Indian summer monsoon (ISM) has profound impact on water resources over the Asian Water Towers (AWTs) and surroundings. Stable isotopes in precipitation (δO and δD) are crucial tracers of ISM moisture transport processes. Here we presented spatiotemporal variations of stable isotopes in precipitation at three stations over Bangladesh in 2017-2018 to evaluate the influence of moisture sources and transport on intra-seasonal variations of stable isotopes in precipitation, combined with local meteorological data, ERA5 reanalysis data and HYSPLIT model. We found Bay of Bengal (BoB), tropical Indian Ocean (TIO) and Arabian Sea (AS) were the primary moisture suppliers throughout the year and moisture uptake process primarily occurred over BoB. The most enriched δO and δD values exist in the pre-monsoon season, associated with >50% contributions from BoB, and gradually decline throughout the monsoon and post-monsoon seasons due to increased contribution of moisture from AS (~30%) and IO (~40%), and reach to their lowest values by the end of the post-monsoon season when >25% contributed from BoB and ~20% from TIO. The strongly positive δO-OLR and negative δO-humidity relationships were found at all three stations showing a decreasing pattern from south to north.δO-temperature (δO-precipitation) relationship was only found at southern stations at local scale. Convective activities over the AS, BoB and northern IO primarily regulate the δO depletion, and a weak (strong) flux-δO relationship for northward (eastward) transport was found. This study could improve understanding of moisture transport by the ISM for our societies to promote the water resource management over AWTs.

We present new measurements of the vertical density profile of the Earth’s atmosphere at altitudes between 70 and 200\,km, based on Earth occultations of the Crab Nebula observed with the X-ray Imaging Spectrometer onboard Suzaku and the Hard X-ray Imager onboard Hitomi. X-ray spectral variation due to the atmospheric absorption is used to derive tangential column densities of the absorbing species, i.e., N and O including atoms and molecules, along the line of sight. The tangential column densities are then inverted to obtain the atmospheric number density. The data from 219 occultation scans at low latitudes in both hemispheres from September 15, 2005 to March 26, 2016 are analyzed to generate a single, highly-averaged (in both space and time) vertical density profile. The density profile is in good agreement with the NRLMSISE-00 model, except for the altitude range of 70–110\,km, where the measured density is $\sim$50\% smaller than the model. Such a deviation is consistent with the recent measurement with the SABER aboard the TIMED satellite (Cheng et al. 2020). Given that the NRLMSISE-00 model was constructed some time ago, the density decline could be due to the radiative cooling/contracting of the upper atmosphere as a result of greenhouse warming in the troposphere. However, we cannot rule out a possibility that the NRL model is simply imperfect in this region. We also present future prospects for the upcoming Japan-US X-ray astronomy satellite, XRISM, which will allow us to measure atmospheric composition with unprecedented spectral resolution of $\Delta E \sim 5$\,eV in 0.3–12\,keV.

The projection of extreme convective precipitation by global climate models (GCM) exhibits significant uncertainty due to coarse resolutions. Direct dynamical downscaling (DDD) of regional climate at kilometer-scale resolutions provides valuable insight into extreme precipitation changes, but its computational expense is formidable. Here we document the effectiveness of machine learning to enable smart dynamical downscaling (SDD), which selects a small subset of GCM data to conduct downscaling. Trained with data for three subtropical/tropical regions, convolutional neural networks (CNNs) retained 92% to 98% of extreme precipitation events (rain intensity higher than the 99th percentile) while filtering out 88% to 95% of circulation data. When applied to reanalysis data sets differing from training data, the CNNs’ skill in retaining extremes decreases modestly in subtropical regions but sharply in the deep tropics. Nonetheless, one of the CNNs can still retain 62% of all extreme events in the deep tropical region in the worst case.

The extreme variability of the cold point tropopause temperature (TCPT) and height (HCPT) are examined over a tropical station, Gadanki (13.45N, 79.2E) using high-resolution radiosonde data during the period 2006-2014. The extreme variabilities such as the coldest (warmest) tropopause is defined if TCPT is lesser (greater) than the lower (upper) limit of its two-sigma level whereas the highest (lowest) tropopause is defined as the HCPTis greater (lesser) than the lower (upper) limit of its two-sigma level. In total 161 extreme cases such as the coldest (52) and warmest (30) TCPT and the highest (57) and lowest (22) HCPT are observed. The coldest (187.2±1.60 K, 17.3±0.52 km), warmest (194.2±1.78 K, 16.9±0.89 km), lowest (191.7±1.78 K, 18.2±0.55 km) and highest (191.8±2.11 K, 16.2±0.38 km) occurs without preference of season. These extreme tropopause cases occur independently. Thermal structure of the coldest tropopause cases reveals that they are often sharper whereas the warmest, highest and lowest tropopause is broader. Water vapor and ozone concentrations are found to be high for the warmest tropopause and low for the coldest tropopause. Under the shallow convection, extreme temperature profiles, in general, show prominent warming between 8-14 km while anomalous cooling (warming) just below (above) the CPT. The occurrence of the tropical cyclones, cirrus clouds and equatorial wave propagation are the possible candidates for the occurrence of the extreme tropopauses.

This is a reply to a comment on the original research study with the title: ‘Unintentional unfairness when applying new greenhouse gas emissions metrics at country level’. This reply responds to some of the claims made in the comment and provides a scientific rebuttal.

Recent work using CMIP5 models under RCP8.5 suggests that individual multimodel-mean changes in precipitation and wind variability associated with the Madden-Julian oscillation (MJO) are not detectable until the end of the 21st century. However, a decrease in the ratio of MJO circulation to precipitation anomaly amplitude is detectable as early as 2021-2040, consistent with an increase in dry static stability as predicted by weak-temperature-gradient balance. Here, we examine MJO activity in multiple reanalyses (ERA5,MERRA-2, and ERA-20C) and find that MJO wind and precipitation anomaly amplitudes have a complicated time evolution over the record. However, a decrease in the ratio of MJO circulation to precipitation anomaly amplitude is detected over the observational period, consistent with the change in dry static stability. These results suggest that weak-temperature-gradient theory may be able to help explain changes in MJO activity in recent decades.

Evaluating the influence of anthropogenic emissions changes on air quality requires accounting for the influence of meteorological variability. Statistical methods such as multiple linear regression (MLR) models with basic meteorological variables are often used to remove meteorological variability and estimate trends in measured pollutant concentrations attributable to emissions changes. However, the ability of these widely-used statistical approaches to correct for meteorological variability remains unknown, limiting their usefulness in the real-world policy evaluations. Here, we quantify the performance of MLR and other quantitative methods using two scenarios simulated by a chemical transport model, GEOS-Chem, as a synthetic dataset. Focusing on the impacts of anthropogenic emissions changes in the US (2011 to 2017) and China (2013 to 2017) on PM2.5 and O3, we show that widely-used regression methods do not perform well in correcting for meteorological variability and identifying long-term trends in ambient pollution related to changes in emissions. The estimation errors, characterized as the differences between meteorology-corrected trends and emission-driven trends under constant meteorology scenarios, can be reduced by 30%-42% using a random forest model that incorporates both local and regional scale meteorological features. We further design a correction method based on GEOS-Chem simulations with constant emission input and quantify the degree to which emissions and meteorological influences are inseparable, due to their process-based interactions. We conclude by providing recommendations for evaluating the effectiveness of emissions reduction policies using statistical approaches.

Predictable internal climate variability on decadal timescales (2-10 years) is associated with large-scale oceanic processes, however these predictable signals may be masked by the noisy climate system. One approach to overcoming this problem is investigating state-dependent predictability - how differences in prediction skill depend on the initial state of the system. We present a machine learning approach to identify state-dependent predictability on decadal timescales in the Community Earth System Model version 2 by incorporating uncertainty estimates into a regression neural network. We leverage the network’s prediction of uncertainty to examine state dependent predictability in sea surface temperatures by focusing on predictions with the lowest uncertainty outputs. In particular, we study two regions of the global ocean - the North Atlantic and North Pacific - and find that skillful initial states identified by the neural network correspond to particular phases of Atlantic multi-decadal variability and the interdecadal Pacific oscillation.

El Niño Southern Oscillation (ENSO) is the leading mode of interannual climate variability, with large socioeconomical and environmental impacts. The main conceptual model for ENSO, the Recharge Oscillator (RO), considers two independent modes: the fast zonal tilt mode in phase with central-eastern Pacific Temperature (Te), and the slow recharge mode in phase quadrature. However, usual indices (western or equatorial sea level/thermocline depth h) do not orthogonally isolate the slow recharge mode, leaving it correlated with Te. Furthermore the optimal index is currently debated. Here, by objectively optimizing the RO equations fit to observations, we develop an improved recharge index. (1) Te-variability is regressed out, building h_ind statistically-independent from Te. Capturing the pure recharge, h_ind reconciles usual indices. (2) The optimum is equatorial plus southwestern Pacific h_ind_eq+sw (because of ENSO Ekman pumping meridional asymmetry). Using h_ind_eq+sw, the RO becomes more consistent with observations. h_ind_eq+sw is more relevant for ENSO operational diagnostics.

This paper describes the first implementation of the d x=3.25 km version of the Energy Exascale Earth System Model (E3SM) global atmosphere model and its behavior in a 40 day prescribed-sea-surface-temperature simulation (Jan 20-Feb 28, 2020). This simulation was performed as part of the DYnamics of the Atmospheric general circulation Modeled On Non-hydrostatic Domains (DYAMOND) phase 2 model intercomparison. Effective resolution is found to be $\sim 6x the horizontal grid resolution despite using a coarser grid for physical parameterizations. Despite this new model being in an immature and untuned state, moving to 3.25 km grid spacing solves several long-standing problems with the E3SM model. In particular, Amazon precipitation is much more realistic, the frequency of light and heavy precipitation is improved, agreement between the simulated and observed diurnal cycle of tropical precipitation is excellent, and the vertical structure of tropical convection and coastal stratocumulus look good. In addition, the new model is able to capture the frequency and structure of important weather events (e.g. hurricanes, midlatitude storms including atmospheric rivers, and cold air outbreaks). Interestingly, this model does not get rid of the erroneous southern branch of the intertropical convergence zone nor the tendency for strongest convection to occur over the Maritime Continent rather than the West Pacific, both of which are classic climate model biases. Several other problems with the simulation are identified, underscoring the fact that this model is a work in progress.

We study the suitability of an initial condition ensemble to form the conceptual basis of defining climate. We point out that the most important criterion is the uniqueness of the probability measure on which the definition relies. We first propose, in harmony with earlier work, to represent such a probability measure by the distribution of ensemble members that have converged to the probability density of the natural probability measure of the so-called snapshot or pullback attractor of the dynamics, which is time dependent in the presence of external forcing. Then we refine the proposal by taking a density that is conditional on the (possibly time-evolving) state of system components with time scales longer than the horizon of a particular study. We discuss the applicability of such a definition in the Earth system and its realistic models, and conclude that micro initialization from observations in slower system components perhaps provides the practically relevant probability density after a few decades of convergence. However, the absence of sufficient time scale separation between system components or regime transitions in slower system components might preclude uniqueness, at least in certain subsystems, and time evolution in slower system components might induce unforced climate changes, leading to the need for targeted investigations to determine the forced response. We propose an initialization scheme for studying all these issues in Earth system models.

Drylands occupy ~40% of the land surface and are thought to dominate global carbon (C) cycle inter-annual variability (IAV). Therefore, it is imperative that global terrestrial biosphere models (TBMs), which form the land component of IPCC earth system models, are able to accurately simulate dryland vegetation and biogeochemical processes. However, compared to more mesic ecosystems, TBMs have not been widely tested or optimized using in situ dryland CO2 fluxes. Here, we address this gap using a Bayesian data assimilation system and 89 site-years of daily net ecosystem exchange (NEE) data from 12 southwest US Ameriflux sites to optimize the C cycle parameters of the ORCHIDEE TBM. The sites span high elevation forest ecosystems, which are a mean sink of C, and low elevation shrub and grass ecosystems that are either a mean C sink or “pivot” between an annual C sink and source. We find that using the default (prior) model parameters drastically underestimates both the mean annual NEE at the forested mean C sink sites and the NEE IAV across all sites. Our analysis demonstrated that optimizing phenology parameters are particularly useful in improving the model’s ability to capture both the magnitude and sign of the NEE IAV. At the forest sites, optimizing C allocation, respiration, and biomass and soil C turnover parameters reduces the underestimate in simulated mean annual NEE. Our study demonstrates that all TBMs need to be calibrated for dryland ecosystems before they are used to determine dryland contributions to global C cycle variability and long-term carbon-climate feedbacks.

Marine heatwaves (MHWs) are extreme oceanic warm water events (above 90th percentile threshold) that significantly impact the marine environment. Several studies have recently explored the genesis and impacts of MHWs though they are least understood in the tropical Indian Ocean. Here we investigate the genesis and trend of MHWs in the Indian Ocean during 1982–2018 and their role in modulating the Indian monsoon. We find that the rapid warming in the Indian Ocean plays a critical role in increasing the number of MHWs. Meanwhile, the El Nino has a prominent influence on the occurrence of MHWs during the summer monsoon. The Indian Ocean warming and the El Nino variability have synergistically resulted in some of the strongest and long-lasting MHWs in the Indian Ocean. The western Indian Ocean (WIO) region experienced the largest increase in MHWs at a rate of 1.2–1.5 events per decade, followed by the north Bay of Bengal at a rate of 0.4–0.5 events per decade. Locally, the MHWs are induced by increased solar radiation, relaxation of winds, and reduced evaporative cooling. In the western Indian Ocean, the decreased winds further restrict the heat transport by ocean currents from the near-equatorial regions towards the north. Our analysis indicates that the MHWs in the western Indian Ocean and the north Bay of Bengal lead to a reduction in monsoon rainfall over the central Indian subcontinent. On the other hand, there is an enhancement of monsoon rainfall over southwest India due to the MHWs in the Bay of Bengal.

The global mean surface temperature (GMST) record exhibits both interannual to multidecadal variability and a long-term warming trend due to external climate forcing. To explore the predictability of temporary slowdowns in decadal warming, we apply an artificial neural network (ANN) to climate model data from the Community Earth System Model Version 2 Large Ensemble. Here, an ANN is tasked with whether or not there will be a slowdown in the rate of the GMST trend by using maps of ocean heat content at the onset. Through a machine learning explainability method, we find the ANN is learning off-equatorial patterns of anomalous ocean heat content that resemble transitions in the phase of the Interdecadal Pacific Oscillation in order to make slowdown predictions. Finally, we test our ANN on observed historical data, which further reveals how explainable neural networks are useful tools for understanding decadal variability in both climate models and observations.

The Yucatan Peninsula (YP) has a complex hydroclimate with many proposed drivers of interannual and longer-term variability, ranging from coupled ocean-atmosphere processes to frequency of tropical cyclones. The mid-Holocene, a time of higher Northern Hemisphere summer insolation, provides an opportunity to test the relationship between Yucatan Peninsula precipitation and ocean temperature. Here we present a new, ~annually resolved speleothem record of stable isotope (δ18O and δ13C) and trace element (Mg/Ca and Sr/Ca) ratios for a section of the mid-Holocene (5.2-5.7 kyr BP), before extensive agriculture began in the region. A meter-long stalagmite from Rio Secreto, a cave system in Playa del Carmen, Mexico, was dated using U-Th geochronology and layer counting, yielding multidecadal age uncertainty (median 2SD of +/- 70 years). New proxy data were compared to an existing late Holocene stalagmite record from the same cave system, allowing us to examine changes in hydrology over time, and to paleoclimate records from the southern YP. The δ18O, δ13C and Mg/Ca data consistently indicate higher mean precipitation and lower precipitation variability during the mid-Holocene compared to the late Holocene. Despite this reduced variability, multidecadal precipitation variations were persistent in regional hydroclimate during the mid-Holocene. We therefore conclude that higher summer insolation led to increased mean precipitation and decreased precipitation variability in the northern YP, but that the region is susceptible to dry periods across climate mean states. Given projected decreases in wet season precipitation in the YP’s near future, we suggest that climate mitigation strategies emphasize drought preparation.

The COVID-19 pandemic led to widespread reductions in mobility and induced observable changes in the atmosphere. Recent work has employed novel mobility datasets as a proxy for trace gas emissions from traffic, yet there has been little work evaluating these emission numbers. Here, we systematically compare these mobility datasets to traffic data from local governments in seven diverse urban and rural regions to characterize the magnitude of errors in emissions that result from using the mobility data. We observe differences in excess of 60% between these mobility datasets and local traffic data, which result in large errors in emission estimates. We could not find a general functional relationship between mobility data and traffic flow over all regions. Future work should be cautious when using these mobility metrics for emission estimates. Further, we use the local government data to identify emission reductions from traffic in the range of 7-22% in 2020 compared to 2019.

Most tree species predominantly associate with a single type of mycorrhizal fungi, which can differentially affect plant nutrient acquisition and biogeochemical cycling. Here, we address for the first time the impact of mycorrhizal distributions on global carbon and nutrient cycling. Using the state-of-the-art carbon-nitrogen economics within the Community Land Model version 5 (CLM5) we found Net Primary Productivity (NPP) increased throughout the 21st century by 20%; however, as soil nitrogen has progressively become limiting, the costs to NPP for nitrogen acquisition — i.e., to mycorrhizae — have increased at a faster rate by 60%. This suggests that nutrient acquisition will increasingly demand a higher portion of assimilated carbon to support the same productivity. Uncertainties in mycorrhizal distributions are non-trivial, however, with uncertainties in NPP by up to 345 Tg C yr-1, depending on which published distribution is used. Remote sensing capabilities for mycorrhizal detection show promise for refining these estimates further.

Atlantic Water (AW) is the largest reservoir of heat in the Arctic Ocean, isolated from the surface and sea-ice by a strong halocline. In recent years AW shoaling and warming are thought to have had an increased influence on sea-ice in the Eurasian Basin. In this study we analyse 59000 profiles from across the Arctic from the 1970s to 2018 to obtain an observationally-based pan-Arctic picture of the AW layer, and to quantify temporal and spatial changes. The potential temperature maximum of the AW (the AW core) is found to be an easily detectable, and generally effective metric for assessments of AW properties, although temporal trends in AW core properties do not always reflect those of the entire AW layer. The AW core cools and freshens along the AW advection pathway as the AW loses heat and salt through vertical mixing at its upper bound, as well as via likely interaction with cascading shelf flows. In contrast to the Eurasian Basin, where the AW warms (by approximately 0.7°C between 2002 and 2018) in a pulse-like fashion and has an increased influence on upper ocean heat content, AW in the Canadian Basin cools (by approximately 0.1°C between 2008 and 2018) and becomes more isolated from the surface due to the intensification of the Beaufort Gyre. These opposing AW trends in the Eurasian and Canadian Basins of the Arctic over the last 40 years suggest that AW in these two regions may evolve differently over the coming decades.

Tropical anvil clouds are an important player in Earth’s climate and climate sensitivity, but simulations of anvil clouds are uncertain. Here we pinpoint one source of uncertainty by demonstrating a marked increase of anvil cloud fraction with resolution in cloud-resolving simulations of radiative-convective equilibrium. This increase in cloud fraction can be traced back to the resolution dependence of horizontal mixing between clear and cloudy air. A mixing timescale is diagnosed for each simulation using the cloud fraction theory of Seeley et al. (2019), and is found to scale linearly with grid spacing, as expected from a simple scaling law. Thus mixing becomes more efficient with increasing resolution, generating more evaporation, decreased precipitation efficiency, greater mass flux, and hence greater detrainment and cloud fraction.The decrease in precipitation efficiency also yields a marked increase in relative humidity with resolution.

Wildfire is an important but understudied natural hazard. Research on wildfire, as with other natural hazards, is all too often conducted at a spatial scale that is too broad to identify local or even regional patterns. This study addresses these research gaps by examining the current and future wildfire risk, considering projections of population and property value, at the census-block level in Louisiana, a U.S. state with relatively dense population and abundant timber resources that would be vulnerable to loss from this hazard. Here wildfire risk is defined as the product of vulnerability to the hazard (which is itself defined as the product of burn probability, damage probability, and percent damaged) and exposure to the hazard, the latter of which is represented here by property value. Historical data (1992-2015) suggest that the highest risk is in southwestern inland, east-central, extreme northwestern, and coastal southwestern Louisiana. Based on existing climate and environmental model output, this research assumes that wildfire will increase by 25 percent by 2050 in Louisiana from current values. When combined with projections of population and property value, it is determined that the geographic distribution of risk by 2050 will remain similar to that today-with highest risk in southwestern inland Louisiana and east-central Louisiana. However, the magnitude of risk will increase across the state, especially in those areas. These results will assist environmental planners in preparing for and mitigating a substantial hazard that often goes underestimated.

Change in global mean surface temperature (GMST), based on a blend of land air and ocean water temperatures, is a widely cited climate change indicator that informs the Paris Agreement goal to limit global warming since preindustrial to “well below” 2°C. Assessment of current GMST enables determination of remaining target-consistent warming and therefore a relevant remaining carbon budget. In recent IPCC reports, GMST was estimated via linear regression or differences between decade-plus period means. We propose non-linear continuous local regression (LOESS) using ±20 year windows to derive GMST across all periods of interest. Using the three observational GMST datasets with almost complete interpolated spatial coverage since the 1950s, we evaluate 1850—1900 to 2019 GMST as 1.14°C with a likely (17—83 %) range of 1.05—1.25°C, based on combined statistical and observational uncertainty, compared with linear regression of 1.05°C over 1880—2019. Performance tests in observational datasets and two model large ensembles demonstrate that LOESS, like period mean differences, is unbiased. However, LOESS also provides a statistical uncertainty estimate and gives warming through 2019, rather than the 1850—1900 to 2010—2019 period mean difference centered at the end of 2014. We derive historical global near-surface air temperature change (GSAT), using a subset of CMIP6 climate models to estimate the adjustment required to account for the difference between ocean water and ocean air temperatures. We find GSAT of 1.21°C (1.11—1.32°C) and calculate remaining carbon budgets. We argue that continuous non-linear trend estimation offers substantial advantages for assessment of long-term observational GMST.

The North Atlantic ocean is key to climate through its role in heat transport and storage. Climate models suggest that the circulation is weakening but the physical drivers of this change are poorly constrained. Here, the root mechanisms are revealed with the explicitly transparent machine learning method Tracking global Heating with Ocean Regimes (THOR). Addressing the fundamental question of the existence of dynamical coherent regions, THOR identifies these and their link to distinct currents and mechanisms such as the formation regions of deep water masses, and the location of the Gulf Stream and North Atlantic Current. Beyond a black box approach, THOR is engineered to elucidate its source of predictive skill rooted in physical understanding. A labeled dataset is engineered using an explicitly interpretable equation transform and k-means application to model data, allowing theoretical inference. A multilayer perceptron is then trained, explaining its skill using a combination of layerwise relevance propagation and theory. With abrupt CO2 quadrupling, the circulation weakens due to a shift in deep water formation regions, a northward shift of the Gulf stream and an eastwards shift in the North Atlantic Current. If CO2 is increased 1% yearly, similar but weaker patterns emerge influenced by natural variability. THOR is scalable and applicable to a range of models using only the ocean depth, dynamic sea level and wind stress, and could accelerate the analysis and dissemination of climate model data. THOR constitutes a step towards trustworthy machine learning called for within oceanography and beyond.

The Atmospheric River (AR) Tracking Method Intercomparison Project (ARTMIP) is a community effort to systematically assess how the uncertainties from AR detectors (ARDTs) impact our scientific understanding of ARs. This study describes the ARTMIP Tier 2 experimental design and initial results using the Coupled Model Intercomparison Project (CMIP) Phases 5 and 6 multi-model ensembles. We show that AR statistics from a given ARDT in CMIP5/6 historical simulations compare remarkably well with the MERRA-2 reanalysis. In CMIP5/6 future simulations, most ARDTs project a global increase in AR frequency, counts, and sizes, especially along the western coastlines of the Pacific and Atlantic oceans. We find that the choice of ARDT is the dominant contributor to the uncertainty in projected AR frequency when compared with model choice. These results imply that new projects investigating future changes in ARs should explicitly consider ARDT uncertainty as a core part of the experimental design.

The characterization of changes over the full distribution of precipitation intensities remains an overlooked and underexplored subject, despite their critical importance to hazard assessments and water resource management. Here, we aggregate daily in situ Global Historical Climatology Network precipitation observations within seventeen internally consistent domains in the United States for two time periods (1951-1980 and 1991-2020). We find statistically significant changes in wet day precipitation distributions in all domains – changes primarily driven by a shift from lower to higher wet day intensities. Patterns of robust change are geographically consistent, with increases in the mean (4.5-5.7%) and standard deviation (4.4-8.7%) of wet day intensity in the eastern U.S., but mixed signals in the western U.S. Beyond their critical importance to the aforementioned impact assessments, these observational results can also inform climate model performance evaluations.