Search

We examine whether the individual ozone (O3) measurements from the four Aura instruments can quantify the stratosphere-troposphere exchange (STE) flux of O3, an important term of the tropospheric O3budget. The level 2 (L2) Aura swath data and the nearly coincident ozone sondes for the years 2005–2006 are compared with the 4-D, high-resolution (1° × 1° × 40-layer × 0.5 h) model simulation of atmospheric ozone for the same period from the University of California, Irvine chemistry transport model (CTM). The CTM becomes a transfer standard for comparing individual profiles from these five, not-quite-coincident measurements of atmospheric ozone. Even with obvious model discrepancies identified here, the CTM can readily quantify instrument-instrument biases in the tropical upper troposphere and mid-latitude lower stratosphere. In terms of STE processes, all four Aura datasets have some skill in identifying stratosphere-troposphere folds, and we find several cases where both model and measurements see evidence of high-O3stratospheric air entering the troposphere. In many cases identified in the model, however, the individual Aura profile retrievals in the upper troposphere and lower stratosphere show too much noise, as expected from their low sensitivity and coarse vertical resolution at and below the tropopause. These model-measurement comparisons of individual profiles do provide some level of confidence in the model-derived STE O3 flux, but it will be difficult to integrate this flux from the satellite data alone.

Problem. Social networks are increasingly becoming an environment for threats, insults, profanity and other destructive manifestations of human communication. Today, a huge number of people are involved in online platforms, and the amount of content created and reactions to it is constantly breaking records. Therefore, there is a need to automate the detection and counteraction of antisocial influences. One of the important areas of such activities is the detection of toxic comments that contain threats, insults, profanity, contempt for others and more. To perform this task, researchers usually build a classifier based on neural networks. And for their training they use a collected or publicly available set of data. The article investigates how different methods of pre-processing of input data affect the final accuracy of the classifier. Previous studies in this direction have confirmed the presence of an impact on the result, but did not allow to draw definitive conclusions about the effectiveness. Goal. Research of preliminary processing of text data methods influence on the destructive messages classifier. Results. It has been shown that the effect of a particular method can be quite dependent on the content in the data set. In addition, it is noted that sometimes the impact may be insignificant, and in some cases may even lead to a worsening of the result. It is also justified the need to pre-check the data set for the percentage of elements that fall under the impact of a particular method. Originality. The methods of data processing are evaluated on the basis of English and Russian data sets. Practical significance. The obtained results allow to make better decisions about the usage of certain pre-processing methods to improve the accuracy of the destructive messages classifier.

National infectious disease incidence rates are often estimated by standardizing locally derived rates using national-level age and race distributions. Data on other factors potentially associated with incidence are often not available in the form of patient-level covariates. Including characteristics of patients' area of residence may improve the accuracy of national estimates. The authors used data from the Centers for Disease Control and Prevention's Active Bacterial Core Surveillance program (2004-2005), adjusted for census-based variables, to estimate the national incidence of invasive pneumococcal disease (IPD). The authors tested Poisson and negative binomial models in a cross-validation procedure to select variables best predicting the incidence of IPD in each county. Including census-level information on race and educational attainment improved the fit of both Poisson and negative binomial models beyond that achieved by adjusting for other census variables or by adjusting for an individual's race and age alone. The Poisson model with census-based predictors led to a national estimate of IPD of 16.0 cases per 100,000 persons as compared with 13.5 per 100,000 persons using an individual's age and race alone. Accuracy of, and confidence intervals for, these estimates can only be determined by obtaining data from other randomly selected US counties. However, incorporating census-derived characteristics should be considered when estimating national incidence of IPD and other diseases.

The development of EGFR kinase inhibitors to treat non-small cell lung cancers is an important medical necessity due to persistent development of resistance by the mutations. In-view of this, a model with a very high predictive ability on thirty (30) EGFR kinase inhibitors was built using QSAR molecular modeling technique. Density Functional Theory method at B3LYP/6-311G* level of theory was used to identify the optimum conformation of the EGFR kinase inhibitors. A Multi-linear regression Genetic Function Algorithm method (MLR-GFA) was used to build five models. The described model among others was selected and reported since it has passed the requirements for good model validation with resulting parameters: R2 of 0.927451, R2adj of 0.901845, Q2CV of 0.869431R2test of 0.767984 and cRp2 of 0.625651. Molecular docking, drug-likeness and ADME studies were carried out to virtually screen these EGFR kinase inhibitors in order to identify a compound with the highest affinity toward the target protein (EGFR kinase enzyme with pdb entry 3IKA) with good pharmacokinetic profile. Compound 11 was identified as the lead compound with the highest binding affinity of −9.8 kcal/mol (and also with good ADME and drug-likeness properties) and was retained as template for designing new EGFR kinase inhibitors. Six new EGFR kinase inhibitors were designed using compound 11 as template. The newly designed compounds were found to have better affinity toward their target protein than the template and Gefitinib (positive control). The reported model predictive performance was further confirmed by predicting the pIC50 of these newly designed compounds.

The article shows the possibilities of Data Set “Interactive Statistics and Intelligent Analytics of the Balanced State of the Regional Economy of Russia in Terms of Big Data and Blockchain – 2020”. For creation of the data set, we formed time rows of the values of the selected indicators, which characterize the balance of Russia's regional economy. The indicators are systematized and classified into two categories. The first category includes the indicators of the level of socio-economic development: GDP per capita and “underdevelopment whirlpools”, balanced financial result of companies’ activities, population's employment level, and balance of the regional budget, calculated by finding the difference between revenues and expenditures. The second category includes the indicators of potential of socio-economic development: investments into fixed capital per capita, share of innovations-active organizations, share of innovative products, and digitization level. The data are collected in a table, based on which – with the help of programming and web-design – a data set is created – an interactive platform for working with data. The data set is available in Russian and English at the official web-site of the Institute of Scientific Communications [12]. The data set unifies and transfers into the digital form the data on the level and potential of Russia's regions’ potential of socio-economic development. The data set allows for flexible setting of data for any research. The data set allows – based on the statistics of a region's socio-economic development – determining its position in the Russian national ranking, category by the “TRMS” matrix, calculating the integral index for each region, and comparing its values. Using the proprietary methodology of calculation of “underdevelopment whirlpools” of Prof. Popkova, the data set reflects the balance of Russia's regional economy in 2005–2024. The data set allows for automatic visualization of data by creating a blockchain polygon of region's socio-economic development, which includes the chronogram of the region's development and the process of region's pulling into an “underdevelopment whirlpool”. The data set's data are presented in the form of an interactive map of Russia's regional economy. Map's color shows categories that are assigned to regions and the borders of regions, as well as information on each region's position in the 2020 rating. The data set allows for large-scale studies of Russia's regional economy with application of technologies of Big Data processing, machine learning, and intellectual analytics.

New burned area datasets and top-down constraints from atmospheric concentration measurements of pyrogenic gases have decreased the large uncertainty in fire emissions estimates. However, significant gaps remain in our understanding of the contribution of deforestation, savanna, forest, agricultural waste, and peat fires to total global fire emissions. Here we used a revised version of the Carnegie-Ames-Stanford-Approach (CASA) biogeochemical model and improved satellite-derived estimates of area burned, fire activity, and plant productivity to calculate fire emissions for the 1997–2009 period on a 0.5° spatial resolution with a monthly time step. For November 2000 onwards, estimates were based on burned area, active fire detections, and plant productivity from the MODerate resolution Imaging Spectroradiometer (MODIS) sensor. For the partitioning we focused on the MODIS era. We used maps of burned area derived from the Tropical Rainfall Measuring Mission (TRMM) Visible and Infrared Scanner (VIRS) and Along-Track Scanning Radiometer (ATSR) active fire data prior to MODIS (1997–2000) and estimates of plant productivity derived from Advanced Very High Resolution Radiometer (AVHRR) observations during the same period. Average global fire carbon emissions according to this version 3 of the Global Fire Emissions Database (GFED3) were 2.0 Pg C year−1 with significant interannual variability during 1997–2001 (2.8 Pg C year−1 in 1998 and 1.6 Pg C year−1 in 2001). Globally, emissions during 2002–2007 were relatively constant (around 2.1 Pg C year−1) before declining in 2008 (1.7 Pg C year−1) and 2009 (1.5 Pg C year−1) partly due to lower deforestation fire emissions in South America and tropical Asia. On a regional basis, emissions were highly variable during 2002–2007 (e.g., boreal Asia, South America, and Indonesia), but these regional differences canceled out at a global level. During the MODIS era (2001–2009), most carbon emissions were from fires in grasslands and savannas (44%) with smaller contributions from tropical deforestation and degradation fires (20%), woodland fires (mostly confined to the tropics, 16%), forest fires (mostly in the extratropics, 15%), agricultural waste burning (3%), and tropical peat fires (3%). The contribution from agricultural waste fires was likely a lower bound because our approach for measuring burned area could not detect all of these relatively small fires. Total carbon emissions were on average 13% lower than in our previous (GFED2) work. For reduced trace gases such as CO and CH4, deforestation, degradation, and peat fires were more important contributors because of higher emissions of reduced trace gases per unit carbon combusted compared to savanna fires. Carbon emissions from tropical deforestation, degradation, and peatland fires were on average 0.5 Pg C year−1. The carbon emissions from these fires may not be balanced by regrowth following fire. Our results provide the first global assessment of the contribution of different sources to total global fire emissions for the past decade, and supply the community with an improved 13-year fire emissions time series.

Images differ in their memorability in consistent ways across observers. What makes an image memorable is not fully understood to date. Most of the current insight is in terms of high-level semantic aspects, related to the content. However, research still shows consistent differences within semantic categories, suggesting a role for factors at other levels of processing in the visual hierarchy. To aid investigations into this role as well as contributions to the understanding of image memorability more generally, we present MemCat. MemCat is a category-based image set, consisting of 10K images representing five broader, memorability-relevant categories (animal, food, landscape, sports, and vehicle) and further divided into subcategories (e.g., bear). They were sampled from existing source image sets that offer bounding box annotations or more detailed segmentation masks. We collected memorability scores for all 10 K images, each score based on the responses of on average 99 participants in a repeat-detection memory task. Replicating previous research, the collected memorability scores show high levels of consistency across observers. Currently, MemCat is the second largest memorability image set and the largest offering a category-based structure. MemCat can be used to study the factors underlying the variability in image memorability, including the variability within semantic categories. In addition, it offers a new benchmark dataset for the automatic prediction of memorability scores (e.g., with convolutional neural networks). Finally, MemCat allows the study of neural and behavioral correlates of memorability while controlling for semantic category.

Behavior recognition is developing rapidly,and a number of behavior recognition algorithms based on deep network automatic learning features have been proposed.The deep learning method requires a large number of data to train,and requires higher computer storage and computing power.After a brief review of the current popular behavior recognition method based on deep network,it focused on the traditional behavior recognition methods.Traditional behavior recognition methods usually followed the processes of video feature extraction,modeling of features and classification.Following the basic process,the recognition process was overviewed according to the following steps,feature sampling,feature descriptors,feature processing,descriptor aggregation and vector coding.At the same time,the benchmark data set commonly used for evaluating the algorithm performance was also summarized.

Long term, high quality estimates of burned area are needed for improving both prognostic and diagnostic fire emissions models and for assessing feedbacks between fire and the climate system. We developed global, monthly burned area estimates aggregated to 0.5° spatial resolution for the time period July 1996 through mid-2009 using four satellite data sets. From 2001–2009, our primary data source was 500-m burned area maps produced using Moderate Resolution Imaging Spectroradiometer (MODIS) surface reflectance imagery; more than 90% of the global area burned during this time period was mapped in this fashion. During times when the 500-m MODIS data were not available, we used a combination of local regression and regional regression trees developed over periods when burned area and Terra MODIS active fire data were available to indirectly estimate burned area. Cross-calibration with fire observations from the Tropical Rainfall Measuring Mission (TRMM) Visible and Infrared Scanner (VIRS) and the Along-Track Scanning Radiometer (ATSR) allowed the data set to be extended prior to the MODIS era. With our data set we estimated that the global annual area burned for the years 1997–2008 varied between 330 and 431 Mha, with the maximum occurring in 1998. We compared our data set to the recent GFED2, L3JRC, GLOBCARBON, and MODIS MCD45A1 global burned area products and found substantial differences in many regions. Lastly, we assessed the interannual variability and long-term trends in global burned area over the past 13 years. This burned area time series serves as the basis for the third version of the Global Fire Emissions Database (GFED3) estimates of trace gas and aerosol emissions.

In supervised learning, a machine learning system requires a data set. In occasions, however, the data set may have learning conflicts that may drastically affect the performance of the learning system. This paper presents a method to analyze the learning conflicts in a data set. Several computer simulations to test and validate our method are performed. Two common functions in the field of optimization are used to create clean data sets. The data sets are, then, contaminated with random data, and the total learning conflict level for each case is computed. The proposed algorithm is used to identify the learning conflicts that are intentionally inserted. Next, an artificial neural network is trained and evaluated using the contaminated data set. The algorithm proposed in this paper is used in a real-world application to detect problems in a data set for a refrigeration system. It is concluded that the algorithm can be used to improve the performance of machine learning systems.

The paper presents View VULMA, a data set specifically designed for training machine-learning tools for elaborating fast vulnerability analysis of existing buildings. Such tools require supervised training via an extensive set of building imagery, for which several typological parameters should be defined, with a proper label assigned to each sample on a per-parameter basis. Thus, it is clear how defining an adequate training data set plays a key role, and several aspects should be considered, such as data availability, preprocessing, augmentation and balancing according to the selected labels. In this paper, we highlight all these issues, describing the pursued strategies to elaborate a reliable data set. In particular, a detailed description of both requirements (e.g., scale and resolution of images, evaluation parameters and data heterogeneity) and the steps followed to define View VULMA are provided, starting from the data assessment (which allowed to reduce the initial sample of about 20.000 images to a subset of about 3.000 pictures), to achieve the goal of training a transfer-learning-based automated tool for fast estimation of the vulnerability of existing buildings from single pictures.

Copyright of Journal of Data Acquisition & Processing / Shu Ju Cai Ji Yu Chu Li is the property of Editorial Department of Journal of Nanjing University of Aeronautics & Astronautics and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

To optimize traumatic spinal cord injury (tSCI) care, administrative and clinical linked data are required to describe the patient's journey. To describe the methods and progress to deterministically link SCI data from multiple databases across the SCI continuum in British Columbia (BC) and Ontario (ON) to answer epidemiological and health service research questions. Patients with tSCI will be identified from the administrative Hospital Discharge Abstract Database using International Classification of Diseases (ICD) codes from Population Data BC and ICES data repositories in BC and ON, respectively. Admissions for tSCI will range between 1995–2017 for BC and 2009-2017 for ON. Linkage will occur with multiple administrative data holdings from Population Data BC and ICES to create the "Admin SCI Cohorts." Clinical data from the Rick Hansen SCI Registry (and VerteBase in BC) will be transferred to Population Data BC and ICES. Linkage of the clinical data with the incident cases and administrative data at Population Data BC and ICES will create subsets of patients referred to as the "Clinical SCI Cohorts" for BC and ON. Deidentified patient-level linked data sets will be uploaded to a secure research environment for analysis. Data validation will include several steps, and data analysis plans will be created for each research question. The creation of provincially linked tSCI data sets is unique; both clinical and administrative data are included to inform the optimization of care across the SCI continuum. Methods and lessons learned will inform future data-linking projects and care initiatives. [ABSTRACT FROM AUTHOR]

Nowadays, Android applications play a major role in software industry. Therefore, having a system that can help companies predict the success probability of such applications would be useful. Thus far, numerous research works have been conducted to predict the success probability of desktop applications using a variety of machine learning techniques. However, since features of desktop programs are different from those of mobile applications, they are not applicable to mobile applications. To our knowledge, there has not been a repository or even a method to predict the success probability of Android applications so far. In this research, we introduce a repository composed of 100 successful and 100 unsuccessful apps of Android operating system in Google PlayStoreTM including 34 features per application. Then, we use the repository to a neural network and other classification algorithms to predict the success probability. Finally, we compare the proposed method with the previous approaches based on the accuracy criterion. Experimental results show that the best accuracy which we achieved is 99.99%, which obtained when we used MLP and PCA, while the best accuracy achieved by the previous work in desktop platforms was 96%. However, the time complexity of the proposed approach is higher than previous methods, since the time complexities of NPR and MLP are O (n 3 ) and O (n p h k o i ), respectively. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Remote Sensing is the property of Editorial Office of Journal of Remote Sensing & Science Publishing Co. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

This paper presents an outlier detection method that is based on a Variable Precision Rough Set Model (VPRSM). This method generalizes the standard set inclusion relation, which is the foundation of the Rough Sets Basic Model (RSBM). The main contribution of this research is an improvement in the quality of detection because this generalization allows us to classify when there is some degree of uncertainty. From the proposed method, a computationally viable algorithm for large volumes of data is also introduced. The experiments performed in a real scenario and a comparison of the results with the RSBM-based method demonstrate the efficiency of both the method and the algorithm in diverse contexts that involve large volumes of data. [ABSTRACT FROM AUTHOR]

Information systems (IS) can significantly support the organization of business processes. However, the proceeding digitalization of processes can also lead to an increasing organizational complexity and the need to more intensely investigate the adherence to external or internal compliance rules. Process-related data from IS and underlying process models can, however, also contribute to an effective compliance checking. This paper summarizes the motivation, the setup, the data set and the results of the 2019 MobIS-Challenge which was conducted as a workshop at WI 2019 in Siegen, Germany. Results submitted to the challenge are presented in detail and directions for future work are discussed. [ABSTRACT FROM AUTHOR]

With the development of modern remote sensing technology, remote sensing images have become one of the powerful tools for people to understand the Earth and its surroundings. However, there is currently no good classification algorithm that can accurately classify images. In order to accurately classify remote sensing images, this paper studies the content of the article by using fuzzy C-means clustering algorithm and radial basis neural network (RBF). The classification accuracy of SIRI-WHU dataset was analyzed by using the classification accuracy evaluation index such as overall accuracy and Kappa coefficient. The Kappa coefficient of vegetation classification in SIRI-WHU dataset was 0.9678, and the overall accuracy reached 97.18%. According to the classification problem of remote sensing image, according to the characteristics of remote sensing image, the improved model Alex Net-10-FCM is used to classify the remote sensing image dataset, and very high classification accuracy is obtained. [ABSTRACT FROM AUTHOR]

Traditional point cloud registration methods require large overlap between scans, which imposes strict constraints on data acquisition. To facilitate registration, users have to carefully position scanners to ensure sufficient overlap. In this article, we propose to use high-level structural information (i.e., plane/line features and their interrelationship) for registration, which is capable of registering point clouds with small overlap, allowing more freedom in data acquisition. We design a novel plane-/line-based descriptor dedicated to establishing structure-level correspondences between point clouds. Based on this descriptor, we propose a simple but effective registration algorithm. We also provide a data set of real-world scenes containing a larger number of scans with a wide range of overlap. Experiments and comparisons with state-of-the-art methods on various data sets reveal that our method is superior to existing techniques. Though the proposed algorithm outperforms state-of-the-art methods on the most challenging data set, the point cloud registration problem is still far from being solved, leaving significant room for improvement and future work. [ABSTRACT FROM AUTHOR]

Copyright of SDU Journal of Engineering Sciences & Design / Mühendislik Bilimleri ve Tasarım Dergisi is the property of Journal of Engineering Sciences & Design and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Parallel end-to-end driving aims to improve the performance of end-to-end driving models using both simulated- and real-world data. However, how to efficiently utilize the data from both the simulated world and the real world remains a difficult issue, since these data are usually not well aligned. In this article, we build a data set called the parallel end-to-end driving data set (PED) for parallel end-to-end driving research. PED consists of 13 000 images from the simulated world and 13 000 images from the real world that are used to train the model, as well as 2700 images from the real world that are used to test the model. The simulated-world data in PED are constructed according to the real world, and each simulated-world image corresponds to a real-world image. PED also contains the vehicle measurement data (GPS, speed, steering angle, and heading direction of the vehicle) related to both the simulated- and real-world images, which are not available in some other data sets. We conduct two types of experiments to illustrate the effectiveness and the superiority of PED and explore some ways to mix the simulated-world data with the real-world data to improve the performance of end-to-end driving models. [ABSTRACT FROM AUTHOR]

Internet of Things (IoT) network is the latest technology which is used to connect all the objects near us. Implementation of IoT technology is latest and growing day-by-day, it is coming with risk itself. So, it required the most efficient model to detect malicious activities as fast as possible and accurate. In our paper, we considered Deep Neural Network (DNN) for identifying the attacks in IoT. Intelligent intrusion detection system can only be built if there is availability of an effective dara set. Performance of DNN to correctly identify the attack has been evaluated on the most used data sets, i.e., KDD-Cup'99, NSL-KDD, and UNSW-NB15. Our experimental results showed the accuracy rate of the proposed method using DNN. It showed that accuracy rate is above 90% with each dataset. [ABSTRACT FROM AUTHOR]

This article presents an agricultural data set collected by Ladybird, an autonomous field robot designed at the Australian Centre for Field Robotics. The data set contains weekly scans of cauliflower and broccoli (Brassica oleracea) covering a 10 week growth cycle from transplant to harvest. The data set includes ground truth; physical characteristics of the crop; environmental data collected by a weather station and a soil sensor network; and scans of the crop performed by Ladybird, which include stereo color, thermal and hyperspectral imagery. The layout of the farm and data collection methodology are described. A description of Ladybird's capabilities and sensors are provided. Benchmark results are provided to illustrate the contents of the data set and how it can be processed. The hyperspectral data are compiled into hypercubes and the pixels are classified into crop, weed, or soil. An object detector is applied to the color imagery to locate the crop. Our intention in releasing the data set is to facilitate robotics and machine learning research activity in agriculture. The data set can be downloaded from https://doi.org/10.25910/5c941d0c8bccb. [ABSTRACT FROM AUTHOR]

Images differ in their memorability in consistent ways across observers. What makes an image memorable is not fully understood to date. Most of the current insight is in terms of high-level semantic aspects, related to the content. However, research still shows consistent differences within semantic categories, suggesting a role for factors at other levels of processing in the visual hierarchy. To aid investigations into this role as well as contributions to the understanding of image memorability more generally, we present MemCat. MemCat is a category-based image set, consisting of 10K images representing five broader, memorability-relevant categories (animal, food, landscape, sports, and vehicle) and further divided into subcategories (e.g., bear). They were sampled from existing source image sets that offer bounding box annotations or more detailed segmentation masks.Wecollected memorability scores for all 10Kimages, each score based on the responses of on average 99 participants in a repeat-detection memory task. Replicating previous research, the collected memorability scores show high levels of consistency across observers. Currently, MemCat is the second largest memorability image set and the largest offering a category-based structure. MemCat can be used to study the factors underlying the variability in image memorability, including the variability within semantic categories. In addition, it offers a new benchmark dataset for the automatic prediction of memorability scores (e.g., with convolutional neural networks). Finally, MemCat allows the study of neural and behavioral correlates of memorability while controlling for semantic category. [ABSTRACT FROM AUTHOR]

In the tropics and subtropics, most fires are set by humans for a wide range of purposes. The total amount of burned area and fire emissions reflects a complex interaction between climate, human activities, and ecosystem processes. Here we used satellite-derived data sets of active fire detections, burned area, precipitation, and the fraction of absorbed photosynthetically active radiation (fAPAR) during 1998–2006 to investigate this interaction. The total number of active fire detections and burned area was highest in areas that had intermediate levels of both net primary production (NPP; 500–1000 g C m−2 year−1) and precipitation (1000–2000 mm year−1), with limits imposed by the length of the fire season in wetter ecosystems and by fuel availability in drier ecosystems. For wet tropical forest ecosystems we developed a metric called the fire-driven deforestation potential (FDP) that integrated information about the length and intensity of the dry season. FDP partly explained the spatial and interannual pattern of fire-driven deforestation across tropical forest regions. This climate-fire link in combination with higher precipitation rates in the interior of the Amazon suggests that a negative feedback on fire-driven deforestation may exist as the deforestation front moves inward. In Africa, compared to the Amazon, a smaller fraction of the tropical forest area had FDP values sufficiently low to prevent fire use. Tropical forests in mainland Asia were highly vulnerable to fire, whereas forest areas in equatorial Asia had, on average, the lowest FDP values. FDP and active fire detections substantially increased in forests of equatorial Asia, however, during El Niño periods. In contrast to these wet ecosystems we found a positive relationship between precipitation, fAPAR, NPP, and active fire detections in arid ecosystems. This relationship was strongest in northern Australia and arid regions in Africa. Highest levels of fire activity were observed in savanna ecosystems that were limited neither by fuel nor by the length of the fire season. However, relations between annual precipitation or drought extent and active fire detections were often poor here, hinting at the important role of other factors, including land managers, in controlling spatial and temporal variability of fire.