Chiang Mai Journal of Science

     Soil salinization is a severe soil degradation process which represents a critical ecological challenge, threatening the sustainable development of agriculture in the Kashgar oasis region of Xinjiang. Therefore, the timely and efficient monitoring and the accurate estimation of soil salinity are highly imperative for the prevention and management of soil salinization. The study presented in this paper involves the development of a new soil salinity inversion model based on the TabNet deep learning algorithm using remote sensing data and environmental variables. The model outperforms common ensemble learning algorithms based on decision trees. This improvement is achieved through the use of attention mechanism and the deep learning architecture in TabNet. In addition, the novelty of the proposed soil salinity inversion model lies in its use of deep learning to construct a inversion model for salinization. The feature variable dataset was initially constructed using the land surface parameters derived from Landsat 8 imagery and other environmental variables influencing soil salinity. This includes data pre-processing for feature selection using the XGBoost model. Separate soil salinity inversion models were developed using XGBoost, LightGBM, CatBoost, CNN and TabNet algorithms, and their performance was compared. The results indicate that TabNet achieved the best predictive performance among the five models, with R² = 0.57, MAE = 8.10, and RMSE = 11.53 on the test dataset. The results of the best performing model, TabNet, and the importance of individual features were subsequently analyzed using SHAP. The effect of some important factors such as groundwater table depth and altitude on salinization is clearly evident. Furthermore, the threshold of groundwater table depth for salinity control in Kashgar was also determined. These results were consistent with the expertise in soil salinization, which further validates the accuracy of the research findings.

     This review traces the development of analytical chemistry as a major research contributor at Chiang Mai University (CMU) over its 60-year life. The evolution of analytical chemistry at CMU has been systematically traced using bibliometric analysis of the scientific literature over each of the six decades. The simplicity and creativity that are features of the research work were born of necessity, given the limitations on equipment and resources in the earlier decades. These underlying themes of simplicity and creativity have been perpetuated throughout the 60-year period despite increasingly sophisticated research, through the use of simple materials for fabrication, natural reagents, common information technology (IT) devices for detection, with an emphasis on green chemistry and sustainable approaches.

    This study investigates the spatial and seasonal distribution of potentially harmful dinoflagellates (PHDs) and their environmental drivers in the coastal waters of Songkhla Province, Thailand. Sampling was conducted across five stations: two semi-enclosed coastal areas (Site A: A1 and A2) and three exposed coastal areas (Site B: B1, B2 and B3). The sampling period, from July 2023 to March 2024, covered the Southwest Monsoon (SWM), Northeast Monsoon (NEM), and Intermediate Dry (IMD) seasons. Twelve PHD species were identified. Seven species (Noctiluca scintillans, Dinophysis caudata, D. miles, Tripos furca, T. fusus, T. tripos, and T. trichoceros) were detected consistently across all sites. Noctiluca scintillans and T. furca were the most abundant and widespread across all sites and seasons. ANOVA revealed significant seasonal effects on the abundance of D. caudata, D. miles, and T. macroceros, while T. fusus showed significant spatial variation (p < 0.05). Canonical Correspondence Analysis (CCA) indicated that dissolved oxygen (DO), pH, temperature, and nitrite concentrations were key variables influencing PHD distribution. Diversity was higher at Site B (H′ = 1.45, E = 0.61, S = 11) than at Site A (H′ = 1.30, E = 0.54, S = 11). Low DO and high nutrient levels in semi-enclosed areas were associated with freshwater inflows and aquaculture activity, whereas exposed coastal stations showed greater physicochemical stability but remained sensitive to nutrient enrichment during the NEM and IMD seasons. Integrated statistical analyses underscored the role of both monsoonal hydrodynamics and anthropogenic nutrient loading in regulating harmful dinoflagellate assemblages along the Songkhla coast. The findings emphasize the importance of site-specific, seasonally adaptive monitoring to the management of coastal water quality and mitigation of harmful algal blooms. The insights provided into the ecological dynamics of tropical coastal systems can inform future strategies for sustainable coastal zone management and support the aims of Sustainable Development Goal (SDG) 14: Life Below Water.

     Bambusicolous fungi comprise a diverse assemblage of species that inhabit a wide range of bamboo species. During an ongoing investigation of bambusicolous fungi in Guangdong Province, China, two interesting strains were isolated from decaying culms of Phyllostachys edulis and Bambusa sinospinosa. Based on morphological characteristics and phylogenetic analysis, the two collections are identified as the sexual and asexual states of a new species, Roussoella yangjiangensis. Roussoella yangjiangensis is characterized by bi-loculate ascostromata, cylindrical asci, fusiform, yellowish brown to dark brown, 1-septate, longitudinally striated ascospores with a sheath, and an asexual morph producing pseudostromatic pycnidia, monophialidic conidiogenous cells, and cylindrical to oblong, brown, aseptate conidia. Phylogenetic analysis of combined LSU, ITS, tef1-α, and rpb2 sequence data reveals that R. yangjiangensis is closely related to R. aseptata and R. yunnanensis, but forms a separate branch. A morphological comparison among these phylogenetically related taxa and other morphologically similar species further support the establishment of the novel species. In this study, the polyphyletic nature of Roussoella is also revealed, and the classification of R. arundinacea, R. chinensis and R. mexicana is discussed.