Modern advances in cloud computing and machine-leaning algorithms are shifting the manner in which Earth-observation (EO) data are used for environmental monitoring, particularly as we settle into the era of free, open-access satellite data streams. Wetland delineation represents a particularly worthy application of this emerging research trend, since wetlands are an ecologically important yet chronically under-represented component of contemporary mapping and monitoring programs, particularly at the regional and national levels. Exploiting Google Earth Engine and R Statistical software, we developed a workflow for predicting the probability of wetland occurrence using a boosted regression tree machine-learning framework applied to digital topographic and EO data. Working in a 13,700 km 2 study area in northern Alberta, our best models produced excellent results, with AUC (area under the receiver-operator characteristic curve) values of 0.898 and explained-deviance values of 0.708. Our results demonstrate the central role of high-quality topographic variables for modeling wetland distribution at regional scales. Including optical and/or radar variables into the workflow substantially improved model performance, though optical data performed slightly better. Converting our wetland probability-of-occurrence model into a binary Wet-Dry classification yielded an overall accuracy of 85%, which is virtually identical to that derived from the Alberta Merged Wetland Inventory (AMWI): the contemporary inventory used by the Government of Alberta. However, our workflow contains several key advantages over that used to produce the AMWI, and provides a scalable foundation for province-wide monitoring initiatives.

Advances in machine learning have changed many fields of study and it has also drawn attention in a variety of remote sensing applications. In particular, deep convolutional neural networks (CNNs) have proven very useful in fields such as image recognition; however, the use of CNNs in large-scale remote sensing landcover classifications still needs further investigation. We set out to test CNN-based landcover classification against a more conventional XGBoost shallow learning algorithm for mapping a notoriously difficult group of landcover classes, wetland class as defined by the Canadian Wetland Classification System. We developed two wetland inventory style products for a large (397,958 km2) area in the Boreal Forest region of Alberta, Canada, using Sentinel-1, Sentinel-2, and ALOS DEM data acquired in Google Earth Engine. We then tested the accuracy of these two products against three validation data sets (two photo-interpreted and one field). The CNN-generated wetland product proved to be more accurate than the shallow learning XGBoost wetland product by 5%. The overall accuracy of the CNN product was 80.2% with a mean F1-score of 0.58. We believe that CNNs are better able to capture natural complexities within wetland classes, and thus may be very useful for complex landcover classifications. Overall, this CNN framework shows great promise for generating large-scale wetland inventory data and may prove useful for other landcover mapping applications.

Abstract:Through the application and use of geospatial data, this study aimed to detect and characterize some of the key environmental drivers contributing to landscape-scale vegetation response patterns in Central Asia. The objectives of the study were to identify the variables driving the year-to-year vegetation dynamics in three regional landscapes (desert, steppe, and mountainous); and to determine if the identified environmental drivers can be used to explain the spatial-temporal variability of these spatio-temporal dynamics over time. It was posed that patterns of change in terrestrial phenology, derived from the 8 km bi-weekly time series of Normalized Difference Vegetation Index (NDVI) data acquired by the Advanced Very High Resolution Radiometer (AVHRR) satellites , can be explained through a multi-scale analysis of a suite of environmental drivers. Multiple linear stepwise regression analyses were used to test the hypotheses and address the objectives of the study. The annually computed phenological response variables or pheno-metrics time (season start, season length, and an NDVI-based productivity metric) were modeled as a function of ten environmental factors relating to soil, topography, and climate. Each of the three studied regional landscapes was shown to be governed by a distinctive suite of environmental drivers. The phenological responses of the steppe landscapes were affected by the year-to-year variation in temperature regimes. The phenology of the mountainous landscapes was influenced primarily by the elevation gradient. The phenological responses of desert landscapes were demonstrated to have the greatest variability over time and seemed to be affected by soil carbon content and OPEN ACCESSRemote Sens. 2011, 3 204 year-to-year variation of both temperature regimes and winter precipitation patterns. Amounts and scales of observed phenological variability over time (measured through coefficient of variation for each pheno-metric time ) in each of the regional landscapes were interpreted in terms of their resistance and resilience capacities under existing and projected environmental settings.

Sustainable land-use planning should consider large-scale landscape connectivity. commonly-used species-specific connectivity models are difficult to generalize for a wide range of taxa. In the context of multi-functional land-use planning, there is growing interest in species-agnostic approaches, modelling connectivity as a function of human landscape modification. We propose a conceptual framework, apply it to model connectivity as current density across Alberta, canada, and assess map sensitivity to modelling decisions. We directly compared the uncertainty related to (1) the definition of the degree of human modification, (2) the decision whether water bodies are considered barriers to movement, and (3) the scaling function used to translate degree of human modification into resistance values. connectivity maps were most sensitive to the consideration of water as barrier to movement, followed by the choice of scaling function, whereas maps were more robust to different conceptualizations of the degree of human modification. We observed higher concordance among cells with high (standardized) current density values than among cells with low values, which supports the identification of cells contributing to larger-scale connectivity based on a cut-off value. We conclude that every parameter in species-agnostic connectivity modelling requires attention, not only the definition of often-criticized expert-based degrees of human modification.Land-use and land-cover changes have impacted many natural ecosystems to provide ecosystem goods and services for an ever-growing human population 1 . This often has unintended consequences that may threaten biodiversity and ecosystem health 2 . Such consequences include changes in local, regional, and global climate 3 , alteration of natural habitats 4 , pollution of land, air, and water 5-7 , and changes in landscape structure, i.e., the total area and spatial configuration of ecosystems 8 .Natural ecosystems offer habitat for many species, and human landscape modification typically involves habitat loss as well as the breaking up of continuous habitats into smaller remnant patches (fragmentation) 9,10 . Consequently, landscapes may lose connectivity, i.e., the degree to which they facilitate movement of organisms and their genes among patches 11,12 . Landscape connectivity can be quantified in three ways: structural landscape connectivity, potential functional connectivity, and actual functional connectivity 13 . Structural landscape connectivity can be determined from physical attributes, based on maps alone without reference to organismal movement behaviour. Potential functional connectivity relies on a set of assumptions on organismal movement behaviour to implement an organism perspective, e.g. by mapping a species' habitat and setting a dispersal threshold. In contrast, actual functional connectivity refers to observed data (e.g., patch occupancy, radio tracking, open Scientific RepoRtS | (2020) 10:6798 | https://doi.org/10.1038/s41598-020-63545-z www.nature.com/scientifi...