Highlight articles

Root-mean-square error (RMSE) or mean absolute error (MAE): when to use them or not

The task of evaluating competing models is fundamental to science. Models are evaluated based on an objective function, the choice of which ultimately influences what scientists learn from their observations. The mean absolute error (MAE) and root-mean-squared error (RMSE) are two such functions. Both are widely used, yet there remains enduring confusion over their use. This article reviews the theoretical justification behind their usage, as well as alternatives for when they are not suitable.

The eWaterCycle platform for open and FAIR hydrological collaboration

With the eWaterCycle platform, we are providing the hydrological community with a platform to conduct their research that is fully compatible with the principles of both open science and FAIR science. The eWatercyle platform gives easy access to well-known hydrological models, big datasets and example experiments. Using eWaterCycle hydrologists can easily compare the results from different models, couple models and do more complex hydrological computational research.

Towards automatic finite-element methods for geodynamics via Firedrake

Firedrake is a state-of-the-art system that automatically generates highly optimised code for simulating finite-element (FE) problems in geophysical fluid dynamics. It creates a separation of concerns between employing the FE method and implementing it. Here, we demonstrate the applicability and benefits of Firedrake for simulating geodynamical flows, with a focus on the slow creeping motion of Earth’s mantle over geological timescales, which is ultimately the engine driving our dynamic Earth.

Training a supermodel with noisy and sparse observations: a case study with CPT and the synch rule on SPEEDO – v.1

In this study, we present a novel formulation to build a dynamical combination of models, the so-called supermodel, which needs to be trained based on data. Previously, we assumed complete and noise-free observations. Here, we move towards a realistic scenario and develop adaptations to the training methods in order to cope with sparse and noisy observations. The results are very promising and shed light on how to apply the method with state of the art general circulation models.

Using neural network ensembles to separate ocean biogeochemical and physical drivers of phytoplankton biogeography in Earth system models

It can be challenging to understand why Earth system models (ESMs) produce specific results because one can arrive at the same result simply by changing the values of the parameters. In our paper, we demonstrate that it is possible to use machine learning to figure out how and why particular components of an ESM (such as biology or ocean circulations) affect the output. This work could be applied to observations to improve the accuracy of the formulations used in ESMs.

The Whole Antarctic Ocean Model (WAOM v1.0): development and evaluation

Here we present an improved model of the Antarctic continental shelf ocean and demonstrate that it is capable of reproducing present-day conditions. The improvements are fundamental and regard the inclusion of tides and ocean eddies. We conclude that the model is well suited to gain new insights into processes that are important for Antarctic ice sheet retreat and global ocean changes. The model will ultimately help to improve projections of sea level rise and climate change.

Impact of increased resolution on long-standing biases in HighResMIP-PRIMAVERA climate models

Climate models do not fully reproduce observations: they show differences (biases) in regional temperature, precipitation, or cloud cover. Reducing model biases is important to increase our confidence in their ability to reproduce present and future climate changes. Model realism is set by its resolution: the finer it is, the more physical processes and interactions it can resolve. We here show that increasing resolution of up to ~ 25 km can help reduce model biases but not remove them entirely.

SELF v1.0: a minimal physical model for predicting time of freeze-up in lakes

The time when lakes freeze varies considerably from year to year. A common way to predict it is to use negative degree days, i.e., the sum of air temperatures below 0°C, a proxy for the heat lost to the atmosphere. Here, we show that this is insufficient as the mixing of the surface layer induced by wind tends to delay the formation of ice. To do so, we developed a minimal model based on a simplified energy balance, which can be used both for large-scale analyses and short-term predictions.

Assessment of the ParFlow–CLM CONUS 1.0 integrated hydrologic model: evaluation of hyper-resolution water balance components across the contiguous United States

Modeling the hydrologic cycle at high resolution and at large spatial scales is an incredible opportunity and challenge for hydrologists. In this paper, we present the results of a high-resolution hydrologic simulation configured over the contiguous United States. We discuss simulated water fluxes through groundwater, soil, plants, and over land, and we compare model results to in situ observations and satellite products in order to build confidence and guide future model development.

The interpretation of temperature and salinity variables in numerical ocean model output and the calculation of heat fluxes and heat content

We show that the way that the air–sea heat flux is treated in ocean models means that the model’s temperature variable should be interpreted as being Conservative Temperature, irrespective of whether the equation of state used in an ocean model is EOS-80 or TEOS-10.

ESSI Earth and Space Science Informatics Highlight articles

Highlight articles

Information for

External links