Grid Cells Research Articles

The last of the maculineans: can we save the emblematic Alcon Blue butterfly Phengaris alcon under climate change when its habitat continues to deteriorate?

AbstractIn the NW-European region of Flanders (northern Belgium), wet heathlands harbour several threatened species among which the emblematic Alcon Blue butterfly Phengaris alcon is rapidly declining. Since 1990, the number of occupied grid cells in Flanders has decreased from 61 to only six in 2023, a decline of 90%. Due to the loss of many populations, the mean distance between populations tripled from 9 km in the period 1950–1990 to 27 km now. Site-wide egg counts in one of the strongholds of the Alcon Blue in Flanders revealed a strong increase after some initial intensive care conservation measures in the 2000s, but since 2013 a steep decline of 99% has been observed. The standardised egg counts in the remaining six Flemish populations (i.e., monitoring within the framework of the Natura2000 reporting) showed a yearly decrease of 18% since 2016. Despite increased conservation attention and targeted management since the end of the 1990s, the current situation of P. alcon in Belgium is more precarious than ever. Causes of this rapid decline are manifold and they are likely to be amplified by extreme weather conditions caused by climate change. More efficient intensive care measures and more intense cooperation across national borders are now quickly required to regionally preserve this habitat specialist with a unique eco-evolutionary profile.Implications for insect conservation By analysing the causes of the decline of the threatened Alcon Blue butterfly in Flanders, we suggest management and policy measures in an attempt to conserve this and other threatened heathland species in a highly anthropogenic region.

Relationship between radar reflectivity thresholds and very low frequency/low frequency total lightning for thunderstorm identification

AbstractTo improve our knowledge of the spatio‐temporal correspondence between radar reflectivity and VLF/LF total lightning data, this paper proposes indexes to quantify the horizontal distribution and temporal evolution characteristics between radar composite reflectivity images and total lightning density map, lightning clusters and radar cells with different radar reflectivity thresholds (30, 35, 40, 45 and 50 dBZ), respectively. It is found that the number of radar grid cells with radar reflectivity over 30 dBZ is ten times the number of lightning grid cells. At the identification stage, the lightning activity regions in a radar cell account for less than 30% using thresholds of 30, 35 and 40 dBZ, respectively, and the radar cell has more than one lightning cluster, which means that the mesoscale convective systems typically have more than one draft, and drafts are only small part of the convection. The majority of the centroid deviations is less than 10 km, indicating that there are some shifts between electrically active regions and convective regions. Results suggest that VLF/LF total lightning data are consistent with radar data and total lightning data can be used individually on a smaller spatio‐temporal scale than radar data.

Areas of Endemism and Biogeographic Regionalization of the Iberian Peninsula Based on Ants (Hymenoptera: Formicidae)

Based on the distribution of 275 species of ants native to the Iberian Peninsula (IP), we identified areas of endemism (AE) within its geographical limits and present a biogeographic regionalization using two complementary methods and two types of operational geographical units. For endemicity analysis (EA), we used a 100 km2 grid cell, and for parsimony analysis of endemicity (PAE), we used hydrological basins as natural units. The EA revealed twelve areas of endemism that were grouped into five consensus areas. These are the Northeastern area, South plateau, Guadalquivir Valley, Baetic System, and Iberian Peninsula (whole). PAE resulted in a cladogram that classified hydrological basins into at least two successively nested subsets: an Atlantic group that is more related to northern European fauna and an Iberian subset that is well supported by synapomorphies. The Iberian subset was differentiated into four main areas: (a) a Northeastern area formed by the Pyrenees and the Catalan Coastal Range, (b) a mainland area containing the Central System and Guadiana and Tajo valleys, (c) a Southern area consisting of the Guadalquivir Valley and the Baetic System, and (d) a Northern coastal area. The results showed congruence between the two methods since many of the synapomorphies are shared among the diagnostic and IP endemic species of the EA. Both EA and PAE showed the relevance of the heterogeneous peninsular orography that combines mountainous regions with valleys and plateaus, which have acted as historical barriers or corridors. The presence of numerous endemic species, particularly in the southern third of the IP, suggests that several Iberian refuges for ants originated during glacial periods. These areas constitute priority sites for the conservation of ants in particular and biodiversity in general on the IP and allow further research about the processes that generated these distributional patterns.

Image Data Stream Organization and Online Analysis Application Based on Data Cube Technology

Abstract. This study aims to explore the important role of data-like cube structures in modern remote sensing data processing and data analysis through ArcPy and Python multiprocessing techniques. A multi-scale spatial data cube is innovatively developed to improve the efficiency of remote sensing data management and optimize data analysis. The core of this study is to define and implement grid cells of different sizes that form the basis of data cube, and to quantify the efficient coverage of specific areas using Python multiprocessing techniques. Experiments were conducted in Hainan Province, and efficient data coverage of the whole Hainan Province was realized using the grid data method, which significantly reduced the amount of remote sensing data and processing time required. This shows that the method has successfully improving data coverage capacity and utilization efficiency. The results of this study not only demonstrate the effective application of data-like cubes in remote sensing data processing and analysis, but also provide new perspectives and methods for future complex spatial data analysis and large-scale remote sensing data processing.

Neuromorphic place cells

Abstract A neuromorphic SLAM system shows potential for more efficient implementation than its traditional counterpart. At the mean time a neuromorphic model of spatial encoding neurons in silicon could provide insights on the functionality and dynamic between each group of cells. Especially when realistic factors including variations and imperfections on the neural movement encoding are presented to challenge the existing hypothetical models for localization. We demonstrate a mixed-mode implementation for spatial encoding neurons including theta cells, egocentric place cells, and the typical allocentric place cells. Together, they form a biologically plausible network that could reproduce the localization functionality of place cells observed in rodents. The system consists of a theta chip with 128 theta cell units and an FPGA implementing 4 networks for egocentric place cells formation that provides the capability for tracking on a 11 by 11 place cell grid. Experimental results validate the robustness of our model when suffering from as much as 18% deviation, induced by parameter variations in analog circuits, from the mathematical model of theta cells. We provide a model for implementing dynamic neuromorphic SLAM systems for dynamic-scale mapping of cluttered environments, even when subject to significant errors in sensory measurements and real-time analog computation. We also suggest a robust approach for the network topology of spatial cells that can mitigate neural non-uniformity and provides a hypothesis for the function of grid cells and the existence of egocentric place cells.

Analysis of terrestrial water storage variations in South Korea using GRACE satellite and GLDAS data in Google Earth Engine

With the requirement of a macroscopic approach to understanding the relationship between water resources and hydrological phenomena, such as severe droughts under climate change, this study investigates South Korea’s long-term terrestrial water storage (TWS) using GRACE satellite data. A detailed analysis of water balance in TWS, based on the GLDAS model outputs in Google Earth Engine, reveals the following results: (1) TWS anomaly shows an averaged decrease of -33.5 mm yr−1 from 2003 to 2016. (2) Spatial shifts in TWS anomaly (-1.176 to -0.424 cm) unveil regional water storage dynamics, indicating negative temporal slope changes per grid cell (-0.393 to -0.225). (3) Contributions of precipitation to TWS are not always straightforward, due to runoff inefficiencies affecting water storage and groundwater. (4) Accessible water, integrating surface water and groundwater linked only to the shallow layer’s soil moisture, constrains deep groundwater accounting, emphasizing the need for local groundwater surveys in further research.

3D magnetotelluric inversion with arbitrary data orientation angles

The survey setup of 3D magnetotelluric inversion is typically chosen to best reflect the underlying geological structure. Typically, the individual observation azimuths are distinct from this preferred survey layout orientation; in fact, measured electric and magnetic fields are aligned with local geomagnetic North. When that happens, either data preprocessing (rotation) is required before the inversion may be performed, or the orientation of the modeling grid is aligned with the data orientations. None of these approaches is optimal, possibly resulting in a reduction of data quality, the loss of statistical accuracy of the error estimates, or in unnecessary grid cells, resulting in increased computation times. In situations when the data quality may be improved by allowing distinct individual site orientations, such as when cultural noise or faulty instrumentation cause a degradation of one of the modes, data loss due to rotation is especially detrimental. In this paper, we present a modified 3D magnetotelluric inversion algorithm, based on the ModEM software, that allows for direct interpretation of observations with arbitrary data orientation angles (ModEM ADORA). In ModEM ADORA, the modeling responses are rotated to the observation orientation angles after each forward calculation and before the data misfit is evaluated. This allows for direct interpretation of observation data with distinct orientations at each site and (as needed for some historical data) at each frequency. The derivative of the penalty functional is computed by pre-multiplying the derivatives of the data functionals (or the complete Jacobian matrix) by the relevant sparse block-diagonal rotation operators. We describe the principle and implementation of ModEM ADORA, and provide a collection of synthetic and real data tests for validation and an analysis of effectiveness of the ADORA option of ModEM. We show that by decoupling modeling grid and sensor orientations, ModEM ADORA supports both computational efficiency and ease of use, as well as allows for the maximum quality and quantity of inverted data.

A hierarchical model with hexagon grids for multi-objective route planning in large-scale off-road environments

Off-road route planning plays a vital role across various domains, including civil transportation, earthquake relief and civil production. Off-road route planning requires the consideration of a diverse range of environmental constraints, which is a considerable challenge compared with conventional route planning along existing networks. We aimed to present a multi-objective route planning method for off-road environments. Specifically, we designed a heuristic function incorporating multiple characteristics to generate reliable route planning solutions based on Pareto optimality. We also constructed a hierarchical hexagonal grid model by merging similar grid cells and reconstructing adjacency relationships with other grid cells to achieve high efficiency. The proposed method was compared with conventional route planning methods using single- and multi-objective approaches in traditional and hierarchical hexagonal grid models. The proposed method can effectively plan off-road routes, improve optimal route distance and trafficability, and provide more efficient route solutions. It reduced the planning time by 97.1% compared with multi-objective route planning based on a traditional hexagonal grid model.

A non-uniform lattice design method for lightweight structures in 3D printing

Lightweight design is an essential topic in aerospace, automotive and other fields. In automobile manufacturing, the engine connecting rod is one of the main components; its lightweight design has a high reference value. And 3D printing provides a feasible solution for designing and manufacturing lightweight structures. Unlike the traditional geometrically homogeneous point design, this study offers a non-homogeneous point design method based on the spatial stress state of additively manufactured components. After numerical simulation of quasi-static stresses on a model of an engine connecting rod, finite element grid cells with different stress values are replaced by lattice cells with different specific stiffnesses at similar local stress levels. The overall specific stiffness of the structure is further improved by continuing the optimized design with the corresponding gradient-type reinforcement of the non-uniform lattice structure. The basic idea of this design is to perform non-uniform adaptive filling of solid parts under localized loading by employing different types of unit cells. Stereolithography 3D printing technology prepares the engine lattice structural parts for quasi-static compression comparison experiments and fracture analysis after failure. The conclusions show that the engine connecting rod members with non-homogeneous lattice have more excellent overall mechanical properties than homogeneous lattice members. This work demonstrates the feasibility of such design methods for 3D printing lightweight structures and optimization.

Visual boundary cues suffice to anchor place and grid cells in virtual reality.

The hippocampal formation contains neurons responsive to an animal's current location and orientation, which together provide the organism with a neural map of space.1,2,3 Spatially tuned neurons rely on external landmark cues and internally generated movement information to estimate position.4,5 An important class of landmark cue are the boundaries delimiting an environment, which can define place cell field position6,7 and stabilize grid cell firing.8 However, the precise nature of the sensory information used to detect boundaries remains unknown. We used 2-dimensional virtual reality (VR)9 to show that visual cues from elevated walls surrounding the environment are both sufficient and necessary to stabilize place and grid cell responses in VR, when only visual and self-motion cues are available. By contrast, flat boundaries formed by the edges of a textured floor did not stabilize place and grid cells, indicating only specific forms of visual boundary stabilize hippocampal spatial firing. Unstable grid cells retain internally coherent, hexagonally arranged firing fields, but these fields "drift" with respect to the virtual environment over periods >5 s. Optic flow from a virtual floor does not slow drift dynamics, emphasizing the importance of boundary-related visual information. Surprisingly, place fields are more stable close to boundaries even with floor and wall cues removed, suggesting invisible boundaries are inferred using the motion of a discrete, separate cue (a beacon signaling reward location). Subsets of place cells show allocentric directional tuning toward the beacon, with strength of tuning correlating with place field stability when boundaries are removed.

Zoogeographical regions in the Atlantic Forest: patterns and potential drivers

AbstractAimTo delineate present‐day zoogeographical regions of terrestrial vertebrates (frogs, lizards, snakes, birds and non‐volant mammals) in the Atlantic Forest. Within each taxonomic group, we examine the relative importance of abrupt climatic transitions, orographic barriers, past climate change and rivers in shaping zoogeographical boundaries.LocationSouth America's Atlantic Forest.MethodsWe applied a network‐based method to delineate zoogeographical regions, using distribution data (range maps) for 455 species of frogs, 103 lizards, 220 snakes, 917 birds and 202 non‐volant mammals, in 50 × 50 km grid cells. We used hierarchical generalized linear mixed‐effects models to test environmental predictors associated with zoogeographical boundaries. Finally, we intersected the bioregion maps delineated for each group to identify general patterns across all vertebrates.ResultsWe identified four zoogeographical regions for birds and snakes, and five for frogs, lizards and non‐volant mammals. Depending on the group, contemporary and past climate conditions, elevation variation and/or rivers were associated with zoogeographical boundaries. The combined maps indicate that the Atlantic Forest retains four spatially cohesive zoogeographical regions based on present‐day distribution of vertebrates.Main conclusionsCross‐taxon congruence indicates that the geographical and environmental characteristics of the Atlantic Forest have a strong influence on the location of zoogeographical regions for vertebrates. In contrast, transition zones appear to be associated with the spatial distribution of life history traits of each group, potentially explaining the observed differences in the number of bioregions across groups and the position of zoogeographical boundaries. This work paves the way for further research into the evolutionary assembly of the Atlantic Forest's zoogeographical regions and may help inform conservation priorities for maintaining their distinctive faunas.

The Impacts of Changing Winter Warm Spells on Snow Ablation Over Western North America

AbstractAn increase in winter air temperature can amplify snowmelt and sublimation in mountain regions with implications to water resources and ecological systems. Winter Warm Spells (WWS) are defined as a winter period (December to February, DJF) of at least 3 consecutive days with daily maximum temperature anomaly above the 90th percentile (using a moving‐average of 15 days between 2001 and 2013). We calculate WWS for every 4‐km grid cell within an atmospheric model over western North America to characterize WWS and analyze snow ablation and their changes in a warmer climate. We find that days with ablation during WWS represent a small fraction of winter days (0.6 days), however, 49% of total winter ablation (33.4 mm/DJF) occurs during WWS. Greater extreme ablation rates (99th percentile) occur 18% more frequently during WWS than during non‐WWS days. Ablation rates during WWS in humid regions are larger (9 mm d−1) than in dry regions (7 mm d−1) in a warmer climate, which can be explained by differences in the energy balance and the snowpack's cold content. We find that warmer (0.8°C), longer (1.8 days) and more frequent (3.7 more events) WWS increase total winter ablation (on average 109% or 18 mm/DJF) in a warmer climate. Winter melt during WWS in warm and humid places is expected to increase about 3 times more than in the cold and dry region. This study provides a comprehensive description of WWS and their impact on snowpack dynamics, which is relevant to reservoir operations and water security.

Impact of Planetary Parameters on Water Clouds Microphysics

Potentially habitable exoplanets are targets of great interest for the James Webb Space Telescope and upcoming mission concepts such as the Habitable Worlds Observatory. Clouds strongly affect climate and habitability, but predicting their properties is difficult. In Global Climate Models (GCMs), especially those aiming at simulating Earth, cloud microphysics is often crudely approximated by assuming that all cloud particles have a single, constant size or a prescribed size distribution and that all clouds in a grid cell are identical. For exoplanets that range over a large phase space of planetary properties, this method could result in large errors. In this work, our goal is to determine how cloud microphysics on terrestrial exoplanets, whose condensable is mainly water vapor, depend on aerosol properties and planetary parameters such as surface pressure, surface gravity, and incident stellar radiation. We use the Community Aerosol and Radiation Model for Atmospheres as a 1D microphysical model to simulate the formation and evolution of clouds including the processes of nucleation, condensation, evaporation, coagulation, and vertical transfer. In these 1D idealized experiments, we find that the parameters that determine the macrophysical thermal structure of the atmospheres, including surface pressure and stellar flux, impact cloud radiative effect (CRE) most significantly. Parameters such as gravity and number density of aerosols working as cloud condensation nuclei affect the microphysical processes of cloud formation, including activation and vertical transfer. They also have a significant, though weaker effect on CRE. This work motivates the development of more accurate GCM cloud schemes and should aid in the interpretation of future observations.

Feedback of efficient shock acceleration on magnetic-field structure inside young type Ia supernova remnants

Abstract Using an effective adiabatic index $\gamma_{\mathrm{eff}}$ to mimic the feedback of efficient shock acceleration, we simulate the temporal evolution of a young type Ia supernova remnant (SNR) with two different background magnetic field (BMF) topologies: a uniform and a turbulent BMF. The density distribution and magnetic-field characteristics of our benchmark SNR are studied with two-dimensional cylindrical magnetohydrodynamic (MHD) simulations. When $\gamma_{\mathrm{eff}}$ is considered, we find that: (1) the two-shock structure shrinks and the downstream magnetic-field orientation is dominated by the Rayleigh-Taylor instability structures; (2) \del{for the uniform case, }there exists more quasi-radial magnetic fields inside the shocked region.\del{ While for the turbulent case, owing to the turbulence in BMF, the downstream magnetic-field orientation is time-dependent}; and 
(3) inside the intershock region, both the quasi-radial magnetic energy density and the total magnetic energy density are enhanced: in radial direction, with $\gamma_{\mathrm{eff}}=$1.1, they are amplified \add{about 10-26 times more than} those with $\gamma_{\mathrm{eff}}=$5/3. While in angular direction, the total magnetic energy densities could be amplified \add{about 350 times more than} those with $\gamma_{\mathrm{eff}}=$5/3, \add{and there are more grid cells within the intershock region where the magnetic energy density is amplified by a factor greater than 100.(3) inside the intershock region, both the quasi-radial magnetic energy density and the total magnetic energy density are enhanced: in radial direction, with $\gamma_{\mathrm{eff}}=$1.1, they are amplified about 6-9 times those with $\gamma_{\mathrm{eff}}=$5/3. While in angular direction, the total magnetic energy densities could be amplified more than 125 times those with $\gamma_{\mathrm{eff}}=$5/3. More grid cells within the intershock region manifest magnetic energy densities surpassing the background magnetic energy density by a factor $>$ 100.

Live Object Recognition Using YOLO

Abstract: Live object recognition refers to the real-time process of identifying and categorizing objects within a given visual input, such as images. This technology utilizes computer vision techniques and advanced algorithms to detect objects, determine their dimension, area and weight and often classify them into defined categories. Our system proposesR-CNN and YOLO to determine the dimensions of the objects in real time. YOLO takes a different approach by treating object detection as a single regression problem. A single neural network is trained to directly predict bounding boxes and class probabilities for multiple objects in an image. The input image is divided into a grid, and each grid cell is responsible for predicting the objects whose center fall within that cell. Live object recognition finds applications in various fields, including autonomous vehicles, surveillance systems, robotics,augmented reality, and more. By providing instantaneous and accurate insights into the surrounding environment, this technology contributes to enhanced decision-making,interaction, and automation across numerous domains. The objects which can be recognized are solid objects which we use daily such as electronic items, stationery items, culinary items and many more. Our system “Live Object Recognition using YOLO” is aimed to detect and determine the objects and dimensions in real time.

Groundwater decline has negatively affected the well-preserved amphibian community of Doñana National Park (SW Spain)

Abstract The success of amphibian conservation efforts is highly dependent on the preservation of amphibian breeding habitats. Doñana National Park contains an abundance of ponds with heterogeneous characteristics, which has historically favoured its amphibian community. However, most of the park’s ponds are groundwater dependent, and aquifer overexploitation outside the park is leading to shorter hydroperiods and even desiccation in the pond network. This problem has been exacerbated by the scarcity of rain over the last decade. In 2021-2022, we surveyed the occurrence of the park’s 11 amphibian species, with a view to comparing our findings with those from the last previous survey in 2003-2004. We mapped the species occurrence within the park for the two survey periods. While all 11 amphibian species could still be found in Doñana, their occurrence had decreased across the board. In 2003-2004, 6 species were present across more than 50% of the sampling area. In contrast, 18 years later, such was the case for only two species (Pelophylax perezi and Hyla meridionalis). Declines were greatest for Epidalea calamita, followed by Pelobates cultripes, Triturus pygmaeus, and Lissotriton boscai. The mean number of species per sampling unit (i.e., grid cell) dropped from 4.3 to 3.1. To preserve Doñana’s amphibian community, it is important to restore the park’s pond network, which implies reducing regional groundwater overexploitation.

Grid Cells in Cognition: Mechanisms and Function.

The activity patterns of grid cells form distinctively regular triangular lattices over the explored spatial environment and are largely invariant to visual stimuli, animal movement, and environment geometry. These neurons present numerous fascinating challenges to the curious (neuro)scientist: What are the circuit mechanisms responsible for creating spatially periodic activity patterns from the monotonic input-output responses of single neurons? How and why does the brain encode a local, nonperiodic variable-the allocentric position of the animal-with a periodic, nonlocal code? And, are grid cells truly specialized for spatial computations? Otherwise, what is their role in general cognition more broadly? We review efforts in uncovering the mechanisms and functional properties of grid cells, highlighting recent progress in the experimental validation of mechanistic grid cell models, and discuss the coding properties and functional advantages of the grid code as suggested by continuous attractor network models of grid cells.

Dynamic Occupancy Grid Map with Semantic Information Using Deep Learning-Based BEVFusion Method with Camera and LiDAR Fusion.

In the field of robotics and autonomous driving, dynamic occupancy grid maps (DOGMs) are typically used to represent the position and velocity information of objects. Although three-dimensional light detection and ranging (LiDAR) sensor-based DOGMs have been actively researched, they have limitations, as they cannot classify types of objects. Therefore, in this study, a deep learning-based camera-LiDAR sensor fusion technique is employed as input to DOGMs. Consequently, not only the position and velocity information of objects but also their class information can be updated, expanding the application areas of DOGMs. Moreover, unclassified LiDAR point measurements contribute to the formation of a map of the surrounding environment, improving the reliability of perception by registering objects that were not classified by deep learning. To achieve this, we developed update rules on the basis of the Dempster-Shafer evidence theory, incorporating class information and the uncertainty of objects occupying grid cells. Furthermore, we analyzed the accuracy of the velocity estimation using two update models. One assigns the occupancy probability only to the edges of the oriented bounding box, whereas the other assigns the occupancy probability to the entire area of the box. The performance of the developed perception technique is evaluated using the public nuScenes dataset. The developed DOGM with object class information will help autonomous vehicles to navigate in complex urban driving environments by providing them with rich information, such as the class and velocity of nearby obstacles.

Underprediction of extirpation and colonisation following climate and land‐use change using species distribution models

AbstractAimTo evaluate the performance of species distribution models in predicting observed colonisations, persistences and extirpations in response to changes in climate and land use over a multi‐decadal period.LocationSweden.MethodsWe use historical (early 20th century) land use and climate data to build species distribution models for 84 plant species across three provinces of Sweden. Model performance was then evaluated internally using a subset of the historical data for cross‐validation, as well as by using the models to project occurrences to the modern day and validating them with observed occurrences from 1990 to 2020. We then analysed predicted and observed occurrences in the modern period in terms of persistence, extirpation (local extinction) and colonisation in relation to species' habitat and climate associations.ResultsWe found overall high agreement between evaluation methods, although internal evaluation gave consistently higher values for model performance (using true skill statistic, TSS). Overall, extirpations were worst predicted, with on average fewer than one‐third of each species' extirpations being foreseen by the models. Colonisations were better predicted, while persistences were relatively well‐predicted. Predictive accuracy of colonisations was higher for species with relatively warmer temperature associations (climate‐driven expansion), while extirpations were better predicted in cool‐related species (retractions at cool edges). Colonisations of forest‐associated species were more common than predicted (underpredicted), despite widespread patterns of afforestation. Assessing grid‐cell level turnover, we found that in grid cells that experienced the largest changes in terms of climate and land use, predicted extirpations were less likely to have happened.Main ConclusionsWe found that commonly applied modelling approaches have limited ability to predict observed changes in species occurrences, especially extirpations. This suggests that we should take predictions of future biodiversity loss very seriously. However, the ability for species to (at least temporarily) persist in unsuitable conditions could be an opportunity for biodiversity conservation.

Uncertainty propagation analysis for distributed hydrological forecasting using a neural network

AbstractThe last few decades have presented a significant increase in hydrological disasters, such as floods. In some countries, most of the environmental, socioeconomic, and biodiversity losses are caused by floods. Thus, flood forecasting is crucial to support an efficient disaster warning system. This work proposes a model for hydrological forecasting based on a neural network with a geographically aligned input named GeoNN. It employs weather radar data to obtain accumulated rainfall in each grid cell of the watershed and make 15‐ and 120‐min predictions of the outlet river level. An uncertainty propagation analysis was performed for GeoNN from a collection of test cases obtained by either using different schemes of the dataset partitioning or introducing different additive‐noise rates to the input data to provide a probability of flood occurrence and also an ensemble prediction. Both this probability and the ensemble were able to detect occurrences of river levels exceeding a given flood threshold.