Estimating land cover map accuracy and area uncertainty using a confusion matrix: A case study in Kalimantan, Indonesia

Remote sensing is widely used to generate land cover maps, but the maps derived from remote sensing often produce accuracy below expectations for map error. Therefore, quantifying map accuracy is essential for reporting the precision of an estimated area. This study describes a simple framework for assessing map accuracy and estimating land cover area uncertainty for a land cover changes map for Kalimantan in 2012-2018. This study compared simple random sampling and stratified random sampling to determine suitable procedures for estimating accuracy and area uncertainty. The validation relies on the visual assessment of high spatial resolution images such as SPOT 6/7 and high-resolution temporal images from Open Foris Collect Earth. Our results showed that the land cover change map assessed using random sampling had an overall accuracy of 74% while using stratified random sampling had an overall accuracy of 75%. Thus, for tropical regions with high cloud cover, we recommend using stratified random sampling. The major source of map error was in differentiating between native forest and plantation areas. Future map improvement requires more accurate differentiation between forest and plantation to better support national forest monitoring systems for sustainable forest management.

Development and optimization of a portable guidance system for the blind and visually impaired with the application of artificial intelligence

This work describes a new type of portable, self-regulating guidance system, which learns to recognize obstacles with the help of a camera, artificial intelligence, and various sensors and thus warn the wearer through audio signals. For obstacle detection, a MobileNetV2 model with an SSD attachment is used which was trained on a custom dataset. Moreover, the system uses the data of motion and distance sensors to improve accuracy. Experimental results confirm that the system can operate with 74.9% mAP accuracy and a reaction time of 0.15 seconds, meeting the performance standard for modern object detection applications. It will also be presented how those affected commented on the device and how the system could be transformed into a marketable product.

De jure property rights and state capacity: evidence from land specification in the Boer Republics

We examine the development of de jure property rights to land by assessing how accurately governments recorded borders of property. We use surveys of farm parcels from two historical states, the Republic of the Orange Free State (OFS) and the South African Republic (ZAR), which are in modern-day South Africa, and employ a descriptive analysis to infer how accurately maps represent parcels of property. We argue that differences in state administrative capacity explains differences in map accuracy and therefore the provision of de jure property rights to land. We find that maps of farms in the ZAR, which had lower administrative capacity, tend to be less accurate than maps of farms in the OFS. Comparisons with military maps compiled under a different administration provide evidence that the costs incurred from previous administrations can limit future attempts to accurately record property. The analysis shows how state administrative capacity can facilitate (or hinder) the provision of property rights to land.

Greek National Map of Soil Organic Carbon

<p>Soil organic carbon (SOC) is the carbon that remains in the soil after the partial decomposition of any material produced by living organisms. It is an essential parameter for agricultural production, the potential sequestration of CO₂ in soil and a vital soil function for global carbon cycle. However, a vast potential of soil carbon is removed from agricultural soils due to non-sustainable soil management practices. Mapping SOC and its changes over time and space is highly valuable for estimating the CO₂ emissions and effects of climate change to the environment. In the present work, the Greek National Map of SOC is presented calculating the SOC stock in 30 arc-seconds spatial resolution using the Global Soil Partnership and Food and Agriculture Organization of the United Nations (FAO) guidelines for SOC mapping. The presented methodology considers the reference framework of the SCORPAN model for digital soil mapping, which can predict SOC stocks in correspondence with soil forming factors. Among the key variables used for estimating SOC stocks are environmental covariates such as climate and meteorological data, thematic maps, digital terrain data, geomorphometry and soil data. Data mining and geostatistical techniques (random forests, support vector machines, regression-kriging) are used to estimate the SOC stocks. Internal and external map accuracy is used to evaluate the performance of the Greek National SOC map. Accuracy of FAO’s methodology was examined herein using different modelling approaches. As indicated in the results, the most accurate map was produced by the random forest technique and an accuracy of FAC2=0.968, RMSE=0.322 and r=0.756. The main findings are also discussed herein covering aspects relevant to the method implementation, validation and feasibility of operational implementation.</p><p><strong>Keywords: </strong>soil organic carbon, climate change, soil management practices, Greek National Map</p>

Refinement of Draft Genome Assemblies of Pigeonpea (Cajanus cajan)

Genome assembly of short reads from large plant genomes remains a challenge in computational biology despite major developments in next generation sequencing. Of late several draft assemblies have been reported in sequenced plant genomes. The reported draft genome assemblies of Cajanus cajan have different levels of genome completeness, a large number of repeats, gaps, and segmental duplications. Draft assemblies with portions of genome missing are shorter than the referenced original genome. These assemblies come with low map accuracy affecting further functional annotation and the prediction of gene components as desired by crop researchers. Genome coverage, i.e., the number of sequenced raw reads mapped onto a certain location of the genome is an important quality indicator of completeness and assembly quality in draft assemblies. The present work aimed to improve the coverage in reported de novo sequenced draft genomes (GCA_000340665.1 and GCA_000230855.2) of pigeonpea, a legume widely cultivated in India. The two recently sequenced assemblies, A1 and A2 comprised 72% and 75% of the estimated coverage of the genome, respectively. We employed an assembly reconciliation approach to compare the draft assemblies and merge them, filling the gaps by employing an algorithm size sorting mate-pair library to generate a high quality and near complete assembly with enhanced contiguity. The majority of gaps present within scaffolds were filled with right-sized mate-pair reads. The improved assembly reduced the number of gaps than those reported in draft assemblies resulting in an improved genome coverage of 82.4%. Map accuracy of the improved assembly was evaluated using various quality metrics and for the presence of specific trait-related functional genes. Employed pair-end and mate-pair local libraries helped us to reduce gaps, repeats, and other sequence errors resulting in lengthier scaffolds compared to the two draft assemblies. We reported the prediction of putative host resistance genes against Fusarium wilt disease by their performance and evaluated them both in wet laboratory and field phenotypic conditions.

Limitations of Predicting Substrate Classes on a Sedimentary Complex but Morphologically Simple Seabed

The ocean floor, its species and habitats are under pressure from various human activities. Marine spatial planning and nature conservation aim to address these threats but require sufficiently detailed and accurate maps of the distribution of seabed substrates and habitats. Benthic habitat mapping has markedly evolved as a discipline over the last decade, but important challenges remain. To test the adequacy of current data products and classification approaches, we carried out a comparative study based on a common dataset of multibeam echosounder bathymetry and backscatter data, supplemented with groundtruth observations. The task was to predict the spatial distribution of five substrate classes (coarse sediments, mixed sediments, mud, sand, and rock) in a highly heterogeneous area of the south-western continental shelf of the United Kingdom. Five different supervised classification methods were employed, and their accuracy estimated with a set of samples that were withheld. We found that all methods achieved overall accuracies of around 50%. Errors of commission and omission were acceptable for rocky substrates, but high for all sediment types. We predominantly attribute the low map accuracy regardless of mapping approach to inadequacies of the selected classification system, which is required to fit gradually changing substrate types into a rigid scheme, low discriminatory power of the available predictors, and high spatial complexity of the site relative to the positioning accuracy of the groundtruth equipment. Some of these issues might be alleviated by creating an ensemble map that aggregates the individual outputs into one map showing the modal substrate class and its associated confidence or by adopting a quantitative approach that models the spatial distribution of sediment fractions. We conclude that further incremental improvements to the collection, processing and analysis of remote sensing and sample data are required to improve map accuracy. To assess the progress in benthic habitat mapping we propose the creation of benchmark datasets.