Title: Comparison of Land Use Land Cover Classifiers Using Different Satellite Imagery and Machine Learning Techniques

Journal: Remote Sensing

DOI: https://doi.org/10.3390/rs14194978

Abstract: Accurate land use land cover (LULC) classification is vital for the sustainable management of natural resources and to learn how the landscape is changing due to climate. For accurate and efficient LULC classification, high-quality datasets and robust classification methods are required. With the increasing availability of satellite data, geospatial analysis tools, and classification methods, it is essential to systematically assess the performance of different combinations of satellite data and classification methods to help select the best approach for LULC classification. Therefore, this study aims to evaluate the LULC classification performance of two commonly used platforms (i.e., ArcGIS Pro and Google Earth Engine) with different satellite datasets (i.e., Landsat, Sentinel, and Planet) through a case study for the city of Charlottetown in Canada. Specifically, three classifiers in ArcGIS Pro, including support vector machine (SVM), maximum likelihood (ML), and random forest/random tree (RF/RT), are utilized to develop LULC maps over the period of 2017–2021. Whereas four classifiers in Google Earth Engine, including SVM, RF/RT, minimum distance (MD), and classification and regression tree (CART), are used to develop LULC maps for the same period. To identify the most efficient and accurate classifier, the overall accuracy and kappa coefficient for each classifier is calculated throughout the study period for all combinations of satellite data, classification platforms, and methods. Change detection is then conducted using the best classifier to quantify the LULC changes over the study period. Results show that the SVM classifier in both ArcGIS Pro and Google Earth Engine presents the best performance compared to other classifiers. In particular, the SVM in ArcGIS Pro shows an overall accuracy of 89% with Landsat, 91% with Sentinel, and 94% with Planet. Similarly, in Google Earth Engine, the SVM shows an accuracy of 87% with Landsat 8 and 92% with Sentinel 2. Furthermore, change detection results show that 13.80% and 14.10% of forest areas have been turned into bare land and urban class, respectively, and 3.90% of the land has been converted into the urban area from 2017 to 2021, suggesting the intensive urbanization. The results of this study will provide the scientific basis for selecting the remote sensing classifier and satellite imagery to develop accurate LULC maps.

The following paper about the comparison of LULC classification methods has been recently accepted for publication by Remote Sensing.

Basheer, S., X. Wang, A. Farooque, R.A. Nawaz, K. Liu, T. Adekanmbi, and S. Liu. Comparison of Land Use Land Cover Classifiers Using Different
Satellite Imagery and Machine Learning Techniques. Remote Sensing, accepted on October 3, 2022.

More details will come soon once the paper is published.

Title: A mixed-level factorial inference approach for ensemble long-term hydrological projections over the Jing River Basin

Journal: Journal of Hydrometeorology

DOI: https://doi.org/10.1175/JHM-D-21-0158.1

Abstract: Long-term hydrological projections can vary substantially depending on the combination of meteorological forcing dataset, hydrologic model, emissions scenario, and natural climate variability. Identifying dominant sources of model spread in an ensemble of hydrologic projections is critically important for developing reliable hydrological projections in support of flooding risk assessment and water resources management; however, it is not well understood due to the multifactor and multi-scale complexities involved in the long-term hydrological projections. Therefore, a stepwise clustered Bayesian (SCB) ensemble method will be first developed to improve the performance of long-term hydrological projections. Meanwhile, a mixed-level factorial inference (MLFI) approach is employed to estimate multiple uncertainties in hydrological projections over the Jing River Basin (JRB). MLFI is able to reveal the main and interactive effects of the anthropogenic emission and model choices on the SCB ensemble projections. The results suggest that the daily maximum temperature under RCP8.5 in the 2050s and 2080s is expected to respectively increase by 3.2 °C and 5.2 °C, which are much higher than the increases under RCP4.5. The maximum increase of the CARM-projected changes in streamflow for the 2050s and 2080s under RCP4.5 is 0.30 and 0.59 × 10³ m³/s in November, respectively. In addition, in a multimodel GCM-RCM-HM ensemble, hydroclimate is found to be most sensitive to the choice of GCM. Moreover, it is revealed that the percentage of contribution of anthropogenic emissions to the changes in monthly precipitation is relatively smaller, but it makes a more significant contribution to the total variance of changes in potential evapotranspiration and streamflow.

Title: Neglected Spatiotemporal Variations of Model Biases in Ensemble-Based Climate Projections

Journal: Geophysical Research Letters

DOI: https://doi.org/10.1029/2022GL098063

Abstract: The Bayesian model averaging (BMA) method has been widely used for generating probabilistic climate projections. However, the averaging weights used in BMA can only reflect the spatially- and temporally-averaged performance of each ensemble member, without the ability to address the spatiotemporal variations of model biases. This can lead to inevitable exaggeration or understatement of the contributions of individual members to the ensemble mean, thus reducing the robustness of the resulting probabilistic projections. Here we propose a new method to help address the neglected spatiotemporal variations of model biases. Through the proposed method, the BMA weights are used as prior distributions to drive the Bayesian discriminant analysis in order to generate refined weights for individual ensemble models according to their spatially- and temporally-clustered performance. Through applying the proposed method to Canada, we demonstrate its effectiveness in generating robust probabilistic climate projections (e.g., the average R² increases from 0.82 to 0.89).

Title: Solar photovoltaics can help China fulfill a net-zero electricity system by 2050 even facing climate change risks

Journal: Resources, Conservation and Recycling

DOI: https://doi.org/10.1016/j.resconrec.2022.106596

Abstract: As China has pledged to become carbon neutral by 2060, electrifying its energy sector is no doubt one of the priority measures to support the transition towards a more sustainable and decarbonized energy system. Solar photovoltaics (PV) has been known as one of the most promising renewable technologies to facilitate the electrification of energy systems. The feasibility of utilizing PV to implement a nationwide decarbonized electricity system now becomes an urgent unanswered question, especially in the context of global climate change and rapid economic growth in China. Here, by using a GIS-based multiple-criteria decision-making approach we address this question by conducting a comprehensive feasibility analysis with consideration of various economic, technological, logistical, and climate change factors. We show that it is feasible for China to fulfill a net-zero electricity system by 2050, through the installation of 7.46 TW solar PV panels on about 1.8% of the national land area (mostly in western China) with a total capital investment of 4.55 trillion USD in the next 30 years. Besides, we show that future climate change may lead to a slight decrease (less than 5%) in solar energy potential, but this would not affect the capability of the nationwide PV system to meet the need for a fully-electrified energy system.