Multiple Temporal Resolutions Research Articles

A New Hybrid Model Based on SCINet and LSTM for Short-Term Power Load Forecasting

A stable and reliable power system is crucial for human daily lives and economic stability. Power load forecasting is the foundation of dynamically balancing between the power supply and demand sides. However, with the popularity of renewable energy sources and electric vehicles, it still struggles to achieve accurate power load forecasting due to the complex patterns and dynamics of load data. To mitigate these issues, this paper proposes a new hybrid model based on a sample convolution and integration network (SCINet) and a long short-term memory network (LSTM) for short-term power load forecasting. Specifically, a feed-forward network (FFN) is first used to enhance the nonlinear representation of the load data to highlight the complex temporal dynamics. The SCINet is then employed to iteratively extract and exchange information about load data at multiple temporal resolutions, capturing the long-term dependencies hidden in the deeper layers. Finally, the LSTM networks are performed to further strengthen the extraction of temporal dependencies. The principal contributions of the proposed model can be summarized as follows: (1) The SCINet with binary tree structure effectively extracts both local and global features, proving advantageous for capturing complex temporal patterns and dynamics; (2) Integrating LSTM into the SCINet-based framework mitigates information loss resulting from interactive downsampling, thereby enhancing the extraction of temporal dependencies; and (3) FNN layers are strategically designed to enhance the nonlinear representations prior to feeding the load data fed into the SCINet and LSTM. Three real-world datasets are used to validate the effectiveness and generalization of the proposed model. Experimental results show that the proposed model has superior performance in terms of evaluation metrics compared with other baseline models.

Transgression–regression cycles drive correlations in Ediacaran–Cambrian rock and fossil records

AbstractStrata of the Ediacaran Period (635–538.8 Ma) yield the oldest known fossils of complex, macroscopic organisms in the geologic record. These “Ediacaran-type” macrofossils (known as the Ediacaran biota) first appear in mid-Ediacaran strata, experience an apparent decline through the terminal Ediacaran, and directly precede the Cambrian (538.8–485.4 Ma) radiation of animals. Existing hypotheses for the origin and demise of the Ediacaran biota include: changing oceanic redox states, biotic replacement by succeeding Cambrian-type fauna, and mass extinction driven by environmental change. Few studies frame trends in Ediacaran and Cambrian macroevolution from the perspective of the sedimentary rock record, despite well-documented Phanerozoic covariation of macroevolutionary patterns and sedimentary rock quantity. Here we present a quantitative analysis of North American Ediacaran–Cambrian rock and fossil records from Macrostrat and the Paleobiology Database. Marine sedimentary rock quantity increases nearly monotonically and by more than a factor of five from the latest Ediacaran to the late Cambrian. Ediacaran–Cambrian fossil quantities exhibit a comparable trajectory and have strong (rs > 0.8) positive correlations with marine sedimentary area and volume flux at multiple temporal resolutions. Even so, Ediacaran fossil quantities are dramatically reduced in comparison to the Cambrian when normalized by the quantity of preserved marine rock. Although aspects of these results are consistent with the expectations of a simple fossil preservation–induced sampling bias, together they suggest that transgression–regression and a large expansion of marine shelf environments coincided with the diversification of animals during a dramatic transition that is starkly evident in both the sedimentary rock and fossil records.

Self-Supervised Motion Perception for Spatiotemporal Representation Learning.

In this study, we propose a novel pretext task and a self-supervised motion perception (SMP) method for spatiotemporal representation learning. The pretext task is defined as video playback rate perception, which utilizes temporal dilated sampling to augment video clips to multiple duplicates of different temporal resolutions. The SMP method is built upon discriminative and generative motion perception models, which capture representations related to motion dynamics and appearance from video clips of multiple temporal resolutions in a collaborative fashion. To enhance the collaboration, we further propose difference and convolution motion attention (MA), which drives the generative model focusing on motion-related appearance, and leverage multiple granularity perception (MG) to extract accurate motion dynamics. Extensive experiments demonstrate SMP's effectiveness for video motion perception and state-of-the-art performance of self-supervised representation models upon target tasks, including action recognition and video retrieval. Code for SMP is available at github.com/yuanyao366/SMP.

Mapping Forest Abrupt Disturbance Events in Southeastern China—Comparisons and Tradeoffs of Landsat Time Series Analysis Algorithms

Forest change monitoring is a fundamental and routine task for forest survey and planning departments, and the resulting forest change information acts as an underlying asset for sustainable forest management strategy development, ecological quality assessment, and carbon cycle research. The traditional ground-based manual monitoring of forest change has the disadvantages of high time and labor costs, low accessibility, and poor timeliness over wide regions. Remote sensing technology has become a popular approach for multi-scale forest change monitoring due to its multiple available spatial, spectral, temporal, and radiometric resolutions and wide coverage. Particularly, the free access policy of long time series archive data of Landsat (around 50 years) has triggered many automated analysis algorithms for landscape-scale forest change analysis, such as VCT, LandTrendr, BFAST, and CCDC. These automated algorithms differ in their principles, parameter settings, execution complexity, and disturbance types to be detected. Thus, selecting a suitable algorithm to satisfy the particular forest management demands is an urgent and challenging task for forest managers in a given forested area. In this study, Lishui City, Zhejiang Province, a typical plantation forest region in Southern China where forest disturbance widely and frequently exists, was selected as the study area. Based on the GEE platform, the algorithmic adaptability of VCT, LandTrendr, and CCDC in monitoring abrupt forest disturbance events was compared and assessed. The results showed that the kappa coefficients of the abrupt disturbance events detected by the three algorithms were at 0.704 (LandTrendr), 0.660 (VCT), and 0.727 (CCDC), and the corresponding overall accuracies were at 0.852, 0.830, and 0.862, respectively. The validated disturbance occurrence time consistency reached nearly 80% for the three algorithms. In light of the characteristics of forest disturbance occurrence in southeastern China as well as the algorithmic complexity, efficiency, and adaptability, LandTrendr was recommended as the most suitable one in this region or other similar regions. Overall, the forest change monitoring process based on GEE is becoming more simplified and easily implemented, and the comparisons and tradeoffs in this study provide a reference for the choice of long time series forest monitoring algorithms.

Lean and interpretable digital twins for building energy monitoring – A case study with smart thermostatic radiator valves and gas absorption heat pumps

The transition to low carbon energy systems poses challenges in terms of energy efficiency. In building refurbishment projects, efficient technologies such as smart controls and heat pumps are increasingly being used as a substitute for conventional technologies with the aim of reducing carbon emissions and determining operational energy and cost savings, together with other benefits. Measured building performance, however, often reveals a significant gap between the predicted energy use (design stage) and actual energy use (operation stage). For this reason, lean and interpretable digital twins are needed for building energy monitoring aimed at persistence of savings and continuous performance improvement. In this research, interpretable regression models are built with data at multiple temporal resolutions (monthly, daily and hourly) and seamlessly integrated with the goal of verifying the performance improvements due to Smart thermostatic radiator valves (TRVs) and gas absorption heat pumps (GAHPs) as well as giving insights on the performance of the building as a whole. Further, as part of modelling research, time of week and temperature (TOWT) approach is reformulated and benchmarked against its original implementation. The case study chosen is Hale Court sheltered housing, located in the city of Portsmouth (UK). This building has been used for the field-testing of innovative technologies such as TRVs and GAHPs within the EU Horizon 2020 project THERMOSS. The results obtained are used to illustrate possible extensions of the use of energy signature modelling, highlighting implications for energy management and innovative building technologies development.

Evaluation of effectiveness in conversations between humans and chatbots using parallel convolutional neural networks with multiple temporal resolutions

Chatbots enable the automation of several components in customer service and allow the support of multiple users. Despite their multiple advantages, due to the large amount of conversations generated by a chatbot, it is difficult to determine whether customer requests are well-addressed. For practical reasons, chatbot’s effectiveness is evaluated manually based upon a small sample (randomly chosen) of conversations or through self-reported user satisfaction. This procedure does not guarantee the correct evaluation of the service because the sample is generally not large enough and self-reports might be influenced by different external factors not directly associated to the chatbot’s functioning. This study proposes a methodology for automatic evaluation of chatbot effectiveness in real production environments. The analysis considers convolutional neural networks adapted for natural language processing, using two parallel convolutional layers to evaluate questions and answers independently. The proposed model also incorporates filters to extract features with multiple temporal resolution. This methodology is tested upon real conversations of chatbots that provide service to two different companies. The results are compared to baseline models based on classical techniques with different pre-trained word embedding models. According to our results, the proposed approach provides accuracies between 78.95% and 80.18%, which outperforms the best result of the baseline models by 2.9%.

Hierarchically nested networks optimize the analysis of audiovisual speech.

In conversational settings, seeing the speaker's face elicits internal predictions about the upcoming acoustic utterance. Understanding how the listener's cortical dynamics tune to the temporal statistics of audiovisual (AV) speech is thus essential. Using magnetoencephalography, we explored how large-scale frequency-specific dynamics of human brain activity adapt to AV speech delays. First, we show that the amplitude of phase-locked responses parametrically decreases with natural AV speech synchrony, a pattern that is consistent with predictive coding. Second, we show that the temporal statistics of AV speech affect large-scale oscillatory networks at multiple spatial and temporal resolutions. We demonstrate a spatial nestedness of oscillatory networks during the processing of AV speech: these oscillatory hierarchies are such that high-frequency activity (beta, gamma) is contingent on the phase response of low-frequency (delta, theta) networks. Our findings suggest that the endogenous temporal multiplexing of speech processing confers adaptability within the temporal regimes that are essential for speech comprehension.

A novel forecast framework for unsteady flows based on a convolutional neural network

Fluid mechanics is an important area where deep learning produces excellent results and can bring about scientific innovation because of its high dimensionality, significant nonlinearity, and ability to process an enormous amount of data. Deep learning technology is currently being used to study fluid mechanics, and its application potential is gradually being demonstrated. We propose a novel multi-resolution convolutional interaction network (MCIN), a hierarchical forecast framework based on a convolutional neural network. This structure can capture temporal dependencies at multiple temporal resolutions to enhance the forecasting performance of the original time series. The high-dimensional data of the flow around a cylinder are projected into a low-dimensional subspace using a variational autoencoder (VAE) as a nonlinear order-reduction technique. Then, the data of the subspace are used as the input to MCIN to forecast future velocity fields. The proposed MCIN is compared to non-intrusive reduced-order models based on dynamic mode decomposition and long short-term memory, combined with a VAE. The results demonstrate that MCIN has superior stability to other models in forecasting the evolution of complicated fluid flows and has the potential to forecast a greater number of future outcomes.

Forest Biodiversity Monitoring Based on Remotely Sensed Spectral Diversity—A Review

Forests are essential for global environmental well-being because of their rich provision of ecosystem services and regulating factors. Global forests are under increasing pressure from climate change, resource extraction, and anthropologically-driven disturbances. The results are dramatic losses of habitats accompanied with the reduction of species diversity. There is the urgent need for forest biodiversity monitoring comprising analysis on α, β, and γ scale to identify hotspots of biodiversity. Remote sensing enables large-scale monitoring at multiple spatial and temporal resolutions. Concepts of remotely sensed spectral diversity have been identified as promising methodologies for the consistent and multi-temporal analysis of forest biodiversity. This review provides a first time focus on the three spectral diversity concepts “vegetation indices”, “spectral information content”, and “spectral species” for forest biodiversity monitoring based on airborne and spaceborne remote sensing. In addition, the reviewed articles are analyzed regarding the spatiotemporal distribution, remote sensing sensors, temporal scales and thematic foci. We identify multispectral sensors as primary data source which underlines the focus on optical diversity as a proxy for forest biodiversity. Moreover, there is a general conceptual focus on the analysis of spectral information content. In recent years, the spectral species concept has raised attention and has been applied to Sentinel-2 and MODIS data for the analysis from local spectral species to global spectral communities. Novel remote sensing processing capacities and the provision of complementary remote sensing data sets offer great potentials for large-scale biodiversity monitoring in the future.

A Differentiable Hydrology Approach for Modeling With Time‐Varying Parameters

AbstractIn the analysis of observational data with mathematical models, it is often desirable to let one or more model parameters vary with time to account for changing environmental conditions or to offer more flexibility. A standard approach for dynamic models is to allow for the number of parameters to grow as a function of time, leading to a high‐dimensional inference problem for long data records. We propose using Hamiltonian Monte Carlo, a gradient‐based Markov chain Monte Carlo (MCMC) method, to perform Bayesian inference for time‐varying parameters (TVP) in hydrology models. As derivatives of model error functions with regard to parameters are not available in closed form, we implemented the GR4J rainfall‐runoff models in Theano/PyMC3, and Jax/NumPyro, allowing for the application of automatic differentiation techniques. A simulation experiment assessing the viability of these methods for recovering underlying temporal variation in parameters indicates that a discrete Gaussian random walk‐based prior appears to be best suited across a range of scenarios while the continuous‐time Gaussian process compares relatively poorly. In an analysis of 20 years of daily real‐world streamflow records from the Model Parameter Estimation Experiment (MOPEX) hydrology data set with increasing data sparsity and at multiple temporal resolutions, we find that TVP inference with gradient‐based MCMC is a flexible and reliable approach for analyzing hydrology models with dynamic parameter sets.

End-to-End Train Horn Detection for Railway Transit Safety.

The train horn sound is an active audible warning signal used for warning commuters and railway employees of the oncoming train(s), assuring a smooth operation and traffic safety, especially at barrier-free crossings. This work studies deep learning-based approaches to develop a system providing the early detection of train arrival based on the recognition of train horn sounds from the traffic soundscape. A custom dataset of train horn sounds, car horn sounds, and traffic noises is developed to conduct experiments and analysis. We propose a novel two-stream end-to-end CNN model (i.e., THD-RawNet), which combines two approaches of feature extraction from raw audio waveforms, for audio classification in train horn detection (THD). Besides a stream with a sequential one-dimensional CNN (1D-CNN) as in existing sound classification works, we propose to utilize multiple 1D-CNN branches to process raw waves in different temporal resolutions to extract an image-like representation for the 2D-CNN classification part. Our experiment results and comparative analysis have proved the effectiveness of the proposed two-stream network and the method of combining features extracted in multiple temporal resolutions. The THD-RawNet obtained better accuracies and robustness compared to those of baseline models trained on either raw audio or handcrafted features, in which at the input size of one second the network yielded an accuracy of 95.11% for testing data in normal traffic conditions and remained above a 93% accuracy for the considerable noisy condition of-10 dB SNR. The proposed THD system can be integrated into the smart railway crossing systems, private cars, and self-driving cars to improve railway transit safety.

Level-Crossing Sampling with Multiple Temporal Resolutions for Speech Signals

Level-crossing sampling (LCS) is an alternative to the uniform sampling method and is suitable for sparse and bursty signals. LCS, which is adaptive to amplitude variations of the signals, mainly focuses on reducing the total number of samples. Presently, speech signals are sampled with LCS methods that have a constant temporal resolution, and they require time-to-digital converters (TDC) with a wide dynamic range. A long sequence of overflow counts of the TDC counter may be generated during silence regions if the dynamic range is reduced. This paper proposes level-crossing samplers with multiple temporal resolutions for speech signals. Frequency scaling is applied to standard LCS, adaptive LCS and peak sampling. During the silence region, the frequency of the TDC clock is scaled down, which enables the representation of longer intervals with smaller words. The number of bits per sample and hence the total number of samples generated are decreased due to the reduction in the dynamic range of the TDC. Simulation results indicate that the proposed method outperforms the conventional level-crossing samplers, which do not employ frequency scaling. In addition to a significant reduction in data size, this method can also be used for real-time automatic detection of silence regions in a speech signal.

EVis: Visually Analyzing Environmentally Driven Events.

Earth scientists are increasingly employing time series data with multiple dimensions and high temporal resolution to study the impacts of climate and environmental changes on Earth's atmosphere, biosphere, hydrosphere, and lithosphere. However, the large number of variables and varying time scales of antecedent conditions contributing to natural phenomena hinder scientists from completing more than the most basic analyses. In this paper, we present EVis (Environmental Visualization), a new visual analytics prototype to help scientists analyze and explore recurring environmental events (e.g. rock fracture, landslides, heat waves, floods) and their relationships with high dimensional time series of continuous numeric environmental variables, such as ambient temperature and precipitation. EVis provides coordinated scatterplots, heatmaps, histograms, and RadViz for foundational analyses. These features allow users to interactively examine relationships between events and one, two, three, or more environmental variables. EVis also provides a novel visual analytics approach to allowing users to discover temporally lagging relationships related to antecedent conditions between events and multiple variables, a critical task in Earth sciences. In particular, this latter approach projects multivariate time series onto trajectories in a 2D space using RadViz, and clusters the trajectories for temporal pattern discovery. Our case studies with rock cracking data and interviews with domain experts from a range of sub-disciplines within Earth sciences illustrate the extensive applicability and usefulness of EVis.

A Double-Layer Multi-Resolution Classification Model for Decoding Spatiotemporal Patterns of Spikes With Small Sample Size.

We build a double-layer, multiple temporal-resolution classification model for decoding single-trial spatiotemporal patterns of spikes. The model takes spiking activities as input signals and binary behavioral or cognitive variables as output signals and represents the input-output mapping with a double-layer ensemble classifier. In the first layer, to solve the underdetermined problem caused by the small sample size and the very high dimensionality of input signals, B-spline functional expansion and L1-regularized logistic classifiers are used to reduce dimensionality and yield sparse model estimations. A wide range of temporal resolutions of neural features is included by using a large number of classifiers with different numbers of B-spline knots. Each classifier serves as a base learner to classify spatiotemporal patterns into the probability of the output label with a single temporal resolution. A bootstrap aggregating strategy is used to reduce the estimation variances of these classifiers. In the second layer, another L1-regularized logistic classifier takes outputs of first-layer classifiers as inputs to generate the final output predictions. This classifier serves as a meta-learner that fuses multiple temporal resolutions to classify spatiotemporal patterns of spikes into binary output labels. We test this decoding model with both synthetic and experimental data recorded from rats and human subjects performing memory-dependent behavioral tasks. Results show that this method can effectively avoid overfitting and yield accurate prediction of output labels with small sample size. The double-layer, multi-resolution classifier consistently outperforms the best single-layer, single-resolution classifier by extracting and utilizing multi-resolution spatiotemporal features of spike patterns in the classification.

Behavior of Overdeformed Shield Tunnel Lining under Grouting Treatment: Field Experiment

Shield tunnel structures are commonly found to be overdeformed in Eastern China soft ground due to adjacent engineering activities. Grouting treatment could serve as a viable solution to actively reduce the excessive diametrical expansion of tunnel linings. However, understanding the grouting-induced effects on adjacent tunnel structures remains preliminary due to the scarcity of well-documented case histories. In this study, a field experiment was conducted to explore a shield metro tunnel’s real performance during grouting treatments. An extensive monitoring scheme was deployed while varied geometrical grouting arrangements were carried out. The tests provided a comprehensive database of the tunnel responses in multiple temporal resolution levels, both during the entire course of the grouting treatment and within single grouting operations. Investigations are carried out over the influencing factors of the grouting efficiency. The experimental observations suggest that grouting efficiency tends to evolve with time and can be affected by various geometrical grouting parameters. Observations are also presented and discussed in detail for the influence of grouting history on the convergence distribution and the tunnel’s performance after the termination of all grouting operations. Time-variant behaviors of the tunnel are depicted in detail, such as the convergence recovery and rebound during the grouting process. Overall, this case study brings new insights into understanding the grouting-induced tunnel lining response. It can also serve as a benchmark for further theoretical and numerical studies.

Drones and Geography: Who Is Using Them and Why?

Drones have equipped geographers with the capacity to collect high-quality geospatial data at multiple spatial, spectral, and temporal resolutions. Although the adoption of drones is increasing across geography, knowledge of those using this technology and their practices is limited. The purpose of this article is to understand who is using drones in geography, how they are using them, and what future opportunities exist. We collected data from eighty-eight survey respondents, predominantly based in the United States but a handful from Australia, Canada, the European Union, and the United Kingdom. The findings from our Web-based survey show that about 85 percent of geographers using drones are White. Female respondents made up only about 30 percent of respondents, although they represented about 75 percent of the eighteen to twenty-four age group. Although the word drone has a negative connotation, most users (∼38 percent) prefer it, followed by unmanned aerial vehicle (∼21 percent) and unmanned aerial systems (∼19 percent). Only 22 percent of geographers have more than six years of drone experience, suggesting the rapid growth in use and popularity among geographers. Off-the-shelf drones are the most desirable, perhaps due to their low cost and ease of use. Overall, drones in geography are considered positive and have introduced a new era of small extent geospatial analyses.

Next Generation Gravity Mission Elements of the Mass Change and Geoscience International Constellation: From Orbit Selection to Instrument and Mission Design

ESA’s Next Generation Gravity Mission (NGGM) is a candidate Mission of Opportunity for ESA–NASA cooperation in the frame of the Mass Change and Geosciences International Constellation (MAGIC). The mission aims at enabling long-term monitoring of the temporal variations of Earth’s gravity field at relatively high temporal (down to 3 days) and increased spatial resolutions (up to 100 km) at longer time intervals. This implies also that time series of GRACE and GRACE-FO can be extended towards a climate series. Such variations carry information about mass change induced by the water cycle and the related mass exchange among atmosphere, oceans, cryosphere, land and solid Earth and will complete our picture of global and climate change. The main observable is the variation of the distance between two satellites measured by a ranging instrument. This is complemented by accelerometers that measure the nongravitational accelerations, which need to be reduced from ranging measurements to obtain the gravity signal. The preferred satellite constellation comprises one satellite pair in a near-polar and another in an inclined circular orbit. The paper focuses on the orbit selection methods for optimizing the spatial sampling for multiple temporal resolutions and then on the methodology for deriving the engineering requirements for the space segment, together with a discussion on the main mission parameters.

Estimating Evapotranspiration of Mediterranean Oak Savanna at Multiple Temporal and Spatial Resolutions. Implications for Water Resources Management

Mediterranean oak savanna is composed of a mixture of scattered oak trees, crops, pasture, and shrubs. It is the most widespread agroforestry landscape in Europe, and its conservation faces multiple threats including water scarcity, which has been exacerbated by global warming and greater climate variability. Evapotranspiration (ET) can be used as a proxy of the vegetation water status and response to water shortage conditions, providing relevant information about the ecosystem stability and its hydrological dynamics. This study evaluates a framework to estimate ET at multiple spatial and temporal scales and applies it to the monitoring of the oak savanna vegetation water consumption for the years 2013–2015. We used a remote sensing-based energy balance model (ALEXI/DisALEXI approach), and the STARFM data fusion technique to provide daily ET estimates at 30 m resolution. The results showed that modeled energy balance components compared well to ground measurements collected by an eddy covariance system, with root mean square error (RMSE) values ranging between 0.60 and 2.18 MJ m−2 d−1, depending on the sensor dataset (MODIS or Landsat) and the flux. The daily 30 m ET series generated by STARFM presented an RMSE value of 0.67 mm d−1, which yielded a slight improvement compared to using MODIS resolution or more simple interpolation approaches with Landsat. However, the major advantage of the high spatio-temporal resolution was found in the analysis of ET dynamics over different vegetation patches that shape the landscape structure and create different microclimates. Fine-scale ET maps (30 m, daily) provide key information difficult to detect at a coarser spatial resolution over heterogeneous landscapes and may assist management decisions at the field and farm scale.

An Explicit Scheme to Represent the Bidirectional Hydrologic Exchanges Between the Vadose Zone, Phreatic Aquifer, and River

AbstractUnderstanding groundwater storage variations as a result of effects from both the vadose zone and the river is of critical importance in the hydraulically connected surface‐subsurface system. In this study, we present a novel Bidirectional Exchange Scheme in Surface and Subsurface (BE3S) that represents bidirectional exchanges between the vadose zone, phreatic aquifer, and river. The approach enables explicit representations of each flow regime while conserving mass, and successfully yields solutions of the coupled system for multiple temporal resolutions. We test the scheme by comparing the BE3S‐derived outputs against other models and corresponding observational data. We apply the scheme to simulate hydrologic states in a reach of the Brazos River in Southeast Texas, such as soil moisture content, groundwater level, and river stage as well as net subsurface discharge fluxes. Good agreements between the simulated and observed data for all the components show the suitability of the proposed scheme in modeling the bidirectional flows and exchanges. We also assess how the bidirectional hydrologic exchanges are affected by adjacent flow domains. We find that vertical percolation is significantly affected by unsaturated soil thickness, resulting in the spatial variability of vertical percolation across sloping topography. Hilltop‐to‐valley convergent groundwater flow is also found to impede vertical percolation in river valleys due to shallow water table while facilitating percolation in the hilltops. The capability of the presented scheme that accounts for the topographically driven lateral groundwater flow and drainage dynamics provides the potential to enhance the representation of the surface‐subsurface system in Earth System Models (ESMs).

Influences of environmental heterogeneity on amphibian composition at breeding sites in a semiarid region of Mexico

Identification of factors that determine the behavior and diversity of species is crucial to our understanding of the ecology of species and communities. Satellite remote sensing (SRS) provides information relevant to the study of biodiversity and associated environmental traits at multiple spatial and temporal resolutions. Pond-breeding amphibians allow for comparative ecological analysis regarding the influence of environmental heterogeneity derived from SRS. Here, we characterize the influence of heterogeneity in abiotic (TWI, LST) and habitat-related (NDVI, NDWI) factors derived at fine spatial resolution on reproductive activity in the anuran community of the South Altiplano semiarid biogeographic province in Mexico. We used anuran reproductive data from seasonal and permanent water bodies and tested for the influence of abiotic and habitat heterogeneity with a spatiotemporal perspective using multivariate analysis, linear mixed-effects models (LMMs) and ecological niche models (ENMs). The results suggest that habitat heterogeneity best explains reproductive activity, whereas abiotic heterogeneity was second in importance. Implementation of ENMs using these abiotic and habitat heterogeneity variables allowed the detection of areas highly suitable for breeding across the study region. This research underlines the importance of environmental heterogeneity for predicting anuran breeding areas and their species compositions in the semiarid South Altiplano region of Mexico.