Search

Around 80 % of global soybean supply is produced in southeast South America (SESA), central Brazil (CB) and the United States (US) alone. This concentration of production in few regions makes global soybean supply sensitive to spatially compounding harvest failures. Weather variability is a key driver of soybean variability, with soybeans being especially vulnerable to hot and dry conditions during the reproductive growth stage in summer. El Niño–Southern Oscillation (ENSO) teleconnections can influence summer weather conditions across the Americas, presenting potential risks for spatially compounding harvest failures. Here, we develop causal structural models to quantify the influence of ENSO on soybean yields via mediating variables like local weather conditions and extratropical sea surface temperatures (SSTs). We show that soybean yields are predominately driven by soil moisture conditions in summer, explaining ∼50 %, 18 % and 40 % of yield variability in SESA, CB and the US respectively. Summer soil moisture is strongly driven by spring soil moisture, as well as by remote extratropical SST patterns in both hemispheres. Both of these soil moisture drivers are again influenced by ENSO. Our causal models show that persistent negative ENSO anomalies of −1.5 standard deviation (SD) lead to a −0.4 SD soybean reduction in the US and SESA. When spring soil moisture and extratropical SST precursors are pronouncedly negative (−1.5 SD), then estimated soybean losses increase to −0.9 SD for the US and SESA. Thus, by influencing extratropical SSTs and spring soil moisture, persistent La Niñas can trigger substantial soybean losses in both the US and SESA, with only minor potential gains in CB. Our findings highlight the physical pathways by which ENSO conditions can drive spatially compounding events. Such information may increase preparedness against climate-related global soybean supply shocks.

Soy harvest failure events can severely impact farmers, insurance companies and raise global prices. Reliable seasonal forecasts of mis-harvests would allow stakeholders to prepare and take appropriate early action. However, especially for farmers, the reliability and lead-time of current prediction systems provide insufficient information to justify within-season adaptation measures. Recent innovations increased our ability to generate reliable statistical seasonal forecasts. Here, we combine these innovations to predict the 1-3 poor soy harvest years in eastern US. We first use a clustering algorithm to spatially aggregate crop producing regions within the eastern US that are particularly sensitive to hot-dry weather conditions. Next, we use observational climate variables (sea surface temperature (SST) and soil moisture) to extract precursor timeseries at multiple lags. This allows the machine learning model to learn the low-frequency evolution, which carries important information for predictability. A selection based on causal inference allows for physically interpretable precursors. We show that the robust selected predictors are associated with the evolution of the horseshoe Pacific SST pattern, in line with previous research. We use the state of the horseshoe Pacific to identify years with enhanced predictability. We achieve very high forecast skill of poor harvests events, even 3 months prior to sowing, using a strict one-step-ahead train-test splitting. Over the last 25 years, 82% of the in February predicted poor harvests were correct. When operational, this forecast would enable farmers (and insurance/trading companies) to make informed decisions on adaption measures, e.g., selecting more drought-resistant cultivars, invest in insurance, change planting management.

Reliable (sub) seasonal (S2S) forecasts remain a huge scientific challenge. The lead-time is too long to benefit from the atmosphere’s inertial memory, but too short for the atmosphere’s boundary conditions to be felt strongly. Only for specific "windows of predictability" (i.e. specific regions, timescales and climatic background states), skillful forecasts are possible, in an otherwise largely unpredictable future. Due to a number of successes in S2S forecasting, the interest in machine learning (ML) is growing fast. However, we argue there is a need for more standardization, consensus on best practices, higher efficiency, and higher reproducibility. Typical S2S ML use-cases, such as (1) pure statistical forecasting based on observations, (2) transfer learning, and (3) post-processing of dynamical model ensembles, require a large coding and preprocessing effort. Such experiments are not trivial to set up, and without sufficient experience and expertise there is a large risk of improper cross-validation and/or improper and non-standard verification.Within a 3-year project, we are developing a high-level Python package called s2spy. Our aim is to make ML workflows more transparent and easier to build, and to facilitate standardization and collaboration across the S2S community. s2spy also contributes to a higher reproducibility and works towards a wider acceptance of standards and best practices. We will present our vision and the capabilities of our package, show-casing that we can build a model from raw climate data up to verification and explanation in only a few lines of code.

Many factors have been suggested to explain variability and change in Sahel rainfall. Of those, sea surface temperature (SST) in the Eastern Mediterranean Sea (EMS) and zonal moisture flux south of the Sahel show strong correlations. Based on observational and reanalysis data on temperature, pressure, wind and moisture flux, this paper identifies a mechanism that explains both correlations. The mechanism hinges on the Jebel Marra massif and the Ethiopian highlands, where the mesoscale convective systems (MCSs) develop that bring most of the rain to the Sahel. We find that cold SST anomalies in the EMS between June and September cause a greater trans-Sahara temperature contrast and coincide with high pressure over Libya, resulting in stronger northerlies towards Sudan. This prevents Tropical Atlantic moisture from reaching the MCS genesis region, which reduces the seasonal northward spread of Sahel rainfall and of the Atlantic intertropical convergence zone, which in turn suppresses the development of the west-African westerly jet and the African westerly jet and inhibits Atlantic moisture from reaching the MCS genesis region, thus further reducing Sahel rainfall. Anomalous moisture transport from the Mediterranean does not play a role. Mediterranean SST variability raises questions about the future development of Sahel rainfall. If a new dry period materialises, this will have substantial implications on food production in the region. There are however opportunities for mitigating against the effects of such a dry period.

Persistent hot–dry or cold–wet summer weather can have significant impacts on agriculture, health, and the environment. For northwestern Europe, these weather regimes are typically linked to, respectively, blocked or zonal jet stream states. The fundamental dynamics underlying these circulation states are still poorly understood. Edward Lorenz postulated that summer circulation may be either fully or almost intransitive, implying that part of the phase space (capturing circulation variability) cannot be reached within one specific summer. If true, this would have major implications for the predictability of summer weather and our understanding of the drivers of interannual variability of summer weather. Here, we test the two Lorenz hypotheses (i.e., fully or almost intransitive) for European summer circulation, capitalizing on a newly available very large ensemble (2000 years) of present-day climate data in the fully coupled global climate model EC-Earth. Using self-organizing maps, we quantify the phase space of summer circulation and the trajectories through phase space in unprecedented detail. We show that, based on Markov assumptions, the summer circulation is strongly dependent on its initial state in early summer with the atmospheric memory ranging from 28 days up to ~45 days. The memory is particularly long if the initial state is either a blocked or a zonal flow state. Furthermore, we identify two groups of summers that are characterized by distinctly different trajectories through phase space, and that prefer either a blocked or zonal circulation state, respectively. These results suggest that intransitivity is indeed a fundamental property of the atmosphere and an important driver of interannual variability.

Global-scale af-/reforestation (A/R) and deforestation substantially changes the Earth’s energy and water fluxes, thereby affecting the large-scale atmospheric circulation and thus have significant impacts on weather systems. During summer, A/R and deforestation induced changes in the soil moisture are shown to have an impact on the planetary wave response through the jet stream. Such changes might lead to high-amplitude, quasi-stationary circumglobal Rossby waves that have been associated with extreme summer heatwaves and persistent high-impact extremes. In this study we investigate how idealized global land use and land management changes can alter the boreal summer circulation with a focus on the response of the jet stream. For the analysis we conducted model experiments with three fully coupled Earth System Models (EC-EARTH, MPI-ESM and CESM). Each scenario run for 160 years from which we analyze the final 30 years. A control run with constant current land use and land management is compared to a global A/R and a global deforestation (global cropland expansion) simulation. In order to assess clean land-atmosphere interactions, all simulations are kept with constant present-day atmospheric forcings (year 2014). We investigate the potential changes in the amplitude of the waves, the likelihood of quasi-stationary wave activity, and of summer blockings within the three different simulations, and the weather consequences that such changes lead to.

High-amplitude quasi-stationary atmospheric Rossby waves with zonal wave numbers 6 to 8 associated with the phenomenon of quasi-resonant amplification (QRA) have been linked to persistent summer extreme weather events in the Northern Hemisphere. We project future occurrence of QRA events based on an index derived from the zonally averaged surface temperature field, comparing results from CMIP5 and CMIP6 (Coupled Model Intercomparison Projects) climate projections. Under the scenarios analyzed, there is a general agreement among models, with most simulations projecting a substantial increase in QRA index. Larger increases are found among CMIP6-SSP585 (42 models, 46 realizations) models with 85% of models displaying a positive trend, as compared with as compared with 60% of CMIP5-RCP85 (35 models, 75 realizations), and a reduced spread among SSP585 models. The CMIP6-SSP370 (24 models, 28 realizations) simulations display qualitatively similar behavior to SSP585, indicating a substantial increase in QRA events under business-as-usual emissions scenarios. Our analysis suggests that anthropogenic warming will likely lead to an even more substantial increase in QRA events (and associated summer weather extremes) than our previous analysis of CMIP5 simulations.

Land cover and land management changes (LCLMC) have often been highlighted as crucial regarding climate change, both for mitigation (e.g. afforestation) and adaptation (e.g. irrigation). In order to understand this role we present fully coupled Earth System Model (ESM) simulations using external forcing conditions from the SSP1-1.9 scenario, except for land cover and land management scenarios that follow differing trajectories. First we conduct a short 30-year historical simulation (histCTL) and a future (years 2015-2100) simulation under SSP1-1.9 conditions but with present day land cover kept at constant end of 2014 conditions (futCTL). These allow us to isolate climate changes in response to the SSP1-19 forcing, but in the absence of land cover changes. Secondly we conduct two simulations under SSP1-1.9 forcing, but with land cover and land management following two different trajectories. These trajectories are derived from the scenarios presented in Humpenöder et al. (2022) and represent two strongly diverging worlds with regard to socio-economic development, environmental protection, and land-based mitigation: (i) the future sustainability scenario (futSust) in which the land sector experiences sustainable development and application of mitigation strategies (such as greenhouse gas emission pricing) in all countries, (ii) the future inequality scenario (futIneq) in which these developments mostly happen in OECD countries, with the rest of the world continuing on current trends (including massive tropical deforestation). Each of these simulations have been run with three different ESMs (CESM, MPI-ESM and EC-EARTH) in order to identify how robust these results are over different ESMs.The results of these simulations can be used to increase our understanding of the role of land cover scenarios within a low-warming future as prescribed by the Paris agreement. We can compare the effects of all other forcings (futCTL- histCTL; CO2, aerosols etc.) to the effects of land cover changes in the different scenarios (futSust – futCTL or futIneq-futCTL) as well as to the difference between the future sustainability and the inequality narratives (futSust-futIneq). These results will be analysed for temperature and moisture fluxes, mainly focusing on warm and dry extremes and how land cover scenarios affect these. ReferencesHumpenöder, F., Popp, A., Schleussner, C. F., Orlov, A., Windisch, M. G., Menke, I., Pongratz, J., Havermann, F., Thiery, W., Luo, F., Jeetze, P. V., Philipp Dietrich, J., Lotze-Campen, H., Weindl, I. & Lejeune, Q. (2022). Overcoming global inequality is critical for land-based mitigation in line with the Paris Agreement. Nature Communications, 13(1), 1-15.

Recent work has shown that (sub-)seasonal variability in tropical Pacific convection, closely linked to ENSO, relates to summertime circulation over the Euro-Atlantic. The teleconnection is non-stationary, probably due to long-term changes in both the tropical Pacific and extra-tropical Atlantic. It also appears imperfectly captured by numerical models. In a previous study we found that the best predictor of errors in sub-seasonal forecasts of European temperature, is a dipole in tropical west Pacific sea surface temperatures (SSTs). In this diagnostic study we use reanalysis data to further investigate the teleconnection pathway and the processes behind its non-stationarity. We show that SST gradients associated with the dipole represent a combination of ENSO variability and west Pacific warming, and have become stronger since 1980. Associated patterns of suppressed and enhanced tropical heating are followed by quasi-stationary waves that linger for multiple weeks. Situations with La Niña-like gradients are followed by high pressure centers over eastern Europe and Russia, three to six weeks later. Inverted situations are followed by high pressure over western Europe, three to six weeks later. The latter situation is however also conditional on a strong meridional tripole in north Atlantic SST and a co-located jet stream. Overall, the sub-seasonal pathway diagnosed in this study connects to patterns detected at seasonal scales, and confirms earlier findings that the summertime connectivity between the Pacific and Europe has shifted in recent decades. It also partly explains the increased occurrence of high sea level pressures and summer temperatures over the European continent.

Over the last 70 years, extreme heat has been increasing at global scale [1,2], with a rapid rate in several regions including Western Europe [3]. Climate models broadly capture heat trends globally [1], but exhibit systematically weaker extreme heat trends than observations in Western Europe [4-6], together with a weaker summer warming [7,8]. The causes of this mismatch, confirmed here by the analysis of 273 latest generation coupled climate simulations, among which only a handful of them overpass observed trends, are not well understood. Here we use a circulation analogue approach [9,10] to identify the dynamical contribution to temperature trends [11-12], and show that a substantial fraction (1.0°C [0.4°-1.6°C] of 3.4°C per global warming degree) of the trend is due to circulation changes, largely due to increases in southerly flows over Western Europe. Their rapid increase in frequency (+63% [20%-106%] since 1950) and persistence (+45% [12%-77%]) are not captured by any of the 32 climate model flow simulations analyzed. The few simulations reaching the observed warming trends in extreme heat have weak dynamical changes, with a decrease in occurrence of southerly flows, indicating that they capture the warming trend for the wrong reasons. These model biases in circulation trends can be due to a systematically underestimated or erroneous representation of the circulation response to external forcing, or to a systematic underestimation of interdecadal variability, or both. The former implies that projections are too conservative, the latter that we are left with deep uncertainty regarding the pace of future summer heat in Europe: the current strong recent trend could weaken or increase in future decades. This calls for caution when interpreting climate projections of heat extremes over Western Europe, in particular in view of adaptation to heat waves.

The Last Interglacial (LIG; ca. 125 ka) is a period of interest for climate research as it is the most recent period of the Earth's history when the boreal climate was warmer than at present. Previous research, based on models and geological evidence, suggests that the LIG may have featured enhanced patterns of ocean storminess, but this remains hotly debated. Here, we apply state-of-the-art climate and hydrodynamic modeling to simulate changes in sea level extremes caused by storm surges, under LIG and pre-industrial climate forcings. Significantly higher seasonal LIG sea level extremes emerge for coastlines along northern Australia, the Indonesian archipelago, much of northern and eastern Africa, the Mediterranean Sea, the Gulf of Saint Lawrence, the Arabian Sea, the east coast of North America, and islands of the Pacific Ocean and of the Caribbean. Lower seasonal LIG sea level extremes emerge for coastlines along the North Sea, the Bay of Bengal, China, Vietnam, and parts of Central America. Most of these anomalies are associated with anomalies in seasonal sea level pressure minima and in eddy kinetic energy calculated from near-surface wind fields, and therefore seem to originate from anomalies in the meridional position and intensity of the predominant wind bands. In a qualitative comparison, LIG sea level extremes seem generally higher than those projected for future warmer climates. These results help to constrain the interpretation of coastal archives of LIG sea level indicators.

Around 80% of global soybean supply is produced in southeast South America (SESA), central Brazil (CB) and the United States (US) alone. This concentration of production in few regions makes global soybean supply sensitive to spatially compounding harvest failures. Weather variability is a key driver of soybean yield variability, with soybean especially vulnerable to hot and dry conditions during the reproductive growth stage in summer. El Niño Southern Oscillation (ENSO) teleconnections can influence summer weather conditions across the Americas presenting potential risks for spatially compounding harvest failures. Here, we develop causal structural models to quantify the influence of ENSO on crop yields via mediating variables like local weather conditions and extratropical sea-surface temperatures (SST). We show that soybean yields are predominately driven by soil moisture conditions in summer explaining ~50 %, 18 % and 40 % of yield variability in SESA, CB and US respectively. Summer soil moisture is strongly driven by spring soil moisture as well as remote extra-tropical SST patterns in both hemispheres. Both of these soil moisture drivers are again influenced by ENSO. Our causal models show that persistent negative ENSO anomalies of -1.5 standard deviation (SD) lead to a -0.4 SD soybean reductions in the US and SESA. When spring soil moisture and extratropical SST precursors are pronouncedly negative (-1.5 SD), then estimated soybean losses increase to -0.9 SD for US and SESA. Thus, by influencing extratropical SSTs and spring soil moisture, persistent La Niña’s can trigger substantial soybean losses in both the US and SESA, with only minor potential gains in CB. Our findings highlight the physical pathways by which ENSO conditions can drive spatially compounding events. Such information may increase preparedness against climate related global soybean supply shocks.

In recent years, East Africa (EA) has been affected by numerous drought events. To better understand the drought hazard and its implications, it is important to analyze the propagation of dry conditions through the hydrological system and identify at what point it leads to negative impacts on society. This way, we define drought events as dry conditions during which actual impacts are felt by communities (i.e. ‘impactful droughts’) rather than using a predefined relative threshold. This study explores the link between standardized and threshold-based drought indices and drought impacts in EA. Drought propagation from meteorological to soil moisture to hydrological drought to finally socio-economic impacts has been investigated in the data-rich regions around the world (i.e. Europe). However, not much has been done in data-poor regions (i.e. Africa), with very different climatic and socio-economic characteristics.We identify meaningful indicators for region- and sector-specific impact occurrences, focusing on three impact types: crop yield and livestock losses, changes in pasture represented by Normalized Drought Vegetative Index (NDVI), and people affected. In this study, we link these impact types to various drought hazard indices calculated from time series of precipitation, evaporation, soil moisture and discharge for every administrative unit in Kenya, Ethiopia and Somalia. The standardized indices are used with varying accumulation periods (1,3,6,12,24,48 months) to find the best predictors. We use independent but complementary machine learning techniques (correlation, logistic regression and random forest modelling) to identify which indices link best to which drought impacts.Using the variable threshold based analysis to analyse drought propagation in Isiolo county in Kenya, we find an increase in frequency and duration of hydrological droughts. This has been confirmed in other studies. The correlations and relationships between crop yield anomalies and indices are different in each of the administrative units. The findings in this research will lead to a greater understanding of the effects of drought hazards to communities, and contribute to improved drought monitoring and forecasting. This research can also form the basis for follow up work to explore the effects of adaptation options (i.e. irrigation) on drought hazard and impacts in East Africa.

Land cover and land management change (LCLMC) has been highlighted for its critical role in mitigation scenarios, both in terms of global mitigation and local adaptation. Yet, the climate effect of individual LCLMC options, their dependence on the background climate and the local vs. non-local responses are still poorly understood across different Earth system models (ESMs). Here we simulate the climatic effects of LCLMC using three state-of-the-art ESMs, including the Community Earth System Model (CESM), the Max Planck Institute for Meteorology Earth System Model (MPI-ESM) and the European Consortium Earth System Model (EC-EARTH). We assess the LCLMC effects using the following four idealized experiments: (i) a fully afforested world, (ii) a world fully covered by cropland, (ii) a fully afforested world with extensive wood harvesting and (iv) a full-cropland world with extensive irrigation. In these idealized sensitivity experiments, performed under present-day climate conditions, the effects of the different LCLMC strategies represent an upper bound for the potential of global mitigation and local adaptation. To disentangle the local and non-local effects from the LCLMC, a checkerboard-like LCLMC perturbation, i.e. alternating grid boxes with and without LCLMC, is applied. The local effects of deforestation on surface temperature are largely consistent across the ESMs and the observations, with a cooling in boreal latitudes and a warming in the tropics. However, the energy balance components driving the change in surface temperature show less consistency across the ESMs and the observations. Additionally, some biases exist in specific ESMs, such as a strong albedo response in CESM mid-latitudes and a soil-thawing-driven warming in boreal latitudes in EC-EARTH. The non-local effects on surface temperature are broadly consistent across ESMs for afforestation, though larger model uncertainty exists for cropland expansion. Irrigation clearly induces a cooling effect; however, the ESMs disagree whether these are mainly local or non-local effects. Wood harvesting is found to have no discernible biogeophysical effects on climate. Overall, our results underline the potential of ensemble simulations to inform decision making regarding future climate consequences of land-based mitigation and adaptation strategies., Earth System Dynamics, 14 (3), ISSN:2190-4987, ISSN:2190-4979

Much of the forecast skill in the mid-latitudes on seasonal timescales originates from deep convection in the tropical belt. For boreal summer, such tropical-extratropical teleconnections are less well understood as compared to winter. Here we validate the representation of boreal tropical – extratropical teleconnections in a general circulation model in comparison with observational data. To characterise variability between tropical convective activity and mid-latitude circulation, we identify the South Asian monsoon (SAM) – circumglobal teleconnection (CGT) pattern and the western North Pacific summer monsoon (WNPSM) – North Pacific high (NPH) pairs as the leading modes of tropical-extratropical coupled variability in both reanalysis (ERA5) and seasonal forecast (SEAS5) data. We calculate causal maps, an application of the PCMCI causal discovery algorithm which identifies causal links in a 2D field, to show the causal effect of each of these patterns on circulation and convection in the Northern Hemisphere. The spatial patterns and signs of the causal links in SEAS5 closely resemble those seen in ERA5, independent of the initialization date of SEAS5. However, the strength of causal links in SEAS5 is often too weak (about two thirds of those in ERA5). By performing a subsampling (over time) experiment, we identify those regions for which SEAS5 data well reproduce ERA5 values, e.g. South-eastern US, and highlight those where the bias is more prominent, e.g. North Africa. We demonstrate that different El Niño – Southern Oscillation phases have only a marginal effect on the strength of these links. Finally, we discuss the potential role of model mean-state biases in explaining differences between SEAS5 and ERA5 causal links.

Land management and anthropogenic land cover change (LMLCC) plays a key role in the global carbon budget. For example, approximately half of the terrestrial biomass has been removed by LMLCC to date. Conversely, large potentials for carbon dioxide removal are invoked when vegetation-based negative emission technologies such as afforestation are discussed. Previous studies on LMLCC effects on the carbon cycle focused on the direct effect of tree removal or regrowth on carbon fluxes. However, a suite of studies has shown that LMLCC has an important influence on climate via biogeophysical effects through changes in energy and water fluxes. This influence can reach far beyond the location of LMLCC, called the "nonlocal effect" of LMLCC on climate. This raises the question if LMLCC can also have non-negligible effects on the carbon cycle remote from the LMLCC location itself. Our study establishes the concept how to investigate strength and patterns of the unintended nonlocal side-effects of LMLCC on carbon stocks and fluxes.Therefore, we conducted three different fully-coupled atmosphere-ocean-land experiments of idealized global cropland expansion with and without cropland irrigation as well as global re-/afforestation starting from today's state over a 150-year period under present day solar and trace gas forcing. All experiments were simulated by three different earth system models (MPI-ESM, EC-EARTH and CESM) to additionally quantify inter-model uncertainty and potentially uncover specific model biases. Here only CESM and MPI-ESM results are presented. To separate the local and nonlocal effects we use a checkerboard approach of grid boxes with and without LMLCC as proposed by Winckler et al., 2017. That is, we separate the carbon stock changes due to LMLCC at the location of LMLCC (local effect) from those induced by climate change caused by remote LMLCC (nonlocal effect). The total effect is the sum of both, the local and nonlocal effect.The results of MPI-ESM (CESM) show that the global nonlocal effect on vegetation carbon (cVeg) accounts for 6% (3%) and 4% (0.6%) of total cVeg changes for crop expansion and afforestation simulation, respectively. Additionally, applying irrigation to crop expansion strongly increases the nonlocal climate induced cVeg change by 52% (610%) of total cVeg change for MPI-ESM (CESM). The nonlocal effect of regions with largest carbon changes exhibit partly much larger nonlocal/total ratio. For instance, the nonlocal cVeg change in the Congo basin after cropland expansion accounts for more than 30% of total cVeg change. Furthermore, in some regions, the nonlocal effect of cVeg can be opposite to the local effect, and may thus reduce the total effect of the LMLCC practice compared to what would be expected from the local effect alone.Overall, the results from MPI-ESM and CESM indicate that the nonlocal carbon effect is important in key regions and can even become globally important for the irrigation practices. In addition to local effects, these unintended nonlocal effects need to be considered when the impacts of a LMLCC practice on the entire carbon cycle (e.g., also with regard to a potential carbon dioxide removal method) will be assessed.

Changes in the Greenland Blocking Index (GBI) present a threat to both the mass loss of the Greenland ice sheet, potentially leading to increased sea level rise, and downstream regions, such as the European continent and Mediterranean Basin that may be affected by associated changes in the frequency and characteristics of weather patterns. While a consistent increase in the summer GBI has been detected in (several) observational and reanalysis datasets, global climate models and seasonal prediction systems so far have failed in capturing this historical trend. The question whether the observed trend is real and not reproduced by models or an artifact of multidecadal variability (and, hence, a transient phenomenon) has so far not been answered. Here we apply the Peter and Clark momentary conditional independence (PCMCI) causal discovery tool to (i) unravel essential causal drivers of the GBI in a set of selected atmospheric fields at intraseasonal (weekly and 3-day averages) time scales in observations and (ii) compare the latter with those found in the ECMWF SEAS5 seasonal forecasts. Observed causal links well reproduced by the dynamical model would be considered indicative of the trend in GBI found in observations more likely emerging because of the selected time frame. By contrast, any differences between the two sets of causal drivers in observations and models would support the hypothesis that some key mechanisms are not appropriately represented in the forecast model, preventing it from correctly reproducing the observed trend.

During boreal summer, large scale subsidence and a persistent northerly flow, known as the Etesians, characterize the tropospheric circulation over the eastern Mediterranean, respectively bringing clear skies and mitigating the emergence of heat waves. Atmospheric drivers over both South Asia and the North Atlantic have been proposed to influence the intraseasonal variability of subsidence and Etesians over the eastern Mediterranean. Here, we employ Causal Effect Networks, obtained by applying the Peter and Clark Momentary Conditional Independence (PCMCI) causal discovery algorithm, to identify causal precursors of subsidence and Etesians in a set of atmospheric fields. We find that both wave train activity over the North Atlantic/North American region and convective activity over the northern Indian Ocean are causally related to the 3-day average 850 hPa meridional wind variations over the eastern Mediterranean at a lag of 3-to-6 days. For 3-day average 500 hPa vertical wind velocity, causal precursors over the Middle East and Arabian Sea similar to those identified for the Etesian are found. We further explore in detail the different nature of the causal precursors to the Etesians and subsidence over the eastern Mediterranean by applying varying averaging in the variables representing the involved phenomena.

Sub-seasonal forecasts are challenging for numerical weather prediction (NWP) and machine learning models alike. Predicting temperature with a lead-time of two or more weeks requires a forward model to integrate multiple complex interactions, like oceanic and land surface conditions that might lead to recurrent or persistent weather patterns. The representation of the relevant interactions is imperfect in NWP models, just as our physical understanding of them. Model predictability can therefore deviate from real predictability for poorly understood reasons, hindering future progress.This paper combines NWP with machine learning to detect and resolve such imperfect representations. We post-process ECMWF extended range forecasts of high summer temperatures in Europe with a shallow artificial neural network (ANN). Predictors are objectively selected from a large set of atmospheric, oceanic and terrestrial sources of predictability from ERA-5 and ECMWF re-forecast output. In the proposed architecture, the ANN learns to ‘update’ a prior ECMWF-given probability of two-meter temperature exceeding a given threshold. Due to the architecture of the network the magnitude of each correction, like increasing underestimated probabilities, can be attributed to specific predictors at initialization- or forecast-time. We interpret the circumstances in which substantial corrections are made. This reveals, e.g., that a tropical west Pacific sea surface temperature pattern is connected to high monthly average European temperature at a two-week lead-time. This teleconnection pattern is underestimated by the dynamical model and by correcting for this bias the ANN-based post-processing can thus improve forecast skill. We further find that the method does not readily increase skill when applied to other combinations of lead-time, averaging period and threshold, possibly due to non-stationarity in the data, lack of real predictability or lack of a re-forecast set of sufficient length.

Land cover and land management changes (LCLMC) have often been highlighted as crucial regarding climate change mitigation (e.g., enhanced carbon uptake on land through afforestation), but their potential for adaptation has also been suggested (e.g., local cooling through irrigation). Regarding the latter, the effects of LCLMC on the climate remain uncertain. LCLMC can have strong implications on surface moisture fluxes and have even been linked to changes in large scale atmospheric circulation. Here, we study the effects of three LCLMC (i) global afforestation, (ii) global cropland expansion and (iii) large-scale irrigation extension on climate by employing three fully coupled Earth System Models (CESM, MPI-ESM, and EC-EARTH). Sensitivity simulations were performed under present-day conditions and extreme LCLMC, of which the effects on moisture fluxes and atmospheric circulation are investigated. We do this by first analyzing the surface moisture fluxes using monthly precipitation and evaporation data to perform a moisture convergence analysis, before performing a moisture tracking analysis with the Water Accounting Model (WAM-2 layers) , this model solves the atmospheric moisture balance and requires sub-daily data from the sensitivity experiments as an input.Here we focus on the results from CESM, cropland expansion has shown to cause an average shift southward of the Intertropical convergence zone as well as a weakening in westerlies strength and consequent decrease in moisture transport. This causes an increase in continental moisture sources over most of the Northern Hemisphere. Afforestation, in contrast, shows an average shift northward of the Intertropical convergence zone and enhanced westerlies and moisture transport. Lastly, irrigation expansion enhances the moisture convergence over areas where irrigation is applied, causing an increase in both precipitation and evapotranspiration.

Heatwaves can have devastating impact on society and reliable early warnings at several weeks lead time are needed. Previous studies showed that north-Pacific sea surface temperatures (SST) can provide long-lead predictability for eastern US temperature, mediated by an atmospheric Rossby wave. The exact mechanisms, however, are not well understood. Here we analyze two different Rossby waves associated with temperature variability in western and eastern US, respectively. Causal discovery analyses reveal that both waves are characterized by positive ocean-atmosphere feedbacks at daily timescales. Only for the eastern US, a long-lead causal link from SSTs to the Rossby wave exists, which generates summer temperature predictability. We show that this SST forcing mechanism originates from the evolution of the winter-to-spring Pacific Decadal Oscillation (PDO). During pronounced winter-to-spring PDO phases (either positive or negative) eastern US summer temperature forecast skill more than doubles, providing a temporary window of enhanced long-lead predictability.

Summer 2010 saw two simultaneous extremes linked by an atmospheric wave train: a record-breaking heatwave in Russia and severe floods in Pakistan. Here, we study this wave event using a large ensemble climate model experiment. First, we show that the circulation in 2010 reflected a recurrent wave train connecting the heatwave and flooding events. Second, we show that the occurrence of the wave train is favored by three drivers: (1) 2010 sea surface temperature anomalies increase the probability of this wave train by a factor 2-to-4 relative to the model’s climatology, (2) early-summer soil moisture deficit in Russia not only increases the probability of local heatwaves, but also enhances rainfall extremes over Pakistan by forcing an atmospheric wave response, and (3) high-latitude land warming favors wave-train occurrence and therefore rainfall and heat extremes. These findings highlight the complexity and synergistic interactions between different drivers, reconciling some seemingly contradictory results from previous studies.

Reliable subseasonal forecasts of high summer temperatures would be very valuable for society. Although state-of-the-art numerical weather prediction (NWP) models have become much better in representing the relevant sources of predictability like land and sea surface states, the subseasonal potential is not fully realized. Complexities arise because drivers depend on the state of other drivers and on interactions over multiple time scales. This study applies statistical modeling to ERA5 data, and explores how nine potential drivers, interacting on eight time scales, contribute to the subseasonal predictability of high summer temperatures in western and central Europe. Features and target temperatures are extracted with two variations of hierarchical clustering, and are fitted with a machine learning (ML) model based on random forests. Explainable AI methods show that the ML model agrees with physical understanding. Verification of the forecasts reveals that a large part of predictability comes from climate change, but that reliable and valuable subseasonal forecasts are possible in certain windows, like forecasting monthly warm anomalies with a lead time of 15 days. Contributions of each driver confirm that there is a transfer of predictability from the land and sea surface state to the atmosphere. The involved time scales depend on lead time and the forecast target. The explainable AI methods also reveal surprising driving features in sea surface temperature and 850 hPa temperature, and rank the contribution of snow cover above that of sea ice. Overall, this study demonstrates that complex statistical models, when made explainable, can complement research with NWP models, by diagnosing drivers that need further understanding and a correct numerical representation, for better future forecasts.