Enhancing Land Cover/Land Use (LCLU) classification through a comparative analysis of hyperparameters optimization approaches for deep neural network (DNN) DOI Creative Commons
Ali Azedou,

Aouatif Amine,

Isaya Kisekka

и другие.

Ecological Informatics, Год журнала: 2023, Номер 78, С. 102333 - 102333

Опубликована: Окт. 11, 2023

Sustainable natural resources management relies on effective and timely assessment of conservation land practices. Using satellite imagery for Earth observation has become essential monitoring cover/land use (LCLU) changes identifying critical areas conserving biodiversity. Remote Sensing (RS) datasets are often quite large require tremendous computing power to process. The emergence cloud-based techniques presents a powerful avenue overcome limitations by allowing machine-learning algorithms process analyze RS the cloud. Our study aimed classify LCLU Talassemtane National Park (TNP) using Deep Neural Network (DNN) model incorporating five spectral indices differentiate six classes Sentinel-2 imagery. Optimization DNN was conducted comparative analysis three optimization algorithms: Random Search, Hyperband, Bayesian optimization. Results indicated that improved classification between with similar reflectance. Hyperband method had best performance, improving accuracy 12.5% achieving an overall 94.5% kappa coefficient 93.4%. dropout regularization prevented overfitting mitigated over-activation hidden nodes. initial results show machine learning (ML) applications can be tools management.

Язык: Английский

Background climate modulates the impact of land cover on urban surface temperature DOI Creative Commons

Marzie Naserikia,

Melissa Hart, Negin Nazarian

и другие.

Scientific Reports, Год журнала: 2022, Номер 12(1)

Опубликована: Сен. 14, 2022

Abstract Cities with different background climates experience thermal environments. Many studies have investigated land cover effects on surface urban heat in individual cities. However, a quantitative understanding of how modify the impact covers remains elusive. Here, we characterise and their impacts temperature (LST) for 54 highly populated cities using Landsat-8 imagery. Results show that characteristics response are distinctly across various climate regimes, largest difference arid climates. Cold seasonal variability, least seasonality tropical In tropical, temperate, cold climates, normalised built-up index (NDBI) is strongest contributor to LST variability during warm months followed by vegetation (NDVI), while bareness (NDBaI) most important factor These findings provide climate-sensitive basis future planning oriented at mitigating local warming.

Язык: Английский

Процитировано

35
Assessment of temporal shifting of PM2.5, lockdown effect, and influences of seasonal meteorological factors over the fastest-growing megacity, Dhaka DOI Open Access
Abdullah-Al- Faisal, Abdulla ‐ Al Kafy, Md. Abdul Fattah

и другие.

Spatial Information Research, Год журнала: 2022, Номер 30(3), С. 441 - 453

Опубликована: Март 19, 2022

Язык: Английский

Процитировано

34
Predicting Microscale Land Use/Land Cover Changes Using Cellular Automata Algorithm on the Northwest Coast of Peninsular Malaysia DOI
Zullyadini A. Rahaman, Abdulla ‐ Al Kafy, Abdullah-Al- Faisal

и другие.

Earth Systems and Environment, Год журнала: 2022, Номер 6(4), С. 817 - 835

Опубликована: Июнь 16, 2022

Язык: Английский

Процитировано

32
The Extreme Heat Wave over Western North America in 2021: An Assessment by Means of Land Surface Temperature DOI Creative Commons
Gabriel I. Cotlier, Juan C. Jiménez‐Muñoz

Remote Sensing, Год журнала: 2022, Номер 14(3), С. 561 - 561

Опубликована: Янв. 25, 2022

In our current global warming climate, the growth of record-breaking heat waves (HWs) is expected to increase in its frequency and intensity. Consequently, considerably growing agglomerated world’s urban population becomes more exposed serious heat-related health risks. this context, study Surface Urban Heat Island (SUHI) intensity during HWs substantial importance due potential vulnerability urbanized areas might have comparison their surrounding rural areas. This article discusses Land Temperatures (LST) reached extreme HW over Western North America boreal summer 2021 using Thermal InfraRed (TIR) imagery acquired from TIR Sensor (TIRS) (30 m spatial resolution) onboard Landsat-8 platform Moderate Resolution Imaging Spectroradiometer (MODIS) (1 km Terra/Aqua platforms. We provide an early assessment maximum LSTs affected areas, as well impacts terms SUHI main cities towns. MODIS series LST 2000 presented highest recorded values late June 2021, with around 50 °C for some cities. High resolution (Landsat-8) were used map assess impact on thermal comfort conditions at intraurban space by means a environmental quality indicator, Field Variance Index (UFTVI). The same high verify existence clusters employ Local Indicator Spatial Association (LISA) quantify degree strength. identified distribution patterns within described behavior across landscape fitting polynomial regression model. also qualitatively analyze relationship between both UFTVI different land cover types. Findings indicate that average daytime studied was typically 1 5 °C, exceptional surpassing 7 9 °C. During night, reduced variations 1–3 value +4 no significant influence maps evidence hotspots much higher located densely built-up while green spaces dense vegetation show lower values. manner, UTFVI has shown “no” vegetated regions, water bodies, low-dense intertwined vegetation, “strongest” observed non-vegetated low albedo material such concrete pavement. evidenced good marker assessing recognizing consequences finding highlights remote-sensing based particularly suitable indicator analysis resolutions. analyzing detecting temperature events seems be promising rapid accurate monitoring mapping.

Язык: Английский

Процитировано

29
Enhancing Land Cover/Land Use (LCLU) classification through a comparative analysis of hyperparameters optimization approaches for deep neural network (DNN) DOI Creative Commons
Ali Azedou,

Aouatif Amine,

Isaya Kisekka

и другие.

Ecological Informatics, Год журнала: 2023, Номер 78, С. 102333 - 102333

Опубликована: Окт. 11, 2023

Sustainable natural resources management relies on effective and timely assessment of conservation land practices. Using satellite imagery for Earth observation has become essential monitoring cover/land use (LCLU) changes identifying critical areas conserving biodiversity. Remote Sensing (RS) datasets are often quite large require tremendous computing power to process. The emergence cloud-based techniques presents a powerful avenue overcome limitations by allowing machine-learning algorithms process analyze RS the cloud. Our study aimed classify LCLU Talassemtane National Park (TNP) using Deep Neural Network (DNN) model incorporating five spectral indices differentiate six classes Sentinel-2 imagery. Optimization DNN was conducted comparative analysis three optimization algorithms: Random Search, Hyperband, Bayesian optimization. Results indicated that improved classification between with similar reflectance. Hyperband method had best performance, improving accuracy 12.5% achieving an overall 94.5% kappa coefficient 93.4%. dropout regularization prevented overfitting mitigated over-activation hidden nodes. initial results show machine learning (ML) applications can be tools management.

Язык: Английский

Процитировано

19