Mangrove-forest classification by using deep learning algorithms has attracted increasing attention but remains challenging. The current studies on the transfer classification of mangrove communities between different regions and different sensors are especially still unclear. To fill the research gap, this study developed a new deep-learning algorithm (encoder–decoder with mixed depth-wise convolution and cascade upsampling, MCCUNet) by modifying the encoder and decoder sections of the DeepLabV3+ algorithm and presented three transfer-learning strategies, namely frozen transfer learning (F-TL), fine-tuned transfer learning (Ft-TL), and sensor-and-phase transfer learning (SaP-TL), to classify mangrove communities by using the MCCUNet algorithm and high-resolution UAV multispectral images. This study combined the deep-learning algorithms with recursive feature elimination and principal component analysis (RFE–PCA), using a high-dimensional dataset to map and classify mangrove communities, and evaluated their classification performance. The results of this study showed the following: (1) The MCCUNet algorithm outperformed the original DeepLabV3+ algorithm for classifying mangrove communities, achieving the highest overall classification accuracy (OA), i.e., 97.24%, in all scenarios. (2) The RFE–PCA dimension reduction improved the classification performance of deep-learning algorithms. The OA of mangrove species from using the MCCUNet algorithm was improved by 7.27% after adding dimension-reduced texture features and vegetation indices. (3) The Ft-TL strategy enabled the algorithm to achieve better classification accuracy and stability than the F-TL strategy. The highest improvement in the F1–score of Spartina alterniflora was 19.56%, using the MCCUNet algorithm with the Ft-TL strategy. (4) The SaP-TL strategy produced better transfer-learning classifications of mangrove communities between images of different phases and sensors. The highest improvement in the F1–score of Aegiceras corniculatum was 19.85%, using the MCCUNet algorithm with the SaP-TL strategy. (5) All three transfer-learning strategies achieved high accuracy in classifying mangrove communities, with the mean F1–score of 84.37~95.25%.

Combining deep learning and UAV images to map wetland vegetation distribution has received increasing attention from researchers. However, it is difficult for one multi-classification convolutional neural network (CNN) model to meet the accuracy requirements for the overall classification of multi-object types. To resolve these issues, this paper combined three decision fusion methods (Majority Voting Fusion, Average Probability Fusion, and Optimal Selection Fusion) with four CNNs, including SegNet, PSPNet, DeepLabV3+, and RAUNet, to construct different fusion classification models (FCMs) for mapping wetland vegetations in Huixian Karst National Wetland Park, Guilin, south China. We further evaluated the effect of one-class and multi-class FCMs on wetland vegetation classification using ultra-high-resolution UAV images and compared the performance of one-class classification (OCC) and multi-class classification (MCC) models for karst wetland vegetation. The results highlight that (1) the use of additional multi-dimensional UAV datasets achieved better classification performance for karst wetland vegetation using CNN models. The OCC models produced better classification results than MCC models, and the accuracy (average of IoU) difference between the two model types was 3.24–10.97%. (2) The integration of DSM and texture features improved the performance of FCMs with an increase in accuracy (MIoU) from 0.67% to 8.23% when compared to RGB-based karst wetland vegetation classifications. (3) The PSPNet algorithm achieved the optimal pixel-based classification in the CNN-based FCMs, while the DeepLabV3+ algorithm produced the best attribute-based classification performance. (4) Three decision fusions all improved the identification ability for karst wetland vegetation compared to single CNN models, which achieved the highest IoUs of 81.93% and 98.42% for Eichhornia crassipes and Nelumbo nucifera, respectively. (5) One-class FCMs achieved higher classification accuracy for karst wetland vegetation than multi-class FCMs, and the highest improvement in the IoU for karst herbaceous plants reached 22.09%.

Global warming has exacerbated the impact of regional drought on vegetation ecosystems, especially in typical karst areas with fragile ecosystems that are more severely affected by drought. However, the response mechanisms of vegetation ecosystems in karst areas to drought stress are still uncertain. With drought stress in the summer of 2022, we examined the spatiotemporal patterns of drought in a World Heritage karst site, Guilin, China, and revealed the exacerbated drought impacts on vegetation ecosystems in karst areas with various vegetation indices. Firstly, we analyzed the spatiotemporal characteristics of drought from 2000 to 2022, utilizing the temperature vegetation dryness index (TVDI), highlighting the intra-annual variability of drought in 2022. Additionally, we compared the responses of different vegetation types to drought stress in karst and non-karst areas and explored the exacerbated impacts of drought stress on vegetation ecosystems within the same year with three vegetation indices, namely, the Normalized Difference Vegetation Index (NDVI), Leaf Area index (LAI), and Gross Primary Production (GPP) in karst areas. The results showed that drought started in July and persisted from August to November at moderate to severe levels (with severe drought in September), eventually easing in December. Karst areas exhibited severe drought (TVDI = 0.76), which more significantly impacted regional vegetation ecosystems than those in non-karst areas. Different vegetation types also experienced greater drought stress in karst areas compared to non-karst areas. The vegetation indices increased at the early- to mid-stages of drought (July to September) compared to those in the baseline year (2020–2021), mainly due to the increase in non-karst areas. However, vegetation indices decreased at the late drought stage (October to November), primarily due to the decrease in karst areas, indicating that the karst topography exacerbated the impact of drought on regional vegetation ecosystems. Since LAI and GPP exhibited similar changing patterns to TVDI, with GPP showing particularly strong alignment, they can be used to reveal the response mechanisms of ecosystems to drought stress in karst areas. We emphasize the importance of monitoring the responses of vegetation ecosystems to climate-induced droughts stress and enhancing their resilience to future climatic challenges, particularly in karst areas.

Rural agricultural activity generates cropland, secondary vegetation and straggling primary forest and can modify the soil seed bank (SSB), potentially impacting the restoration of preferred species. The interaction between vegetation and seed banks during the recovery process is dependent on management practices and recovery pathways. This study was carried out in Guilin of southwest China to assess the variation in plant diversity and species composition of both aboveground and soil seed banks across three typical vegetation types with different human interventions: orchard, bamboo shrub and primary forest. The results show that there were significant differences in the species composition and diversity of aboveground vegetation and SSB, as well as in soil properties among three typical vegetation types. The primary forest had the highest aboveground species diversity, while the orchard had the highest species diversity and seed density of SSB. In addition, principal component analysis (PCA) and canonical correspondence analyses (CCAs) showed that the species composition and plant life forms of the three typical vegetation types were significantly influenced by soil properties. Based on these findings, the characteristics of aboveground vegetation and the soil seed bank and their correlations with soil properties are expected to drastically change with human intervention. These results imply that unsustainable land use has greatly impacted soil properties, and consequently, the aboveground vegetation and SSB. Nevertheless, vegetation will recover quickly after farming is abandoned. The successful restoration of fragmented ecosystems requires the addition of seeds and seedlings of target species, especially perennial woody plants from the relevant natural ecosystems, to accelerate succession from bamboo shrub to forest.

Ecosystem water-use efficiency ($WUE$) has been central in revealing the variability in terrestrial carbon and water cycles. Short-rotation plantations such as Eucalyptus plantations can simultaneously impact net primary production ($NPP$) and actual evapotranspiration (${ET}_a$), components of $WUE$, resulting in changes in terrestrial carbon and water cycles. However, there are few detailed studies on the changes in the $WUE$ of Eucalyptus plantations at the catchment scale with high spatial remote sensing imagery. Here, we present the changes in the $WUE$ of Eucalyptus plantations and its driving factors (i.e., $NPP$ and ${ET}_a$) using satellite-based models combined with 5-m spatial resolution RapidEye imagery in a small county in South China. The increases in ${ET}_a$ of Eucalyptus plantations are primarily the result of climate warming and result in low $WUE$ of Eucalyptus plantations. The management practice used (short rotation in this study) can enhance the effect of climate warming on $WUE$ by varying the $NPP$ of Eucalyptus plantations. A high value of $NPP$ leads to a high $WUE$ of Eucalyptus plantations at the end of a short rotation, while a low value of $NPP$ results in a low $WUE$ at the beginning of another short rotation. Changes in the $WUE$ of Eucalyptus plantations indicated large spatial and temporal variability, associated with climate warming and short-rotation practices.