Week 3–4 Prediction of Wintertime CONUS Temperature Using Machine Learning Techniques

This paper shows that skillful week 3–4 predictions of a large-scale pattern of 2 m temperature over the US can be made based on the Nino3.4 index alone, where skillful is defined to be better than climatology. To find more skillful regression models, this paper explores various machine learning strategies (e.g., ridge regression and lasso), including those trained on observations and on climate model output. It is found that regression models trained on climate model output yield more skillful predictions than regression models trained on observations, presumably because of the larger training sample. Nevertheless, the skill of the best machine learning models are only modestly better than ordinary least squares based on the Nino3.4 index. Importantly, this fact is difficult to infer from the parameters of the machine learning model because very different parameter sets can produce virtually identical predictions. For this reason, attempts to interpret the source of predictability from the machine learning model can be very misleading. The skill of machine learning models also are compared to those of a fully coupled dynamical model, CFSv2. The results depend on the skill measure: for mean square error, the dynamical model is slightly worse than the machine learning models; for correlation skill, the dynamical model is only modestly better than machine learning models or the Nino3.4 index. In summary, the best predictions of the large-scale pattern come from machine learning models trained on long climate simulations, but the skill is only modestly better than predictions based on the Nino3.4 index alone.

Modelling the Vadret da Tschierva, Switzerland: calibration with the historical length record and future response to climate change

The Vadret da Tschierva (Vd Tschierva) is a 4 km long glacier in the Swiss Alps spanning an altitude range of 2400–4049 m a.s.l. Length observations since 1855 show steady retreat interrupted by a period of advance from 1965 until 1985. The total retreat is ~2200 m (period 1855–2018). We have studied the Vd Tschierva with a flowline model, combined with ‘buckets’ that represent steep hanging glaciers and ice-free rock faces delivering mass to the main stream. The model is calibrated by a control method, in which an ELA history is objectively determined by finding the best match between observed and simulated glacier length. There is a modest correlation between the reconstructed ELA and an ELA record based on meteorological observations at Segl-Maria (only 8 km away from the glacier). It is difficult to reproduce the observed length record when the glacier model is driven by climate model output (Coupled Model Intercomparison Project 5). We have calculated the future evolution of the Vd Tschierva for different rates of ELA rise. For a constant rise of 4 ${\rm m\;}{\rm a}^{ \hbox{-} 1}$ , we predict that the glacier length will change from the current 3.2 km to ~1.7 km in the year 2100.

Constructing High-Resolution, Bias-Corrected Climate Products: A Comparison of Methods

High-resolution, bias-corrected climate data are necessary for climate impact studies at local scales. Gridded historical data are convenient for bias correction but may contain biases resulting from interpolation. Long-term, quality-controlled station data are generally superior climatological measurements, but because the distribution of climate stations is irregular, station data are challenging to incorporate into downscaling and bias-correction approaches. Here, we compared six novel methods for constructing full-coverage, high-resolution, bias-corrected climate products using daily maximum temperature simulations from a regional climate model (RCM). Only station data were used for bias correction. We quantified performance of the six methods with the root-mean-square-error (RMSE) and Perkins skill score (PSS) and used two ANOVA models to analyze how performance varied among methods. We validated the six methods using two calibration periods of observed data (1980–89 and 1980–2014) and two testing sets of RCM data (1990–2014 and 1980–2014). RMSE for all methods varied throughout the year and was larger in cold months, whereas PSS was more consistent. Quantile-mapping bias-correction techniques substantially improved PSS, while simple linear transfer functions performed best in improving RMSE. For the 1980–89 calibration period, simple quantile-mapping techniques outperformed empirical quantile mapping (EQM) in improving PSS. When calibration and testing time periods were equivalent, EQM resulted in the largest improvements in PSS. No one method performed best in both RMSE and PSS. Our results indicate that simple quantile-mapping techniques are less prone to overfitting than EQM and are suitable for processing future climate model output, whereas EQM is ideal for bias correcting historical climate model output.

Deep Learning for Climate Model Output Statistics

<p>Climate models are an important tool for the assessment of prospective climate change effects but they suffer from systematic and representation errors, especially for precipitation. Model output statistics (MOS) reduce these errors by fitting the model output to observational data with machine learning. In this work, we explore the feasibility and potential of deep learning with convolutional neural networks (CNNs) for MOS. We propose the CNN architecture ConvMOS specifically designed for reducing errors in climate model outputs and apply it to the climate model REMO. Our results show a considerable reduction of errors and mostly improved performance compared to three commonly used MOS approaches.</p>

The impact of climate change on soil erosion: a systematic review

<p>Climate change is expected to cause an increase of extreme precipitation and consequently an increase of soil erosion in many regions worldwide, although large differences are reported. Therefore, this study systematically reviews research presenting projected changes in soil erosion under climate change, focussing on studies that forced soil erosion models with precipitation from climate model output. A total of 766 documents were analysed and further evaluated based on predetermined inclusion criteria, resulting in a selection of 168 documents published between 1995 and 2021. From these documents a total of 35 variables were recorded, including information related to bibliography, objective, study site, climate model, soil erosion model, land use change scenarios, soil and water conservation techniques, and the projected change in soil erosion under climate change. Studies were performed on all continents, with the majority in Europe (32%), Asia (29%) and North America (23%). The study sites were mainly located in humid continental (28%) and humid subtropical climates (22%). The studies were equally distributed over the future periods (i.e. near-, mid- and end-century) and emissions scenarios (i.e. low, intermediate and high). The majority of the studies were forced by a single climate model (44%), while 67% of the studies used a climate model ensemble smaller than 5. MUSLE (31%), RUSLE (18%) and WEPP (9%) are the most applied soil erosion models. Of these models, most were applied with a daily time step (65%). In addition to climate, the impacts of land use change and soil and water conservation techniques were considered in 13% and 17% of the studies, respectively.</p><p>Climate model output is an important source of uncertainty, therefore, we used the climate model ensemble size as a measure for uncertainty, assigning studies based on a larger climate model ensemble a larger weight in the estimation of the (weighted) median change in soil erosion under climate change. Soil erosion is projected to increase from near-century (+5% with respect to the reference period) to mid- and end-century (+17% and +15%, respectively). Soil erosion is projected to increase most in semi-arid (+23%) and humid continental climates (+20%), while soil erosion is projected to decrease in Mediterranean climates (-2%). Higher increase of soil erosion is projected for models that apply sub-daily (+26%) and daily time steps (+14%), than monthly (0%) and yearly time steps (+8%). Significantly different results were obtained between studies using bias-correction methods based on delta change (+9%) and quantile mapping (+37%). On the other hand, no significant differences were obtained between the emission scenarios. Our review further highlights that changes in land use or soil and water conservation measures can either mitigate (i.e. no tillage, agricultural abandonment, reforestation) or aggravate (i.e. agricultural expansion) the impacts of climate change. This review illustrates that most studies project an increase of soil erosion under future climate change, while environmental (e.g. climate, land use) and methodological (e.g. erosion model, bias-correction, climate ensemble) differences between studies determine the strength and significance of the projected impacts.</p><p>We acknowledge funding from the Spanish Ministry of Science, Inovation and Universities (PID2019-109381RB-I00/AEI/10.13039/501100011033).</p>

ClimateNet: an expert-labeled open dataset and deep learning architecture for enabling high-precision analyses of extreme weather

Identifying, detecting, and localizing extreme weather events is a crucial first step in understanding how they may vary under different climate change scenarios. Pattern recognition tasks such as classification, object detection, and segmentation (i.e., pixel-level classification) have remained challenging problems in the weather and climate sciences. While there exist many empirical heuristics for detecting extreme events, the disparities between the output of these different methods even for a single event are large and often difficult to reconcile. Given the success of deep learning (DL) in tackling similar problems in computer vision, we advocate a DL-based approach. DL, however, works best in the context of supervised learning – when labeled datasets are readily available. Reliable labeled training data for extreme weather and climate events is scarce. We create “ClimateNet” – an open, community-sourced human-expert-labeled curated dataset that captures tropical cyclones (TCs) and atmospheric rivers (ARs) in high-resolution climate model output from a simulation of a recent historical period. We use the curated ClimateNet dataset to train a state-of-the-art DL model for pixel-level identification – i.e., segmentation – of TCs and ARs. We then apply the trained DL model to historical and climate change scenarios simulated by the Community Atmospheric Model (CAM5.1) and show that the DL model accurately segments the data into TCs, ARs, or “the background” at a pixel level. Further, we show how the segmentation results can be used to conduct spatially and temporally precise analytics by quantifying distributions of extreme precipitation conditioned on event types (TC or AR) at regional scales. The key contribution of this work is that it paves the way for DL-based automated, high-fidelity, and highly precise analytics of climate data using a curated expert-labeled dataset – ClimateNet. ClimateNet and the DL-based segmentation method provide several unique capabilities: (i) they can be used to calculate a variety of TC and AR statistics at a fine-grained level; (ii) they can be applied to different climate scenarios and different datasets without tuning as they do not rely on threshold conditions; and (iii) the proposed DL method is suitable for rapidly analyzing large amounts of climate model output. While our study has been conducted for two important extreme weather patterns (TCs and ARs) in simulation datasets, we believe that this methodology can be applied to a much broader class of patterns and applied to observational and reanalysis data products via transfer learning.

A climatology of tropical wind shear produced by clustering wind profiles from a climate model

A procedure for producing a climatology of tropical wind shear from climate-model output is presented. The procedure is designed to find grid columns in the model where the organization of convection may be present. The climate-model output consists of east–west and north–south wind profiles at 20 equally spaced pressure levels from 1000 hPa to 50 hPa, and the Convective Available Potential Energy (CAPE) as diagnosed by the model’s Convection Parametrization Scheme (CPS). The procedure begins by filtering the wind profiles based on their maximum shear, and on a CAPE threshold of 100 J kg−1. The filtered profiles are normalized using the maximum wind speed at each pressure level, and rotated to align the wind at 850 hPa. From each of the filtered profiles, a sample has been produced with 40 dimensions (20 for each wind direction). The number of dimensions is reduced by using Principal Component Analysis (PCA), where the requirement is that 90 % of the variance must be explained by the principal components. This requires keeping the first seven leading principal components. The samples, as represented by their principal components, can then be clustered using the K-Means Clustering Algorithm (KMCA). 10 clusters are chosen to represent the samples, and the median of each cluster defines a Representative Wind Profile (RWP) – a profile that represents the shear conditions of the wind profiles produced by the climate model. The RWPs are analysed, first in terms of their vertical structure, and then in terms of their geographical and temporal distributions. We find that the RWPs have some features often associated with the organization of convection, such as low-level and mid-level shear. Some of the RWPs can be matched with wind profiles taken from case studies of organization of convection, such as squall lines seen in Tropical Ocean Global Atmosphere, Coupled Atmosphere Ocean Research Experiment (TOGA–COARE). The RWPs’ geographical distributions show that each RWP occurs preferentially in certain regions. Six of the RWPs occur preferentially over land, while three occur preferentially over oceans. The temporal distribution of RWPs shows that they occur preferentially at certain times of the year, with the distributions having mainly one or two modes. Their geographical and temporal distributions are compared with those seen in previous studies of organized convection, and some broad and specific similarities are noted. By performing the analysis on climate-model output, we lay the foundations for the development of the representation of shear-induced organization in a CPS. This would use the same methodology to diagnose where the organization of convection occurs, and modify the CPS in an appropriate manner to represent it.

The Arctic Ocean Observation Operator for 6.9 GHz (ARC3O) – Part 1: How to obtain sea ice brightness temperatures at 6.9 GHz from climate model output

We explore the feasibility of an observation operator producing passive microwave brightness temperatures for sea ice at a frequency of 6.9 GHz. We investigate the influence of simplifying assumptions for the representation of sea ice vertical properties on the simulation of microwave brightness temperatures. We do so in a one-dimensional setup, using a complex 1D thermodynamic sea ice model and a 1D microwave emission model. We find that realistic brightness temperatures can be simulated in cold conditions from a simplified linear temperature profile and a simplified salinity profile as a function of depth in the ice. These realistic brightness temperatures can be obtained based on profiles interpolated to as few as five layers. Most of the uncertainty resulting from the simplifications is introduced by the simplification of the salinity profiles. In warm conditions, the simplified salinity profiles lead to brine volume fractions that are too high in the subsurface layer. To overcome this limitation, we suggest using a constant brightness temperature for the ice during warm conditions and treating melt ponds as water surfaces. Finally, in our setup, we cannot assess the effect of wet snow properties. As periods of snow with intermediate moisture content, typically occurring in spring and fall, locally last for less than a month, our approach allows one to estimate realistic brightness temperatures at 6.9 GHz from climate model output for most of the year.