Honduras is among the ten largest palm oil producers in the world. Accurately detecting the current plantation status and oil palm distribution is crucial to optimize the planting layout. However, current information on single oil palm trees in Honduras is scarce. Here we used submeter-resolution RGB satellite images in 2022 to detect the spatial distribution and single oil palm tree in Honduras with the four-direction gradient difference, single-tree feature extraction, and double-ring detection methods. In the accuracy validation zones, the recall, precision, and F1-score metrics for the spatial distribution of oil palm were 86.50%, 86.57%, and 86.53%, respectively; counting accuracy for single oil palm trees was 86.75%; and localization accuracy was approximately 6 m. We estimated 0.25 million hectares (ha) of oil palm plantation area in Honduras, corresponding to 28.27 million trees and an average density of 113 trees/ha. Furthermore, we categorized planted land into highly or moderately suitable zones for future expansion. Among the 0.25 million ha of potentially available land, the highly and moderately suitable zones represented 10,501 ha and 242,120 ha, respectively, with the potential to carry approximately 28.55 million oil palms. For future oil palm expansion, priority should be given to highly suitable zones. This study not only characterized the distribution and single oil palm properties in Honduras but also suggested potentially suitable zones for future plantation expansion, thereby providing important methodological references and data support to assess the impact of oil palm planting on the ecological environment and socioeconomic aspects at the country scale.

As one of the most important application for remote sensing monitoring, land cover mapping has witnessed notable advancements in data acquisition, algorithmic diversity, and classification accuracy. Despite the instrumental role data-driven algorithms have played in the development of global land cover products, their inherent limitations as “black box” methods often fall short of meeting end-users’ specific requirements. In this study, built upon the foundation of the earlier land cover monitoring platform [FROM-GLC plus(FGP)], a data and knowledge dual-driven framework (FGP 2.0) was developed as a user-adaptive framework for intelligent remote sensing land cover mapping. By incorporating ontology-based semantic descriptions with advanced data-driven algorithms, FGP 2.0 provides the capacity for both traditional annual mapping and emerging dynamic mapping. Our results illustrate that FGP 2.0 significantly improves the overall accuracy of annual maps by ~5%, and dynamic maps by ~20% compared to FGP. Moreover, an operational dynamic mapping tool has been developed on the Google Earth engine (GEE), enabling the generation of near-real-time land cover maps for any given place. With an extensible and flexible mapping framework, FGP 2.0 demonstrates the potential of customized land cover monitoring results to suit different application scenarios. This innovative approach not only meets the current demand for reliable annual and dynamic land cover maps but also sets a new benchmark for the integration of geoscientific expertise with machine learning techniques in remote sensing monitoring.

High-resolution land cover mapping (LCM) is an important application for studying and understanding the change of the earth surface. While deep learning (DL) methods demonstrate great potential in analyzing satellite images, they largely depend on massive high-quality labels. This paper proposes SW-LCM, a Scalable and Weakly-supervised two-stage Land Cover Mapping method on a new Sunway Supercomputer. Our method consists of a k-means clustering module as a first stage, and an iterative deep learning module as a second stage. With the k-means module providing a good enough starting point (taking inaccurate results as noisy labels), the deep learning module improves the classification results in an iterative way, without any labelling efforts required for processing large scenarios. To achieve efficiency for country-level land cover mapping, we design a customized data partition scheme and an on-the-fly assembly for k-means. Through careful parallelization and optimization, our k-means module scales to 98,304 computing nodes (over 38 million cores), and provides a sustained performance of 437.56 PFLOPS, in a real LCM task of the entire region of China; the iterative updating part scales to 24,576 nodes, with a performance of 11 PFLOPS. We produce a 10-m resolution land cover map of China, with an accuracy of 83.5% (10-class) or 73.2% (25-class), 7% to 8% higher than best existing products, paving ways for finer land surveys to support sustainability-related applications.

Providing accurate and timely oil palm information on a large scale is essential for both economic development and ecological significance. However, owing to different sensors, photograph acquisition conditions, and environmental heterogeneity, the large volume and the variety of the data make it extremely challenging for large-scale and cross-regional oil palm tree detection. It is computationally expensive to train a model from images covering large heterogeneous regions and all environmental conditions for continuously accumulated multisource remote sensing data. In this letter, we propose a new multisource domain generalization (DG) method, Maximum Mean Discrepancy Deep Reconstruction Classification Network (MMD-DRCN). It learns representations from multiple source domains and obtains inspiring performance in an unknown and “unseen” target domain. Besides classification loss, our MMD-DRCN distills more representative features through reconstruction loss and aligns multisource latent features by MMD loss, both of which effectively enhance the capacity of generalization. MMD-DRCN achieves an average F1-score of 82.70% in all transfer tasks, attaining a 5.83% gain compared to Baseline (a straightforward convolutional neural network (CNN) model). Experimental results demonstrate DG poses a promising potential for large-scale and cross-regional oil palm tree detection without any information of the target domain.

Synthetic aperture radar (SAR) plays an essential role in earth observation and projection due to its capability to penetrate clouds, which makes it possible to monitor terrestrial surfaces under all weather conditions. Multiplicative noise often occurs in the SAR signal, hampering the retrieval of information from SAR imagery. Convolutional neural networks (CNNs) have been used in many computer vision tasks and are helpful in image denoising. However, current CNN-based denoising approaches inevitably lead to a “washed out” effect that loses spatial details. Another limitation is that most typical CNN-based denoising models require a noise-free image for training. To address these issues, we propose a novel end-to-end self-supervised SAR denoising model: Enhanced Noise2Noise (EN2N), which can be trained without a noise-free image. To enhance the quality of the result images, the perceptual features from a pre-learned CNN are introduced to restore the spatial details by a hybrid loss function. Experiments show that our proposed method outperforms the typical denoising methods in terms of noise reduction and feature preservation based on image quality metrics. Also, the new hybrid loss could enhance the spatial details significantly. The good performance maintains the robustness throughout time, which reduces the uncertainty in time-series SAR caused by random noise. Benefiting from optimization of graphics processing unit (GPU) and multi-threading, the proposed method has higher computation efficiency than traditional methods. This study demonstrates the great potential of using our self-supervised deep learning approaches for SAR image denoising in the future.

The Coconut tree is of great importance in economic values and ecological impacts for many tropical developing countries and lots of islands in the Pacific Ocean. Detecting and counting coconut is a meaningful and valuable research. In this paper, we present a coconut tree crown detection method to detect and count the coconut trees in the Tenarunga from high-resolution satellite images acquired by Google Earth. Our coconut tree detection method contains three major procedures: feature extraction, a multi-level Region Proposal Network (RPN) and a large-scale coconut tree detection workflow. We manually annotate all coconut trees for our study regions in the Tenarunga. Eventually, we achieve a higher average F1-score of 77.14% in our four test regions than pure Faster R-CNN. Experiment results demonstrate the potential for large-scale individual coconut tree detection and counting from high-resolution satellite images using deep learning.

This paper proposed a deep convolutional neural network (DCNN) based framework for large-scale oil palm tree detection using high-resolution remote sensing images in Malaysia. Different from the previous palm tree or tree crown detection studies, the palm trees in our study area are very crowded and their crowns often overlap. Moreover, there are various land cover types in our study area, e.g. impervious, bare land, and other vegetation, etc. The main steps of our proposed method include large-scale and multi-class sample collection, AlexNet-based DCNN training and optimization, sliding window-based label prediction, and post-processing. Compared with the manually interpreted ground truth, our proposed method achieves detection accuracies of 92%-97% in our study area, which are greatly higher than the accuracies obtained from another two detection methods used in this paper.

Urban cellular automata (CA) models propagate and accumulate errors during the modeling process due to the model structure or stochastic processes involved. It is feasible to assimilate real-time observations into an urban CA model to reduce model uncertainties. However, the assimilation performance is sensitive to the spatio-temporal units in the assimilation algorithm, that is, spatial block size and window length (temporal interval). In this study, we coupled an assimilation model, an ensemble Kalman filter (EnKF) and a Logistic-CA model to simulate the urban dynamic in Beijing over a period of two decades. Our results indicate that the coupled EnKF-CA model outperforms the CA-alone counterpart by about 10% in terms of the figure of merit, which reflects the agreement of modeled pixels. We also find that the assimilation performance using a finer block (1 km) is better than that using a coarser block (5 km and 10 km) because of the better depiction of spatial heterogeneity using a finer block. Moreover, the improvement of intermediate outputs using the coupled EnKF-CA model is effective for a certain period (e.g. 5 years). This implies that a high-frequency assimilation may not significantly improve the model performance. The sensitivity analyses of spatio-temporal assimilation in the EnKF-CA model provide a better understanding of the assimilation mechanism that couples with land-use change models.

Cellular automata (CA) have emerged as a primary tool for urban growth modeling due to its simplicity, transparency, and ease of implementation. Sensitivity analysis is an important component in CA modeling for a better understanding of errors or uncertainties and their propagation. Most studies on sensitivity analyses in urban CA modeling focus on specific component such as neighborhood configuration or stochastic perturbation. However, sensitivity analysis of transition rules, which is one of the core components in CA models, has not been systematically done. This article proposes a systematic sensitivity analysis of major operational components in urban CA modeling using a stepwise comparison approach. After obtaining transition rules, three stages (i.e. static calibration of transition rules, dynamic evolution with varied time steps, and incorporation with stochastic perturbation) are designed to facilitate a comprehensive analysis. This scheme implemented with a case study in Guangzhou City (China) reveals that gaps in performance from static calibration with different transition rules can be reduced when dynamic evolution is considered. Moreover, the degree of stochastic perturbation is closely related to obtain urban morphology. However, a more realistic (i.e. fragmented) urban landscape is achieved at the cost of decreasing pixel-based accuracy in this study. Thus, a trade-off between pixel-based and pattern-based comparisons should be balanced in practical urban modeling. Finally, experimental results illustrate that models for transition rules extraction with good quality can do an assistance for urban modeling through reducing errors and uncertainty range. Additionally, ensemble methods can feasibly improve the performance of CA models when coupled with nonparametric models (i.e. classification and regression tree).

A spatial multi-objective land use optimization model defined by the acronym ‘NSGA-II-MOLU’ or the ‘non-dominated sorting genetic algorithm-II for multi-objective optimization of land use’ is proposed for searching for optimal land use scenarios which embrace multiple objectives and constraints extracted from the requirements of users, as well as providing support to the land use planning process. In this application, we took the MOLU model which was initially developed to integrate multiple objectives and coupled this with a revised version of the genetic algorithm NSGA-II which is based on specific crossover and mutation operators. The resulting NSGA-II-MOLU model is able to offer the possibility of efficiently searching over tens of thousands of solutions for trade-off sets which define non-dominated plans on the classical Pareto frontier. In this application, we chose the example of Tongzhou New Town, China, to demonstrate how the model could be employed to meet three conflicting objectives based on minimizing conversion costs, maximizing accessibility, and maximizing compatibilities between land uses. Our case study clearly shows the ability of the model to generate diversified land use planning scenarios which form the core of a land use planning support system. It also demonstrates the potential of the model to consider more complicated spatial objectives and variables with open-ended characteristics. The breakthroughs in spatial optimization that this model provides lead directly to other properties of the process in which further efficiencies in the process of optimization, more vivid visualizations, and more interactive planning support are possible. These form directions for future research.