Terrestrial photography is a cost-effective and easy-to-use method for measuring and monitoring spatially distributed land surface variables. It can be used to continuously investigate remote and often inaccessible terrain. We focus on the observation of snow cover patterns in high mountainous areas. The high temporal and spatial resolution of the photographs have various applications, for example validating spatially distributed snow hydrological models. However, the analysis of a photograph requires a preceding georectification of the digital camera image. To accelerate and simplify the analysis, we have developed the "Photo Rectification And ClassificaTIon SoftwarE" (PRACTISE) that is available as a Matlab code. The routine requires a digital camera image, the camera location and its orientation, as well as a digital elevation model (DEM) as input. If the viewing orientation and position of the camera are not precisely known, an optional optimisation routine using ground control points (GCPs) helps to identify the missing parameters. PRACTISE also calculates a viewshed using the DEM and the camera position. The visible DEM pixels are utilised to georeference the photograph which is subsequently classified. The resulting georeferenced and classified image can be directly compared to other georeferenced data and can be used within any geoinformation system. The Matlab routine was tested using observations of the north-eastern slope of the Schneefernerkopf, Zugspitze, Germany. The results obtained show that PRACTISE is a fast and user-friendly tool, able to derive the microscale variability of snow cover extent in high alpine terrain, but can also easily be adapted to other land surface applications

Abstract. Knowledge of current snow cover extent is essential for characterizing energy and moisture fluxes at the Earth's surface. The snow-covered area (SCA) is often estimated by using optical satellite information in combination with the normalized-difference snow index (NDSI). The NDSI thereby uses a threshold for the definition if a satellite pixel is assumed to be snow covered or snow free. The spatiotemporal representativeness of the standard threshold of 0.4 is however questionable at the local scale. Here, we use local snow cover maps derived from ground-based photography to continuously calibrate the NDSI threshold values (NDSI thr ) of Landsat satellite images at two European mountain sites of the period from 2010 to 2015. The Research Catchment Zugspitzplatt (RCZ, Germany) and Vernagtferner area (VF, Austria) are both located within a single Landsat scene. Nevertheless, the long-term analysis of the NDSI thr demonstrated that the NDSI thr at these sites are not correlated (r = 0.17) and different than the standard threshold of 0.4. For further comparison, a dynamic and locally optimized NDSI threshold was used as well as another locally optimized literature threshold value (0.7). It was shown that large uncertainties in the prediction of the SCA of up to 24.1 % exist in satellite snow cover maps in cases where the standard threshold of 0.4 is used, but a newly developed calibrated quadratic polynomial model which accounts for seasonal threshold dynamics can reduce this error. The model minimizes the SCA uncertainties at the calibration site VF by 50 % in the evaluation period and was also able to improve the results at RCZ in a significant way. Additionally, a scaling experiment shows that the positive effect of a locally adapted threshold diminishes using a pixel size of 500 m or larger, underlining the general applicability of the standard threshold at larger scales.

Abstract. Terrestrial photography combined with the recently presented Photo Rectification And ClassificaTIon SoftwarE (PRACTISE V.1.0) has proven to be a valuable source to derive snow cover maps in a high temporal and spatial resolution. The areal coverage of the used digital photographs is however strongly limited. Satellite images on the other hand can cover larger areas but do show uncertainties with respect to the accurate detection of the snow covered area. This is especially the fact if user defined thresholds are needed, e.g. in case of the frequently used normalizeddifference snow index (NDSI). The definition of this value is often not adequately defined by either a general value from literature or over the impression of the user, but not by reproducible independent information. PRACTISE V.2.1 addresses this important aspect and shows additional improvements. The Matlab-based software is now able to automatically process and detect snow cover in satellite images. A simultaneously captured camera-derived snow cover map is in this case utilized as in situ information for calibrating the NDSI threshold value. Moreover, an additional automatic snow cover classification, specifically developed to classify shadow-affected photographs, was included. The improved software was tested for photographs and Landsat 7 Enhanced Thematic Mapper (ETM+) as well as Landsat 8 Operational Land Imager (OLI) scenes in the Zugspitze massif (Germany). The results show that using terrestrial photography in combination with satellite imagery can lead to an objective, reproducible, and user-independent derivation of the NDSI threshold and the resulting snow cover map. The presented method is not limited to the sensor system or the threshold used in here but offers manifold application options for other scientific branches.

Abstract. Snow cover variability has a significant impact on climate and the environment and is of great socioeconomic importance for the European Alps. Terrestrial photography offers a high potential to monitor snow cover variability, but its application is often limited to small catchment scales. Here, we present a semiautomatic procedure to derive snow cover maps from publicly available webcam images in the Swiss Alps and propose a procedure for the georectification and snow classification of such images. In order to avoid the effort of manually setting ground control points (GCPs) for each webcam, we implement a novel registration approach that automatically resolves camera parameters (camera orientation; principal point; field of view, FOV) by using an estimate of the webcams' positions and a high-resolution digital elevation model (DEM). Furthermore, we propose an automatic image-to-image alignment to correct small changes in camera orientation and compare and analyze two recent snow classification methods. The resulting snow cover maps indicate whether a DEM grid is snow-covered, snow-free, or not visible from webcams' positions. GCPs are used to evaluate our novel automatic image registration approach. The evaluation reveals a root mean square error (RMSE) of 14.1 m for standard lens webcams (FOV<48∘) and a RMSE of 36.3 m for wide-angle lens webcams (FOV≥48∘). In addition, we discuss projection uncertainties caused by the mapping of low-resolution webcam images onto the high-resolution DEM. Overall, our results highlight the potential of our method to build up a webcam-based snow cover monitoring network.

No abstract