Quantitatively revealing the response of carbon stocks to land use change (LUCC) and analyzing the vulnerability of ecosystem carbon stock (ECS) services are of great significance for maintaining the carbon cycle and ecological security. For this study, China’s Guizhou Province was the study area. Land use data in 2000, 2010, and 2020 were selected to explore the impacts of LUCC on carbon stocks in multiple scenarios by combining the PLUS and InVEST models and then analyzing the vulnerability of ECS services. The results show that forest land plays an important role in improving ECS services in karst plateau mountainous areas. In 2000–2020, forest land expansion offset the carbon stock reduced by the expansion of built-up land, greatly improving the regional ECS function. Following the natural trend (NT), the total carbon stock in Guizhou Province will decrease by 1.86 Tg; however, under ecological protection (EP) measures, the ECS service performs a positive function for LUCC. Focusing on socioeconomic development (ED) will increase the vulnerability of the regional ECS service. In the future, the forest land area size should be increased, and built-up land should be restricted to better improve the service function of ECS in karst plateau mountainous areas.

Exploring the evolution of livelihood patterns of farm households relocated for poverty alleviation in three prefectures of Guizhou Province, China, can provide a theoretical basis and practical value for achieving sustainable livelihoods of relocated farm households in ethnic minority areas. This study took three ethnic minority autonomous prefectures in Guizhou as the research area. The types and evolution pathways of farm households’ livelihood patterns were divided using 355 tracking and monitoring data from the four phases of relocated farm households’ livelihoods. The evolution characteristics of farm households’ livelihood patterns and changes in their evolution pathways from “before relocation” to 2021 were analyzed. An indicator system of influencing factors was established based on the sustainable livelihood framework. The influencing factors of the evolution path selection and transformation of livelihood patterns were identified using the binary and multiple logistic regression models. The main results are as follows: (1) Implementing the poverty alleviation relocation project shifted the livelihood activities of farm households from agricultural production to non-agricultural industries. The transformation trend of livelihood patterns shifted from agro-dominated, agricultural, agricultural-diversified and balanced types to highly diversified, deeply diversified and subsidy-dependent types. (2) The transformation mode of the evolution pathways of farm households’ livelihood patterns mainly included the stable transformation modes of “active$\to$retention$\to$active” and “active$\to$retention$\to$retention” and the fluctuating transformation modes of “active$\to$retention$\to$fallback”, “active$\to$fallback$\to$active” and “retention$\to$fallback$\to$active”. (3) Natural capital promoted farm households to choose fallback pathways, while material and human capital showed an inhibitory impact. In contrary to human capital, social capital promoted farm households to choose retention pathways. Financial and human capital promoted farm households to choose active pathways. (4) Per capita income, the household head age and relocation years positively affected the evolution pathways of the livelihood pattern from the fallback type to the retention type. The number of means of transport, per capita education level, transportation and communication cost, and the presence of civil servants in the household positively affected the evolution pathways of the livelihood pattern from the fallback type to the active type. Policy suggestions are as follows: Expanding local and nearby employment channels, strengthening vocational skills training for farm households, expanding financial access for farm households, improving policy support for ethnic characteristic industries, implementing bottom-line guarantee policies and exploring ethnic interactive activities for resettlement farm households.

Rapidly and accurately extracting tobacco plant information can facilitate tobacco planting management, precise fertilization, and yield prediction. In the karst mountainous of southern China, tobacco plant identification is affected by large ground undulations, fragmented planting areas, complex and diverse habitats, and uneven plant growth. This study took a tobacco planting area in Guizhou Province as the research object and used DJI UAVs to collect UAV visible light images. Considering plot fragmentation, plant size, presence of weeds, and shadow masking, this area was classified into eight habitats. The U-Net model was trained using different habitat datasets. The results show that (1) the overall precision, recall, F1-score, and Intersection over Union (IOU) of tobacco plant information extraction were 0.68, 0.85, 0.75, and 0.60, respectively. (2) The precision was the highest for the subsurface-fragmented and weed-free habitat and the lowest for the smooth-tectonics and weed-infested habitat. (3) The weed-infested habitat with smaller tobacco plants can blur images, reducing the plant-identification accuracy. This study verified the feasibility of the U-Net model for tobacco single-plant identification in complex habitats. Decomposing complex habitats to establish the sample set method is a new attempt to improve crop identification in complex habitats in karst mountainous areas.

Rapidly and accurately extracting tobacco plant information can facilitate tobacco planting man-agement, precise fertilization and yield prediction. In karst plateau of southern China, tobacco plant identification is affected by large ground undulations, fragmented planting areas, complex and diverse habitats and uneven plant growth. This study took a tobacco planting area in Gui-zhou Province as the research object and used DJI UAVs to collect UAV visible light images. Considering plot fragmentation, plant size, presence of weeds and shadow masking, this area was classified into eight habitats. The datasets of different habitats were constructed to train the U-Net model. The results show that (1) the overall precision, recall, F1-score and IOU of tobacco plant information extraction were 0.68, 0.85, 0.75 and 0.60, respectively. (2) The precision was the highest for the subsurface-fragmented and weed-free habitat and the lowest for the smooth-tectonics and weed-infested habitat. (3) The weed-infested habitat with smaller tobacco plants can cause blurred images, reducing the plant identification accuracy. This study verified the feasibility of the U-Net model for tobacco single-plant identification in complex habitats. Decomposing complex habitats to establish the sample set method is a new attempt to improve crop identification in complex habitats in karst mountainous areas.