Manitoba Hydro

• Significant transparency and responsible modelling are vital for high-stakes AI forecasting. • The reported forecast accuracy of Google’s AI flood model is substantially overstated when evaluated against real observational data. • Google’s AI flood model relies on several subjective and questionable evaluation choices. The development of AI models is increasing at a rapid rate. However, when are they ready to be deployed in real-world operational settings? In this paper, we introduce a framework to support such assessments and apply it to Google’s recently released AI-based flood prediction system, which is claimed to achieve “reliability in predicting extreme riverine events” and provide “accurate and timely warnings” that are available “earlier and over larger and more impactful events in ungauged basins”. The system has been integrated into an operational early-warning platform producing open, real-time forecasts in more than 80 countries. While this development promises to usher in a new and exciting age in global flood forecasting, the supporting evidence relies heavily on several subjective choices, the implications of which have not been acknowledged or assessed. Here, we evaluate the consequences of these choices on claims of operational deployment readiness across four dimensions: predictive accuracy, forecast timeliness, the characterization of extreme events, and benchmarking against state-of-the-art models. Our assessment reveals that the system’s actual predictive accuracy is likely to be substantially lower than reported—particularly for extreme events—raising concerns about responsible practices across modelling and publicity in high-stakes applications. The deployment of the Google AI model therefore risks misinforming those who depend on its outputs for evacuation and preparedness decisions, particularly in less-developed countries such as those targeted by the enterprise, given its alarmingly high (>90%) rates of false positives and false negatives. Beyond the immediate operational consequences, if left unaddressed, these outcomes may erode public trust in AI within hydrological sciences. We conclude by calling for greater transparency, accountability, and methodological rigor in the integration of AI into flood forecasting.

ABSTRACT Ethiopia's Awash River Basin (ARB) data scarcity and quality concerns limit effective planning and research. This study evaluated 15 streamflow gauging stations through a two-week field inspection following World Meteorological Organization (WMO) protocols, combined with observer feedback and four statistical homogeneity tests. This study also conducts analysis of streamflow trends using daily data from 15 gauging stations over the period of 1965–2015 using Mann-Kendall test and Sen's slope estimator. Field assessments revealed outdated equipment, inadequate site conditions, and low observer satisfaction, often leading to errors in water level measurement. Homogeneity analysis showed that approximately 25, 40, and 33% of the stations in the Upper, Middle, and Lower Awash Basins, respectively, exhibit inhomogeneous data records, undermining long-term hydrological analyses. The study found a statistically significant increasing trend in annual streamflow in most middle Awash basin, while upper and lower basins showed insignificant trends.These findings highlight significant spatial variability in data reliability across the basin. The study concludes that upgrading gauging networks with telemetry, improving rating curve updates, and enhancing observer support are urgent priorities. Strengthening institutional coordination and capacity building will be critical to ensure reliable streamflow records, thereby improving hydrological forecasting and sustainable basin-wide water resources management.

U-Pb-Lu-Hf-trace element analysis of detrital zircon from two metasedimentary rocks of the Armstrong assemblage in the northern Pikwitonei granulite domain of the NW Superior Province (Manitoba, Canada) are used to explore the Eo-Paleoarchean crustal formation history and zircon petrogenetic processes of the proto-Superior craton. The zircon age spectrum from each sample spans a billion-year period from 3.85 to 2.65 Ga, with a predominant 3.85–3.40 Ga detrital zircon age population. The ca. 3.85 Ga detrital zircon grains from each sample have subchondritic εHf(t) values between −3.9 and − 1.0, consistent with initial crystallization from a suite of late-Hadean to early-Eoarchean protoliths. Semi-continuous Eo-Paleoarchean crustal reworking is subsequently inferred based on the single uniform Hf isotope evolution trend for 3.85–3.20 Ga detrital zircon grains, which is comparable to igneous zircon populations from the Nuvvuagittuq greenstone belt on the eastern shore of Hudson Bay in Quebec. Trace element analysis of detrital zircon grains from both metasedimentary samples indicate the limited presence of detrital metamorphic zircon at 3.67 Ga, 3.5 Ga and 3.4 Ga, which reveal potential metamorphic events within the Eo-Paleoarchean Superior Province. In addition, the detrital igneous zircon grains have trace element compositions consistent with either partial melting of hydrous source rocks or magmatic reworking/assimilation of felsic rocks. The trace element compositions are broadly similar to ca. 3.9–3.2 Ga igneous zircon populations from Acasta and Saglek regions, while demonstrating the importance of calculating initial U concentrations for Hadean – Archean zircon. • Pikwitonei domain and Nuvvuagittuq belt share a similar zircon age and Hf isotope history. • Eo-Paleoarchean detrital and igneous zircon suggest intermediate-felsic crust. • Multiple episodes of crustal rejuvenation facilitated growth of the largest craton. • Archean metasedimentary rocks may contain rare detrital metamorphic zircon. • Hadean – Archean zircon TE discrimination requires initial U and Th concentrations.

• Multiple textural varieties of spodumene and quartz exist at the Tanco pegmatite. • Spodumene-quartz intergrowths form from the breakdown of metastable virgilite. • Spodumene-quartz symplectites form from dissolution-reprecipitation reactions. • As spodumene crystal sizes get smaller in intergrowths, issues arise in metallurgy. The Archean Tanco deposit in southeastern Manitoba is a complexly zoned petalite-subtype pegmatite. Lithium mineralization primarily occurs as spodumene-quartz intergrowths (SQUI), which are historically interpreted as pseudomorphs of petalite breakdown. This breakdown would produce a spodumene-to-quartz ratio of 56.3:43.7 vol%, volume reduction textures, and low trace element contents. However, SQUI at Tanco shows variable spodumene-to-quartz ratios, remarkably little petalite preservation or volume reduction textures and green spodumene with up to 8330 ppm Fe. Three textural varieties of spodumene-quartz associations were identified and described in zones 45 and 50 of the pegmatite: (1) Classic SQUI with c-axis-oriented, oscillatory-zoned spodumene and quartz, which locally occur within relict crystal outlines; (2) spodumene-quartz symplectites (SQS) characterized by pervasive symplectic intergrowths with elevated Cs and depleted Mn and Sn; and (3) macro spodumene-quartz intergrowths with randomly oriented, stubby crystals and chaotic zonation. Metastable virgilite exists as a solid solution with quartz and is known to crystallize from experimental pegmatite crystallization studies. The breakdown of virgilite into classic SQUI could explain the lack of petalite and volume reduction textures, variable mineralogical ratios, and iron-bearing spodumene in SQUI. SQS likely resulted from coupled dissolution-reprecipitation in the presence of a fluid, while macro intergrowths represent primary magmatic textures. The occurrence of SQS at Tanco poses a metallurgical issue due to the fine sizes and hardness of the intergrowths leading to Li recovery issues. Understanding the origins of lithium mineralization is important as the quartz-spodumene textural associations have metallurgical recovery implications, affecting the overall economic viability of deposits.

The escalating threat of wildfires has been inflicting growing harm on power systems in recent years. Transmission lines, especially those supported by wooden poles, are susceptible to destruction by wildfires, leading to loss of load (loss of energy). It is, therefore, crucial to incorporate the unavailability of transmission lines due to wildfire into the quantitative risk assessment of power systems.Burn-P3 (probability, prediction, and planning) is a spatial fire simulation model that is commonly used in land-management planning and wildfire research. Its application can be extended to simulate the ignition and spread of wildfires in transmission corridors. The Burn-P3 outputs are quantitative, however, they do not represent the absolute annual likelihood of experiencing a wildfire. Instead, the outputs provide a measure of relative likelihood, subject to landscape alterations. A process/method is, therefore, needed to convert the burn probability map to the annual unavailability of a transmission line for risk assessment in power systems. A method is proposed to align a map with wood pole transmission lines and their corresponding burn probabilities in the area. The annual unavailability of a transmission line is assessed and applied in the loss of load and/or loss of energy studies to quantify the wildfire-related risks. The output can be utilized for capital project assessment and prioritization.

Estimating the electrical model parameters of photovoltaic (PV) modules, which are normally not provided by manufacturers, is crucial for modeling PV systems in simulation studies. This work proposes a novel robust approach to identify parameters of the single diode model (SDM) of a PV module. The equivalent electrical circuit parameters are extracted by solving a set of equations through the application of the Lambert W function for a given value of ideality factor, Ai, based on data found in manufacturer's datasheets under standard test conditions (STC). To find the optimal ideality factor, parameters are computed for different Ai values within a defined range, and their fitness is assessed using a multi-criteria objective function. Irradiance and temperature dependence of the parameters are estimated using the temperature coefficients available on datasheets and the irradiation coefficient of open circuit voltages, calculated using the data at Normal Operating Cell Temperature (NOCT). By utilizing the estimated values for the parameters, the current-voltage (1- V) characteristics at different levels of irradiation and temperatures, and the temperature coefficient of maximum power for several PV modules, were obtained. The errors expressed in terms of root means square deviations compared to the datasheet values were less than 4 % for the five different PV modules considered.

As the penetration of complex software-driven inverter-based resources (IBRs) rapidly increases in power systems around the world, the need for modeling large areas of these systems in a time-domain electromagnetic transient (EMT) environment has also increased. Since 2016, the Australian Energy Market Operator (AEMO) has been developing “large-scale” EMT models of parts of Australia’s interconnected Eastern Australian power system, known as the <i>National Electricity Market</i> ( <i>NEM</i> ), for use across many of AEMO’s functions as the NEM’s independent system operator. Due to the size and complexity of the models, these simulations have required large computational requirements and were typically very slow, taking more than 24 h for a 30-s simulation in 2016.