Table 1.

Different AI applications in climate impacts with their strengths and limitations

S/N	AI models	Applications	Strength	Limitations	Ref
1	Recurrent Neural Networks (RNNs)	Flood prediction using rainfall data	- Suitable for sequential data processes like time series prediction and natural language processing - It can learn long-term dependencies in data	- Times series prediction (e.g., extreme weather conditions prediction) - Natural language processing (e.g., information extraction from environmental reports)	(Ren et al. 2020)
2		Prediction of river flow level			(Liu et al. 2020)
3	Decision Trees	Prediction of water quality in rivers	- It is relatively simple to train and interpret - It can be used for a variety of data types	- Water quality monitoring - Detection and monitoring of deforestation - Air quality monitoring	(Nouraki et al. 2021)
4		Assessment of water quality			(Bui et al. 2020; Nasir et al. 2022)
5	Convolutional Neural Networks (CNNs)	Monitoring of coastal erosion	- It can learn complex patterns from images without explicit programming - Most suitable for image classification and other assignments that involve spatial data	- Image classification (e.g., wildlife identification, deforestation detection)	(Scardino et al. 2022)
6		Monitoring coral reef health through underwater images			(Burns et al. 2022)
7	Support Vector Machines (SVMs)	Air quality monitoring	- Suitable for varying data types, including text and images - It can process high-dimensional data - It can assimilate complex relationships within variables	- Natural language processing (e.g., information extraction from environmental reports) - Time series prediction (e.g., forecasting extreme weather conditions) - Image classification (e.g., wildlife identification, deforestation detection)	(Castelli et al. 2020; Leong et al. 2020)
8		Prediction of harmful algal biomass in lakes			(Henrique et al. 2021)
9	Random Forests	Modeling air quality index in urban areas	- Have better strength for overfitting than individual decision trees - An ensemble learning technique that integrates several decision trees' predictions	- Image classification (e.g., wildlife identification, deforestation detection) - Natural language processing (e.g., information extraction from environmental reports) - Time series prediction (e.g., forecasting extreme weather conditions)	(Alsaber et al. 2021)
10		Prediction of forest fire			(Decastro et al. 2022; Pham et al. 2020)
11	Hybrid models	Combination of RNNs and CNNs to predict air pollution	- Combined the strength of deep learning and machine learning - More robust and accurate than individual models	Not suitable for environmental task monitoring that requires high accuracy and robustness	(Tsokov et al. 2022)
		A combination of CNNs and RNNs is used to predict urban heat islands			(Li and Zheng 2023)