Laboratory for
Climatology and Remote Sensing

Abstract:

The lack of long term and well distributed precipitation observations on the Tibetan Plateau (TiP) with its complex terrain raises the need for other sources of precipitation data for this area. Satellite-based precipitation retrievals can fill those data gaps. Before precipitation rates can be retrieved from satellite imagery, the precipitating area needs to be classified properly. Here, we present a feasibility study of a precipitation area delineation scheme for the TiP based on multispectral data with data fusion from the geostationary orbit (GEO, Insat-3D and Elektro-L2) and a machine learning approach (Random Forest, RF). The GEO data are used as predictors for the RF model, extensively validated by independent GPM (Global Precipitation Measurement Mission) IMERG (Integrated Multi-satellitE Retrievals for GPM) gauge calibrated microwave (MW) best-quality precipitation estimates. To improve the RF model performance, we tested different optimization schemes. Here, we find that (1) using more precipitating pixels and reducing the amount of non-precipitating pixels during training greatly improved the classification results. The accuracy of the precipitation area delineation also benefits from (2) changing the temporal resolution into smaller segments. We particularly compared our results to the Infrared (IR) only precipitation product from GPM IMERG and found a markedly improved performance of the new multispectral product (Heidke Skill Score (HSS) of 0.19 (IR only) compared to 0.57 (new multispectral product)). Other studies with a precipitation area delineation obtained a probability of detection (POD) of 0.61, whereas our POD is comparable, with 0.56 on average. The new multispectral product performs best (worse) for precipitation rates above the 90th percentile (below the 10th percentile). Our results point to a clear strategy to improve the IMERG product in the absence of MW radiances.

Abstract:

In the Atacama Desert, cyanobacteria grow on various substrates such as soils (edaphic) and quartz or granitoid stones (lithic). Both edaphic and lithic cyanobacterial communities have been described but no comparison between both communities of the same locality has yet been undertaken. In the present study, we compared both cyanobacterial communities along a precipitation gradient ranging from the arid National Park Pan de Azúcar (PA), which resembles a large fog oasis in the Atacama Desert extending to the semiarid Santa Gracia Natural Reserve (SG) further south, as well as along a precipitation gradient within PA. Various microscopic techniques, as well as culturing and partial 16S rRNA sequencing, were applied to identify 21 cyanobacterial species; the diversity was found to decline as precipitation levels decreased. Additionally, under increasing xeric stress, lithic community species composition showed higher divergence from the surrounding edaphic community, resulting in indigenous hypolithic and chasmoendolithic cyanobacterial communities. We conclude that rain and fog water, respectively, cause contrasting trends regarding cyanobacterial species richness in the edaphic and lithic microhabitats.

Abstract:

The Atacama Desert is one of the driest and probably oldest deserts on Earth where only a few extremophile organisms are able to survive. This study investigated two terricolous and two epiphytic lichens from the fog oasis “Las Lomitas” within the National Park Pan de Azúcar which represents a refugium for a few vascular desert plants and many lichens that can thrive on fog and dew alone. Ecophysiological meas‐ urements and climate records were combined with molecular data of the mycobiont, their green algal photobionts and lichenicolous fungi to gain information about the ecology of lichens within the fog oasis. Phylogenetic and morphological investiga‐ tions led to the identification and description of the new lichen species Acarospora conafii sp. nov. as well as the lichenicolous fungi that accompanied them and revealed the trebouxioid character of all lichen photobionts. Their photosynthetic responses were compared during natural scenarios such as reactivation by high air humidity and in situ fog events to elucidate the activation strategies of this lichen community. Epiphytic lichens showed photosynthetic activity that was rapidly induced by fog and high relative air humidity whereas terricolous lichens were only activated by fog.

Abstract:

Hyperspectral remote sensing is a promising tool for a variety of applications including ecology, geology, analytical chemistry and medical research. This article presents the new hsdar package for R statistical software, which performs a variety of analysis steps taken during a typical hyperspectral remote sensing approach. The package introduces a new class for efficiently storing large hyperspectral data sets such as hyperspectral cubes within R. The package includes several important hyperspectral analysis tools such as continuum removal, normalized ratio indices and integrates two widely used radiation transfer models. In addition, the package provides methods to directly use the functionality of the caret package for machine learning tasks. Two case studies demonstrate the package’s range of functionality: First, plant leaf chlorophyll content is estimated and second, cancer in the human larynx is detected from hyperspectral data.

Abstract:

Tropical montane forests, particularly Andean rainforest, are important ecosystems for regional carbon and water cycles as well as for biological diversity and speciation. Owing to their remoteness, however, ecological key-processes are less understood as in the tropical lowlands. Remote sensing allows modeling of variables related to spatial patterns of carbon stocks and fluxes (e.g., biomass) and ecosystem functioning (e.g., functional leaf traits). However, at a landscape scale most studies conducted so far are based on airborne remote sensing data which is often available only locally and for one time-point. In contrast, multispectral satellites at moderate spectral and spatial resolutions are able to provide spatially continuous and repeated observations. Here, we investigated the effectiveness of Landsat-8 imagery in modeling tropical montane forest biomass, its productivity and selected canopy traits. Topographical, spectral and textural metrics were derived as predictors. To train and validate the models, in-situ data was sampled in 54 permanent plots in forests of southern Ecuador distributed within three study sites at 1000?m, 2000?m and 3000?m a.s.l. We used partial least squares regressions to model and map all response variables. Along the whole elevation gradient biomass and productivity models explained 31%, 43%, 69% and 63% of variance in aboveground biomass, annual wood production, fine litter production and aboveground net primary production, respectively. Regression models of canopy traits measured as community weighted means explained 62%, 78%, 65% and 65% of variance in leaf toughness, specific leaf area, foliar N concentration, and foliar P concentration, respectively. Models at single study sites hardly explained variation in aboveground biomass and the annual wood production indicating that these measures are mainly determined by the change of forest types along with elevation. In contrast, the models of fine litter production and canopy traits explained between 8%–85% in variation depending on the study site. We found spectral metrics, in particular a vegetation index using the red and the green band to provide complementary information to topographical metrics. The model performances for estimating leaf toughness, biochemical canopy traits and related fine litter production all improved when adding spectral information. Our findings therefore revealed that differences in fine litter production and canopy traits in our study area are driven by local changes in vegetation edaphically induced by topography. We conclude that Landsat-derived metrics are useful in modeling fine litter production and biochemical canopy traits, in a topographically and ecologically complex tropical montane forest.

Abstract:

Fast Fog and Low Stratus Detection (FFLSD) is a processing tool enabling detection of fog and low stratus (FLS) on very large time-series of Meteosat Second Generation data. We combine approaches for FLS detection by day and night into one homogeneous and efficient processing tool. The main unification improvements are: (1) consistent spatial tiling instead of various pixel clustering approaches, (2) uniform generation of sun and viewing angle dependent thresholds for individual tiles, (3) flexible study areas instead of area dependence, (4) parallel raster processing using the Open Computing Language (OpenCL), (5) and efficient algorithms for complex processing steps like histogram generation and analysis. As a result, users are enabled to generate products like long-term FLS climatologies with reasonable resources and short processing times. FFLSD is available as open source and allows reuse and extensions for other tasks.

Abstract:

Fog plays a major role in many ecological aspects and it influences human life in various ways. In this study, a temporally highly resolved and spatially explicit anal- ysis of variations in fog occurrence was conducted for Europe and links to general weather conditions were investigated. To this end, a high-resolution fog product based on Meteosat Second Generation data was developed. Characteristic fog distri- butions were identified by applying a Self Organizing Map approach to the dataset. It was found that the resulting fog patterns are primarily determined by terrain char- acteristics. Simultaneous occurrences between these patterns and the predominant general weather situations were analyzed. The results show that the general weather situations can be categorized into three main groups, each responsible for the forma- tion of a different group of fog patterns. Additionally, distinct regional differences could be identified in the diurnal and annual fog frequency cycles and the derived region-specific frequency variations were used to draw conclusions about the fog types prevailing in these regions.

Abstract:

The meteorological measurements in Brno, Czech Republic, is among the world's oldest measurements, operating since 1799. Like many others, station was initially installed in the city center, relocated several times, and currently operates at an airport outside the city. These geographical changes potentially bias the temperature record due to different station surroundings and varying degrees of urban heat island effects. Here, we assess the influence of land cover on spatial temperature variations in Brno, capitol of Moravia and the second largest city of the Czech Republic. We therefore use a unique dataset of half-hourly resolved measurements from 11 stations spanning a period of more than 3.5 years and apply this information to reduce relocation biases in the long-term temperature record from 1799 to the present. Regression analysis reveals a significant warming influence from nearby buildings and a cooling influence from vegetation, explaining up to 80{\%} of the spatial variability within our network. The influence is strongest during the warm season and for land cover changes between 300 and 500 m around stations. Relying on historical maps and recent satellite data, it was possible to capture the building densities surrounding the past locations of the meteorological station. Using the previously established land cover--temperature relation, the anthropogenic warming for each measurement site could be quantified and hence eliminated from the temperature record accordingly, thereby increasing the long-term warming trend.

Abstract:

With 450,000?km2Kobresia (syn. Carex) pygmaea dominated pastures in the eastern Tibetan highlands are the world's largest pastoral alpine ecosystem forming a durable turf cover at 3000–6000?m?a.s.l. Kobresia's resilience and competitiveness is based on dwarf habit, predominantly below-ground allocation of photo assimilates, mixture of seed production and clonal growth, and high genetic diversity. Kobresia growth is co-limited by livestock-mediated nutrient withdrawal and, in the drier parts of the plateau, low rainfall during the short and cold growing season. Overstocking has caused pasture degradation and soil deterioration over most parts of the Tibetan highlands and is the basis for this man-made ecosystem. Natural autocyclic processes of turf destruction and soil erosion are initiated through polygonal turf cover cracking, and accelerated by soil-dwelling endemic small mammals in the absence of predators. The major consequences of vegetation cover deterioration include the release of large amounts of C, earlier diurnal formation of clouds, and decreased surface temperatures. These effects decrease the recovery potential of Kobresia pastures and make them more vulnerable to anthropogenic pressure and climate change. Traditional migratory rangeland management was sustainable over millennia, and possibly still offers the best strategy to conserve and possibly increase C stocks in the Kobresia turf.

Abstract:

Despite many eorts of the radar community, quantitative precipitation estimation (QPE) from weather radar data remains a challenging topic. The high resolution of X-band radar imagery in space and time comes with an intricate correction process of reflectivity. The steep and high mountain topography of the Andes enhances its complexity. This study aims to optimize the rainfall derivation of the highest X-band radar in the world (4450 m a.s.l.) by using a random forest (RF) model and single Plan Position Indicator (PPI) scans. The performance of the RFmodel was evaluated in comparison with the traditional step-wise approach by using both, the Marshall-Palmer and a site-specific Z–R relationship. Since rain gauge networks are frequently unevenly distributed and hardly available at real time in mountain regions, bias adjustment was neglected. Results showed an improvement in the step-wise approach by using the site-specific (instead of the Marshall-Palmer) Z–R relationship. However, both models highly underestimate the rainfall rate (correlation coecient < 0.69; slope up to 12). Contrary, the RF model greatly outperformed the step-wise approach in all testing locations and on dierent rainfall events (correlation coecient up to 0.83; slope = 1.04). The results are promising and unveil a dierent approach to overcome the high attenuation issues inherent to X-band radars.

Abstract:

The present investigation evaluates the accuracy of estimating above-ground biomass (AGB) by means of two dierent sensors installed onboard an unmanned aerial vehicle (UAV) platform (DJI Inspire I) because the high costs of very high-resolution imagery provided by satellites or light detection and ranging (LiDAR) sensors often impede AGB estimation and the determination of other vegetation parameters. The sensors utilized included an RGB camera (ZENMUSE X3) and a multispectral camera (Parrot Sequoia), whose images were used for AGB estimation in a natural tropical mountain forest (TMF) in Southern Ecuador. The total area covered by the sensors included 80 ha at lower elevations characterized by a fast-changing topography and dierent vegetation covers. From the total area, a core study site of 24 ha was selected for AGB calculation, applying two dierent methods. The firstmethod used the RGB images and applied the structure formotion (SfM) process to generate point clouds for a subsequent individual tree classification. Per the classification at tree level, tree height (H) and diameter at breast height (DBH) could be determined, which are necessary input parameters to calculate AGB (Mg ha 1) by means of a specific allometric equation for wet forests. The second method used the multispectral images to calculate the normalized dierence vegetation index (NDVI), which is the basis for AGB estimation applying an equation for tropical evergreen forests. The obtained results were validated against a previous AGB estimation for the same area using LiDAR data. The study found two major results: (i) The NDVI-based AGB estimates obtained by multispectral drone imagery were less accurate due to the saturation eect in dense tropical forests, (ii) the photogrammetric approach using RGB images provided reliable AGB estimates comparable to expensive LiDAR surveys (R2: 0.85). However, the latter is only possible if an auxiliary digital terrain model (DTM) in very high resolution is available because in dense natural forests the terrain surface (DTM) is hardly detectable by passive sensors due to the canopy layer, which impedes ground detection.

Abstract:

Weather radar networks are an excellent tool for quantitative precipitation estimation (QPE), due to their high resolution in space and time, particularly in remote mountain areas such as the Tropical Andes. Nevertheless, reduction of the temporal and spatial resolution might severely reduce the quality of QPE. Thus, the main objective of this study was to analyze the impact of spatial and temporal resolutions of radar data on the cumulative QPE. For this, data from the world’s highest X-band weather radar (4450 m a.s.l.), located in the Andes of Ecuador (Paute River basin), and from a rain gauge network were used. Dierent time resolutions (1, 5, 10, 15, 20, 30, and 60 min) and spatial resolutions (0.5, 0.25, and 0.1 km) were evaluated. An optical flow method was validated for 11 rainfall events (with dierent features) and applied to enhance the temporal resolution of radar data to 1-min intervals. The results show that 1-min temporal resolution images are able to capture rain event features in detail. The radar–rain gauge correlation decreases considerably when the time resolution increases (r from 0.69 to 0.31, time resolution from 1 to 60 min). No significant dierence was found in the rain total volume (3%) calculated with the three spatial resolution data. A spatial resolution of 0.5 km on radar imagery is suitable to quantify rainfall in the AndesMountains. This study improves knowledge on rainfall spatial distribution in the Ecuadorian Andes, and it will be the basis for future hydrometeorological studies.

Abstract:

Provisioning services from grassland ecosystems are strongly linked to physical and chemical grassland traits, which are affected by atmospheric CO2 concentrations ([CO2]s). The influences of increased [CO2]s ([eCO2]s) are typically investigated in Free Air Carbon dioxide Enrichment (FACE) studies via destructive sampling methods. This traditional approach is restricted to sampling plots and harvest dates, while hyperspectral approaches provide new opportunities as they are rapid, non-destructive and cost-effective. They further allow a high temporal resolution including spatially explicit information. In this study we investigated the hyperspectral predictability of 14 grassland traits linked to forage quality and quantity within a FACE experiment in central Germany with three plots under ambient atmospheric [CO2]s, and three plots at [eCO2]s (∼20% above ambient [CO2]s). We analysed the suitability of various normalisation and feature selection techniques to link comprehensive laboratory analyses with two years of hyperspectral measurements (spectral range 600–1600 nm). We applied partial least squares regression and found good to excellent predictive performances (0.49 ≤ leave one out cross-validation R2≤ 0.94), which depended on the normalisation method applied to the hyperspectral data prior to model training. Noteworthy, the models' predictive performances were not affected by the different [CO2]s, which was anticipated due to the altered plant physiology under [eCO2]s. Thus, an accurate monitoring of grassland traits under different [CO2]s (present-day versus future, or within a FACE facility) is promising, if appropriate predictors are selected. Moreover, we show how hyperspectral predictions can be used e.g., within a future phenotyping approach, to monitor the grassland on a spatially explicit level and on a higher temporal resolution compared to conventional destructive sampling techniques. Based on the information during the vegetation period we show how hyperspectral monitoring might be used e.g., to adapt harvest practices or gain deeper insights into physiological plant alterations under [eCO2]s.

Abstract:

Information about clouds is important for observing and predicting weather and climate as well as for generating and distributing solar power. Most existing approaches extract cloud information from satellite data by classifying individual pixels instead of using closely integrated spatial information, ignoring the fact that clouds are highly dynamic, spatially continuous entities. This paper proposes a novel cloud classification method based on deep learning. Relying on a Convolutional Neural Network (CNN) architecture for image segmentation, the presented Cloud Segmentation CNN (CS-CNN), classifies all pixels of a scene simultaneously rather than individually. We show that CS-CNN can successfully process multispectral satellite data to classify continuous phenomena such as highly dynamic clouds. The proposed approach produces excellent results on Meteosat Second Generation (MSG) satellite data in terms of quality, robustness, and runtime compared to other machine learning methods such as random forests. In particular, comparing CS-CNN with the CLAAS-2 cloud mask derived from MSG data shows high accuracy (0.94) and Heidke Skill Score (0.90) values. In contrast to a random forest, CS-CNN produces robust results and is insensitive to challenges created by coast lines and bright (sand) surface areas. Using GPU acceleration, CS-CNN requires only 25 ms of computation time for classification of images of Europe with 508  508 pixels.

Abstract:

Atmospheric carbon (CO2) exchange, evapotranspiration (ET) processes, and their interactions with climatic drivers across tropical alpine grasslands are poorly understood. This lack of understanding is particularly evident for the páramo, the highest vegetated frontier in the northern Andes, the main source of water for inter-Andean cities, and a large carbon storage area. Studies of CO2 and ET fluxes via the standard Eddy Covariance (EC) technique have never been applied to this region, limiting the understanding of diurnal / nocturnal exchanges and budget estimations. In this paper, we report the first EC analysis conducted on the Andean páramo (3765?m a.s.l.); this analysis measured CO2, ET, and micrometeorological variables over two years (2016–2018) to understand their interactions with climatic / biophysical controls. The páramo was found to be a source of CO2 and exhibited a net positive exchange (mean = +99?±?30 gC m?2 per year). The light-responses of net CO2 exchange and the primary productivity were correlated and model-parameterized. Evapotranspiration was 635?±?9?mm per year (51% of the annual rainfall total), and we obtained crop coefficients for the dominant vegetation (Tussock grass) based on reference-ET models FAO56 and ASCE-ERWI (0.90 and 0.78, respectively). We also compared our results to those from other high-altitude (alpine) and high-latitude grasslands (tundra). Finally, we demonstrate that our measurement period is representative of the páramo’s longer-term climate dynamics. Our investigation contributes to the body of knowledge on the land surface-atmosphere processes of the tropical Andes and supports decision-making about ecosystem services management and the preservation of this vulnerable biome.

Abstract:

In tropical agriculture, the vigorously growing Bracken fern causes severe problems by invading pastures and out-competing the common pasture grasses. Due to infestation by that weed, pastures are abandoned after a few years, and as a fatal consequence, the biodi- versity-rich tropical forest is progressively cleared for new grazing areas. Here we present a broad physiological comparison of the two plant species that are the main competitors on the pastures in the tropical Ecuadorian Andes, the planted forage grass Setaria sphacelata and the weed Bracken (Pteridium arachnoideum).With increasing elevation, the competitive power of Bracken increases as shown by satellite data of the study region. Using data obtained from field measurements, the annual biomass production of both plant species, as a measure of their competitive strength, was modeled over an elevational gradient from 1800 to 2800 m. The model shows that with increasing elevation, biomass production of the two species shifts in favor of Bracken which, above 1800 m, is capable of outgrowing the grass. In greenhouse experiments, the effects on plant growth of the presumed key vari- ables of the elevational gradient, temperature and UV radiation, were separately analyzed. Low temperature, as well as UV irradiation, inhibited carbon uptake of the C4-grass more than that of the C3-plant Bracken. The less temperature-sensitive photosynthesis of Bracken and its effective protection from UV radiation contribute to the success of the weed on the highland pastures. In field samples of Bracken but not of Setaria, the content of flavo- noids as UV-scavengers increased with the elevation. Combining modeling with measure- ments in greenhouse and field allowed to explain the invasive growth of a common weed in upland pastures. The performance of Setaria decreases with elevation due to suboptimal photosynthesis at lower temperatures and the inability to adapt its cellular UV screen