgiu, Mariana
Csillik, Ovidiu
[J]. REMOTE SENSING OF ENVIRONMENT, 2018, 204 : 509 - 523
[7] Bochenek Z., 2017, Geoinformation, V1, P15, DOI [10.34867/gi.2017.2, DOI 10.34867/GI.2017.2]
[8] Random forests
Breiman, L
[J]. MACHINE LEARNING, 2001, 45 (01) : 5 - 32
[9] A Comparative Assessment of Machine-Learning Techniques for Land Use and Land Cover Classification of the Brazilian Tropical Savanna Using ALOS-2/PALSAR-2 Polarimetric Images
Camargo, Flavio F.
Sano, Edson E.
Almeida, Claudia M.
Mura, Jose C.
Almeida, Tati
[J]. REMOTE SENSING, 2019, 11 (13)
[10] Campbell J, 2007, TLS-TIMES LIT SUPPL, P5

← 1 2 3 4 5 6 →