Search

The Digital Earth project aimed for the integration of data science and Earth science. Here, we reflect on the main lessons learned that include the need for interdisciplinary collaboration, thinking out of the box, the concept of ‘thinking in workflows’ and models for the sustainable implementation of scientific software, data infrastructure and policies.

Digital Earth is a project funded by the German Helmholtz Association with all centers of the research field Earth and Environment involved. The main goal of the Digital Earth project is to develop and bundle data science methods in extendable and maintainable scientific workflows that enable natural scientists in collaboration with data scientists to achieve a deeper understanding of the Earth system. This has been achieved by developing solutions for data analysis and exploration with visual and computational approaches with data obtained in a SMART monitoring approach and modeling studies, accompanied by a continuous evaluation of the collaboration processes. In this chapter, the history, setup, and focus of the Digital Earth project are described.

Digital Earth is a project funded by the German Helmholtz Association with all centers of the research field Earth and Environment involved. The main goal of the Digital Earth project is to develop and bundle data science methods in extendable and maintainable scientific workflows that enable natural scientists in collaboration with data scientists to achieve a deeper understanding of the Earth system. This has been achieved by developing solutions for data analysis and exploration with visual and computational approaches with data obtained in a SMART monitoring approach and modelingstudies, accompanied by acontinuous evaluation of the collaboration processes. In this chapter, the history, setup, and focus of the Digital Earth project are described.

Data-driven science has turned into a fourth fundamental paradigm of performing research. Earth System Science, following a holistic approach in unraveling the complex network of processes and interactions shaping system Earth, particularly profits from embracing data-driven approaches next to observation and modeling. At the end, increasing digitalization of Earth sciences will lead to cultural transformation towards a Digital Earth Culture.

The Digital Earth project aimed for the integration of data science and Earth science. Here, we reflect on the main lessons learned that include the need for interdisciplinary collaboration, thinking out of the box, the concept of ‘thinking in workflows’ and models for the sustainable implementation of scientific software, data infrastructure and policies.

Artificial intelligence and machine learning (ML) methods are increasinglyappliedinEarthsystemresearch,forimprovingdataanalysis,andmodelperformance,andeventuallysystemunderstanding.IntheDigitalEarthproject,severalML approaches have been tested and applied, and are discussed in this chapter. These include data analysis using supervised learning and classification for detection of river levees and underwater ammunition; process estimation of methane emissions andforenvironmentalhealth;point-to-spaceextrapolationofvaryingobservedquantities; anomaly and event detection in spatial and temporal geoscientific datasets. We present the approaches and results, and finally, we provide some conclusions on the broad applications of these computational data exploration methods and approaches.

The 2015/2016 Arctic winter was one of the coldest stratospheric winters in recent years. A stable vortex formed by early December and the early winter was exceptionally cold. Cold pool temperatures dropped below the nitric acid trihydrate (NAT) existence temperature of about 195 K, thus allowing polar stratospheric clouds (PSCs) to form. The low temperatures in the polar stratosphere persisted until early March, allowing chlorine activation and catalytic ozone destruction. Satellite observations indicate that sedimentation of PSC particles led to denitrification as well as dehydration of stratospheric layers. Model simulations of the 2015/2016 Arctic winter nudged toward European Centre for Medium-Range Weather Forecasts (ECMWF) analysis data were performed with the atmospheric chemistry–climate model ECHAM5/MESSy Atmospheric Chemistry (EMAC) for the Polar Stratosphere in a Changing Climate (POLSTRACC) campaign. POLSTRACC is a High Altitude and Long Range Research Aircraft (HALO) mission aimed at the investigation of the structure, composition and evolution of the Arctic upper troposphere and lower stratosphere (UTLS). The chemical and physical processes involved in Arctic stratospheric ozone depletion, transport and mixing processes in the UTLS at high latitudes, PSCs and cirrus clouds are investigated. In this study, an overview of the chemistry and dynamics of the 2015/2016 Arctic winter as simulated with EMAC is given. Further, chemical–dynamical processes such as denitrification, dehydration and ozone loss during the 2015/2016 Arctic winter are investigated. Comparisons to satellite observations by the Aura Microwave Limb Sounder (Aura/MLS) as well as to airborne measurements with the Gimballed Limb Observer for Radiance Imaging of the Atmosphere (GLORIA) performed aboard HALO during the POLSTRACC campaign show that the EMAC simulations nudged toward ECMWF analysis generally agree well with observations. We derive a maximum polar stratospheric O3 loss of ∼ 2 ppmv or 117 DU in terms of column ozone in mid-March. The stratosphere was denitrified by about 4–8 ppbv HNO3 and dehydrated by about 0.6–1 ppmv H2O from the middle to the end of February. While ozone loss was quite strong, but not as strong as in 2010/2011, denitrification and dehydration were so far the strongest observed in the Arctic stratosphere in at least the past 10 years.

This paper presents the spatio-temporal distribution of NOy species at altitudes between 14 and 31 km as measured with the MIPAS-B instrument on the morning of 21 March 2003 in northern Scandinavia. At lower altitudes (below about 22 km), temperature variations, the distribution of ClONO2, and the tracer N2O reveal the dynamics through the edge of the late arctic polar vortex. At higher altitudes, continuous measurement before, during, and after sunrise provides information about photochemistry illustrating the evolution of the photochemically active gases NO2 and N2O5 around sunrise. The measured temporal evolution of NO2 and N2O5 is compared to box modelling that is run along backward calculated trajectories. While the comparison of measured and modelled N2O5 reveals significant differences, there is a good agreement between the model and observations for NO2 in terms of volume mixing ratios but the simulated decrease shortly after sunrise is underestimated compared to the measurements. The differences are attributed to the photolysis rates used in the box model calculations.

Trends of hydrogen chloride (HCl), chlorine nitrate (ClONO2), and hydrogen fluoride (HF) total column abundances above Kiruna (Northern Sweden, 67.84$^\circ\text{N}$, 20.41$^\circ\text{E}$) derived from nearly 14 years (1996–2009) of measurement and model data are presented. The measurements have been performed with a Bruker 120HR (later Bruker 125 HR) Fourier transform infrared (FTIR) spectrometer and the chemistry-transport model (CTM) used was KASIMA (KArlsruhe SImulation model of the Middle Atmosphere). The total column abundances of ClONO2 and HF calculated by KASIMA agree quite well with the FTIR measurements while KASIMA tends to underestimate the HCl columns. To calculate the long-term trends, a linear function combined with an annual cycle was fitted to the data using a least squares method. The precision of the resulting trends was estimated with the bootstrap resampling method. For HF, both model and measurements show a positive trend that seems to decrease in the last few years. This suggests a stabilisation of the HF total column abundance. Between 1996 and 2009, KASIMA simulates an increase of (+1.51±0.07) %/yr which exceeds the FTIR result of (+0.65±0.25) %/yr. The trends determined for HCl and ClONO2 are significantly negative over the time period considered here. This is expected because the emission of their precursors (chlorofluorocarbons and hydrochlorofluorocarbons) has been restricted in the Montreal Protocol in 1987 and its amendments and adjustments. The trend for ClONO2 from the FTIR measurements amounts to (−3.28±0.56)%/yr and the one for HCl to (−0.81±0.23)%/yr. KASIMA simulates a weaker decrease: For ClONO2, the result is (−0.90±0.10) %/yr and for HCl (−0.17±0.06) %/yr. Part of the difference between measurement and model data can be explained by sampling and the stronger annual cycle indicated by the measurements. There is a factor of about four between the trends of HCl and ClONO2 above Kiruna for both measurement and model data.

We present an original multi-spectrum fitting procedure to retrieve volume mixing ratio (VMR) profiles of carbonyl fluoride (COF2) from ground-based high resolution Fourier transform infrared (FTIR) solar spectra. The multi-spectrum approach consists of simultaneously combining, during the retrievals, all spectra recorded consecutively during the same day and with the same resolution. Solar observations analyzed in this study with the SFIT-2 v3.91 fitting algorithm correspond to more than 2900 spectra recorded between January 2000 and December 2007 at high zenith angles, with a Fourier Transform Spectrometer operated at the high-altitude International Scientific Station of the Jungfraujoch (ISSJ, 46.5° N latitude, 8.0° E longitude, 3580m altitude), Switzerland. The goal of the retrieval strategy described here is to provide information about the vertical distribution of carbonyl fluoride. The microwindows used are located in the υ1 or in the υ4 COF2 infrared (IR) absorption bands. Averaging kernel and eigenvector analysis indicates that our FTIR retrieval is sensitive to COF2 inversion between 17 and 30 km, with the major contribution to the retrieved information always coming from the measurement. Moreover, there was no significant bias between COF2 partial columns, total columns or VMR profiles retrieved from the two bands. For each wavenumber region, a complete error budget including all identified sources has been carefully established. In addition, comparisons of FTIR COF2 17–30 km partial columns with KASIMA and SLIMCAT 3-D CTMs are also presented. If we do not notice any significant bias between FTIR and SLIMCAT time series, KASIMA COF2 17–30 km partial columns are lower of around 25%, probably due to incorrect lower boundary conditions. For each times series, linear trend estimation for the 2000–2007 time period as well as a seasonal variation study are also performed and critically discussed. For FTIR and KASIMA time series, very low COF2 growth rates (0.4±0.2%/year and 0.3±0.2%/year, respectively) have been derived. However, the SLIMCAT data set gives a slight negative trend (−0.5±0.2%/year), probably ascribable to discontinuities in the meteorological data used by this model. We further demonstrate that all time series are able to reproduce the COF2 seasonal cycle, which main seasonal characteristics deduced from each data set agree quite well.

International audience; The Atmospheric Chemistry Experiment (ACE) satellite was launched on 12 August 2003. Its two instruments measure vertical profiles of over 30 atmospheric trace gases by analyzing solar occultation spectra in the ultraviolet/visible and infrared wavelength regions. The reservoir gases HNO3, ClONO2, and N2O5 are three of the key species provided by the primary instrument, the ACE Fourier Transform Spectrometer (ACE-FTS). This paper describes the ACE-FTS version 2.2 data products, including the N2O5 update, for the three species and presents validation comparisons with available observations. We have compared volume mixing ratio (VMR) profiles of HNO3, ClONO2, and N2O5 with measurements by other satellite instruments (SMR, MLS, MIPAS), aircraft measurements (ASUR), and single balloon-flights (SPIRALE, FIRS-2). Partial columns of HNO3 and ClONO2 were also compared with measurements by ground-based Fourier Transform Infrared (FTIR) spectrometers. Overall the quality of the ACE-FTS v2.2 HNO3 VMR profiles is good from 18 to 35 km. For the statistical satellite comparisons, the mean absolute differences are generally within ±1 ppbv (±20%) from 18 to 35 km. For MIPAS and MLS comparisons only, mean relative differences lie within ±10% between 10 and 36 km. ACE-FTS HNO3 partial columns (~15–30 km) show a slight negative bias of -1.3% relative to the ground-based FTIRs at latitudes ranging from 77.8° S–76.5° N. Good agreement between ACE-FTS ClONO2 and MIPAS, using the Institut für Meteorologie und Klimaforschung and Instituto de Astrofisica de Andalucia (IMK-IAA) data processor is seen. Mean absolute differences are typically within ±0.01 ppbv between 16 and 27 km and less than +0.09 ppbv between 27 and 34 km. The ClONO2 partial column comparisons show varying degrees of agreement, depending on the location and the quality of the FTIR measurements. Good agreement was found for the comparisons with the midlatitude Jungfraujoch partial columns for which the mean relative difference is 4.7%. ACE-FTS N2O5 has a low bias relative to MIPAS IMK-IAA, reaching -0.25 ppbv at the altitude of the N2O5 maximum (around 30 km). Mean absolute differences at lower altitudes (16–27 km) are typically -0.05 ppbv for MIPAS nighttime and ±0.02 ppbv for MIPAS daytime measurements.

International audience; The Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) instrument was launched aboard the environmental satellite ENVISAT into its sun-synchronous orbit on 1 March 2002. The short-lived species NO2 is one of the key target products of MIPAS that are operationally retrieved from limb emission spectra measured in the stratosphere and mesosphere. Within the MIPAS validation activities, a large number of independent observations from balloons, satellites and ground-based stations have been compared to European Space Agency (ESA) version 4.61 operational NO2 data comprising the time period from July 2002 until March 2004 where MIPAS measured with full spectral resolution. Comparisons between MIPAS and balloon-borne observations carried out in 2002 and 2003 in the Arctic, at mid-latitudes, and in the tropics show a very good agreement below 40 km altitude with a mean deviation of roughly 3%, virtually without any significant bias. The comparison to ACE satellite observations exhibits only a small negative bias of MIPAS which appears not to be significant. The independent satellite instruments HALOE, SAGE II, and POAM III confirm in common for the spring-summer time period a negative bias of MIPAS in the Arctic and a positive bias in the Antarctic middle and upper stratosphere exceeding frequently the combined systematic error limits. In contrast to the ESA operational processor, the IMK/IAA retrieval code allows accurate inference of NO2 volume mixing ratios under consideration of all important non-LTE processes. Large differences between both retrieval results appear especially at higher altitudes, above about 50 to 55 km. These differences might be explained at least partly by non-LTE under polar winter conditions but not at mid-latitudes. Below this altitude region mean differences between both processors remain within 5% (during night) and up to 10% (during day) under undisturbed (September 2002) conditions and up to 40% under perturbed polar night conditions (February and March 2004). The intercomparison of ground-based NDACC observations shows no significant bias between the FTIR measurements in Kiruna (68° N) and MIPAS in summer 2003 but larger deviations in autumn and winter. The mean deviation over the whole comparison period remains within 10%. A mean negative bias of 15% for MIPAS daytime and 8% for nighttime observations has been determined for UV-vis comparisons over Harestua (60° N). Results of a pole-to-pole comparison of ground-based DOAS/UV-visible sunrise and MIPAS mid-morning column data has shown that the mean agreement in 2003 falls within the accuracy limit of the comparison method. Altogether, it can be indicated that MIPAS NO2 profiles yield valuable information on the vertical distribution of NO2 in the lower and middle stratosphere (below about 45 km) during day and night with an overall accuracy of about 10–20% and a precision of typically 5–15% such that the data are useful for scientific studies. In cases where extremely high NO2 occurs in the mesosphere (polar winter) retrieval results in the lower and middle stratosphere are less accurate than under undisturbed atmospheric conditions.

Altitude profiles of ClONO2 retrieved with the IMK (Institut für Meteorologie und Klimaforschung) science-oriented data processor from MIPAS/Envisat (Michelson Interferometer for Passive Atmospheric Sounding on Envisat) mid-infrared limb emission measurements between July 2002 and March 2004 have been validated by comparison with balloon-borne (Mark IV, FIRS2, MIPAS-B), airborne (MIPAS-STR), ground-based (Spitsbergen, Thule, Kiruna, Harestua, Jungfraujoch, Izaña, Wollongong, Lauder), and spaceborne (ACE-FTS) observations. With few exceptions we found very good agreement between these instruments and MIPAS with no evidence for any bias in most cases and altitude regions. For balloon-borne measurements typical absolute mean differences are below 0.05 ppbv over the whole altitude range from 10 to 39 km. In case of ACE-FTS observations mean differences are below 0.03 ppbv for observations below 26 km. Above this altitude the comparison with ACE-FTS is affected by the photochemically induced diurnal variation of ClONO2. Correction for this by use of a chemical transport model led to an overcompensation of the photochemical effect by up to 0.1 ppbv at altitudes of 30–35 km in case of MIPAS-ACE-FTS comparisons while for the balloon-borne observations no such inconsistency has been detected. The comparison of MIPAS derived total column amounts with ground-based observations revealed no significant bias in the MIPAS data. Mean differences between MIPAS and FTIR column abundances are 0.11±0.12×1014 cm−2 (1.0±1.1%) and −0.09±0.19×1014 cm−2 (−0.8±1.7%), depending on the coincidence criterion applied. χ2 tests have been performed to assess the combined precision estimates of MIPAS and the related instruments. When no exact coincidences were available as in case of MIPAS – FTIR or MIPAS – ACE-FTS comparisons it has been necessary to take into consideration a coincidence error term to account for χ2 deviations. From the resulting χ2 profiles there is no evidence for a systematic over/underestimation of the MIPAS random error analysis.