Machine Learning for Urban Heat Island (UHI) Analysis: Predicting Land Surface Temperature (LST) in Urban Environments

Urban Climate, Journal Year: 2024, Volume and Issue: 55, P. 101962 - 101962

Published: May 1, 2024

Language: Английский

---

Optuna-DFNN: An Optuna framework driven deep fuzzy neural network for predicting sintering performance in big data

Alexandria Engineering Journal, Journal Year: 2024, Volume and Issue: 97, P. 100 - 113

Published: April 16, 2024

Language: Английский

---

Machine Learning-Based Rice Field Mapping in Kulon Progo using a Fusion of Multispectral and SAR Imageries

Forum Geografi, Journal Year: 2023, Volume and Issue: 37(2)

Published: Dec. 29, 2023

The land-conversion of rice fields can reduce production and negatively impact food security. Consequently, monitoring is essential to prevent the loss productive agricultural land. This study uses a combination Sentinel-2 MSI, Sentinel-1 SAR, along with SRTM (elevation slope data) monitor land-conversion. NDVI, NDBI NDWI indices are transformed from annual median composite MSI images used identify different another object. A monthly SAR data cropping patterns in inundation phase. classification performed by using Random Forest machine learning algorithm Google Earth Engine (GEE) platform. run 1000 trees, 70:30 ratio training testing sample features extracted visual interpretation high resolution imagery. In this study, effective computing amount multi-temporal multi-sensory map rice-field land conversion an accuracy rate 96.16% (2021) 95.95% (2017) for mapping paddy fields. From multitemporal field maps 2017—2021, 826.66 hectares rice-fields non-rice was identified. Based on spatial distribution, higher at area near roads, built Yogyakarta International Airport. Therefore, it important assess ensure that National Strategic Projects managed due regard environmental impacts

Language: Английский

---

Assessing and segmenting salt-affected soils using in-situ EC measurements, remote sensing, and a modified deep learning MU-NET convolutional neural network

Ecological Informatics, Journal Year: 2024, Volume and Issue: 81, P. 102652 - 102652

Published: May 26, 2024

Language: Английский

---

Tradeoffs among multi-source remote sensing images, spatial resolution, and accuracy for the classification of wetland plant species and surface objects based on the MRS_DeepLabV3+ model

Ecological Informatics, Journal Year: 2024, Volume and Issue: 81, P. 102594 - 102594

Published: April 8, 2024

Classification of wetland plant species (PlatSpe) and surface objects (SurfObj) in remote sensing images faces significant challenges due to the high diversity PlatSpe fragmented nature SurfObj. Unmanned aerial vehicle (UAV) satellite are primary data sources for classification However, there is still insufficient research on effect various spatial resolutions results. This study essentially focuses Huixian Wetland Guilin, Guangxi, China through utilizing UAV with varying as sources. To this end, MRS_DeepLabV3+ model constructed based multi-resolution segmentation DeepLabV3+, SurfObj appropriately classified model. The obtained results reveal that: (1) optimal scale parameter (SP) capable achieving higher accuracy compared DeepLabV3+. SPs both gradually lessen decreasing resolution, require larger images. (2) In image models, OA kappa exhibit a trend reduction resolution. (3) overall accuracies models superior resolution intervals 2 16 m. investigation can be regarded valuable reference selecting classification.

Language: Английский

---

Comparison of machine and deep learning algorithms using Google Earth Engine and Python for land classifications

Frontiers in Environmental Science, Journal Year: 2024, Volume and Issue: 12

Published: May 28, 2024

Classifying land use and cover (LULC) is essential for various environmental monitoring geospatial analysis applications. This research focuses on classification in District Sukkur, Pakistan, employing the comparison between machine deep learning models. Three satellite indices, namely, NDVI, MNDWI, NDBI, were derived from Landsat-8 data utilized to classify four primary categories: Built-up Area, Water Bodies, Barren Land, Vegetation. The main objective of this study evaluate compare effectiveness models including Random Forest achieved an overall accuracy 91.3% a Kappa coefficient 0.90. It accurately classified 2.7% area as 1.9% 54.8% 40.4% While slightly less accurate, Decision Tree model provided reliable classifications. Deep showed significant accuracy, Convolutional Neural Networks (CNN) Recurrent (RNN). CNN impressive 97.3%, excelling classifying Bodies with User Producer Accuracy exceeding 99%. RNN model, 96.2%, demonstrated strong performance categorizing These findings offer valuable insights into potential applications perfect classifications, implications management analysis. rigorous validation comparative these contribute advancing remote sensing techniques their utilization tasks. presents contribution field underscores importance precise context sustainable conservation.

Language: Английский

---

Bare ground classification using a spectral index ensemble and machine learning models optimized across 12 international study sites

Geocarto International, Journal Year: 2025, Volume and Issue: 40(1)

Published: March 1, 2025

Language: Английский

---

An improved graph neural network integrating indicator attention and spatio-temporal correlation for dissolved oxygen prediction

Ecological Informatics, Journal Year: 2025, Volume and Issue: unknown, P. 103126 - 103126

Published: April 1, 2025

Language: Английский

---

Application of the DKPR Method to Tropical Conditions Using an Integrated Approach to Assess the Vulnerability of Soubré Lake (Southwest, Côte d’Ivoire)

Water Conservation Science and Engineering, Journal Year: 2025, Volume and Issue: 10(1)

Published: April 1, 2025

Language: Английский

---

Assessing Land Cover Classification Accuracy: Variations in Dataset Combinations and Deep Learning Models

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(14), P. 2623 - 2623

Published: July 18, 2024

This study evaluates land cover classification accuracy through adjustments to the deep learning model (DLM) training process, including variations in loss function, rate scheduler, and optimizer, along with diverse input dataset compositions. DLM datasets were created by integrating surface reflectance (SR) spectral data from satellite imagery textural information derived gray-level co-occurrence matrix, yielding four distinct datasets. The U-Net served as baseline, models A B configured adjusting parameters. Eight classifications generated two conditions. Model B, utilizing a comprising spectral, textural, terrain information, achieved highest overall of 90.3% kappa coefficient 0.78. Comparing different compositions, incorporating alongside SR significantly enhanced accuracy. Furthermore, using combination multiple functions or dynamically effectively mitigated overfitting issues, enhancing compared single function.

Language: Английский

---