Machine learning analyses of remote sensing measurements establish strong relationships between vegetation and snow depth in the boreal forest of Interior Alaska

2021 ◽  
Author(s):  
Thomas Douglas ◽  
Caiyun Zhang
Keyword(s):  
Machine Learning ◽  
Remote Sensing ◽  
Snow Depth ◽  
Spatial Scales ◽  
Critical Role ◽  
Winter Season ◽  
Thermal Conditions ◽  
Near Surface ◽  
Lidar Measurements ◽  
Remote Sensing Methods

The seasonal snowpack plays a critical role in Arctic and boreal hydrologic and ecologic processes. Though snow depth can be different from one season to another there are repeated relationships between ecotype and snowpack depth. Alterations to the seasonal snowpack, which plays a critical role in regulating wintertime soil thermal conditions, have major ramifications for near-surface permafrost. Therefore, relationships between vegetation and snowpack depth are critical for identifying how present and projected future changes in winter season processes or land cover will affect permafrost. Vegetation and snow cover areal extent can be assessed rapidly over large spatial scales with remote sensing methods, however, measuring snow depth remotely has proven difficult. This makes snow depth–vegetation relationships a potential means of assessing snowpack characteristics. In this study, we combined airborne hyperspectral and LiDAR data with machine learning methods to characterize relationships between ecotype and the end of winter snowpack depth. Our results show hyperspectral measurements account for two thirds or more of the variance in the relationship between ecotype and snow depth. An ensemble analysis of model outputs using hyperspectral and LiDAR measurements yields the strongest relationships between ecotype and snow depth. Our results can be applied across the boreal biome to model the coupling effects between vegetation and snowpack depth.

Application of Remote Sensing in Agriculture

2014 ◽  
Vol 13 (1) ◽  
Author(s):  
Jan Piekarczyk
Keyword(s):  
Remote Sensing ◽  
Precision Agriculture ◽  
Crop Production ◽  
Satellite Images ◽  
Crop Yields ◽  
Spatial Scales ◽  
Strong Relationship ◽  
Physical Parameters ◽  
Agricultural Crop ◽  
Remote Sensing Methods

AbstractWith increasing intensity of agricultural crop production increases the need to obtain information about environmental conditions in which this production takes place. Remote sensing methods, including satellite images, airborne photographs and ground-based spectral measurements can greatly simplify the monitoring of crop development and decision-making to optimize inputs on agricultural production and reduce its harmful effects on the environment. One of the earliest uses of remote sensing in agriculture is crop identification and their acreage estimation. Satellite data acquired for this purpose are necessary to ensure food security and the proper functioning of agricultural markets at national and global scales. Due to strong relationship between plant bio-physical parameters and the amount of electromagnetic radiation reflected (in certain ranges of the spectrum) from plants and then registered by sensors it is possible to predict crop yields. Other applications of remote sensing are intensively developed in the framework of so-called precision agriculture, in small spatial scales including individual fields. Data from ground-based measurements as well as from airborne or satellite images are used to develop yield and soil maps which can be used to determine the doses of irrigation and fertilization and to take decisions on the use of pesticides.

scholarly journals Distributed snow and rock temperature modelling in steep rock walls using Alpine3D

2017 ◽  
Vol 11 (1) ◽  
pp. 585-607 ◽  
Author(s):  
Anna Haberkorn ◽  
Nander Wever ◽  
Martin Hoelzle ◽  
Marcia Phillips ◽  
Robert Kenner ◽  
...  
Keyword(s):  
Energy Balance ◽  
Snow Cover ◽  
Snow Depth ◽  
Spatial Scales ◽  
Swiss Alps ◽  
Balance Model ◽  
Near Surface ◽  
Surface Rock ◽  
Depth Data ◽  
Rock Temperature

Abstract. In this study we modelled the influence of the spatially and temporally heterogeneous snow cover on the surface energy balance and thus on rock temperatures in two rugged, steep rock walls on the Gemsstock ridge in the central Swiss Alps. The heterogeneous snow depth distribution in the rock walls was introduced to the distributed, process-based energy balance model Alpine3D with a precipitation scaling method based on snow depth data measured by terrestrial laser scanning. The influence of the snow cover on rock temperatures was investigated by comparing a snow-covered model scenario (precipitation input provided by precipitation scaling) with a snow-free (zero precipitation input) one. Model uncertainties are discussed and evaluated at both the point and spatial scales against 22 near-surface rock temperature measurements and high-resolution snow depth data from winter terrestrial laser scans.In the rough rock walls, the heterogeneously distributed snow cover was moderately well reproduced by Alpine3D with mean absolute errors ranging between 0.31 and 0.81 m. However, snow cover duration was reproduced well and, consequently, near-surface rock temperatures were modelled convincingly. Uncertainties in rock temperature modelling were found to be around 1.6 °C. Errors in snow cover modelling and hence in rock temperature simulations are explained by inadequate snow settlement due to linear precipitation scaling, missing lateral heat fluxes in the rock, and by errors caused by interpolation of shortwave radiation, wind and air temperature into the rock walls.Mean annual near-surface rock temperature increases were both measured and modelled in the steep rock walls as a consequence of a thick, long-lasting snow cover. Rock temperatures were 1.3–2.5 °C higher in the shaded and sunny rock walls, while comparing snow-covered to snow-free simulations. This helps to assess the potential error made in ground temperature modelling when neglecting snow in steep bedrock.

scholarly journals Evaluating remote sensing of deciduous forest phenology at multiple spatial scales using PhenoCam imagery

2014 ◽  
Vol 11 (16) ◽  
pp. 4305-4320 ◽  
Author(s):  
S. T. Klosterman ◽  
K. Hufkens ◽  
J. M. Gray ◽  
E. Melaas ◽  
O. Sonnentag ◽  
...  
Keyword(s):  
Remote Sensing ◽  
Time Series ◽  
Curve Fitting ◽  
Satellite Remote Sensing ◽  
Vegetation Index ◽  
Deciduous Forest ◽  
Spatial Scales ◽  
Visual Assessment ◽  
Statistical Uncertainty ◽  
Near Surface

Abstract. Plant phenology regulates ecosystem services at local and global scales and is a sensitive indicator of global change. Estimates of phenophase transition dates, such as the start of spring or end of fall, can be derived from sensor-based time series, but must be interpreted in terms of biologically relevant events. We use the PhenoCam archive of digital repeat photography to implement a consistent protocol for visual assessment of canopy phenology at 13 temperate deciduous forest sites throughout eastern North America, and to perform digital image analysis for time-series-based estimation of phenophase transition dates. We then compare these results to remote sensing metrics of phenophase transition dates derived from the Moderate Resolution Imaging Spectroradiometer (MODIS) and Advanced Very High Resolution Radiometer (AVHRR) sensors. We present a new type of curve fit that uses a generalized sigmoid function to estimate phenology dates, and we quantify the statistical uncertainty of phenophase transition dates estimated using this method. Results show that the generalized sigmoid provides estimates of dates with less statistical uncertainty than other curve-fitting methods. Additionally, we find that dates derived from analysis of high-frequency PhenoCam imagery have smaller uncertainties than satellite remote sensing metrics of phenology, and that dates derived from the remotely sensed enhanced vegetation index (EVI) have smaller uncertainty than those derived from the normalized difference vegetation index (NDVI). Near-surface time-series estimates for the start of spring are found to closely match estimates derived from visual assessment of leaf-out, as well as satellite remote-sensing-derived estimates of the start of spring. However late spring and fall phenology metrics exhibit larger differences between near-surface and remote scales. Differences in late spring phenology between near-surface and remote scales are found to correlate with a landscape metric of deciduous forest cover. These results quantify the effect of landscape heterogeneity when aggregating to the coarser spatial scales of remote sensing, and demonstrate the importance of accurate curve fitting and vegetation index selection when analyzing and interpreting phenology time series.

scholarly journals Variability scaling and consistency in airborne and satellite altimetry measurements of Arctic sea ice

2020 ◽  
Vol 14 (2) ◽  
pp. 751-767
Author(s):  
Shiming Xu ◽  
Lu Zhou ◽  
Bin Wang
Keyword(s):  
Remote Sensing ◽  
Sea Ice ◽  
Snow Depth ◽  
Spatial Scales ◽  
Ice Cover ◽  
Arctic Sea Ice ◽  
The Arctic ◽  
First Year ◽  
Noise Levels ◽  
Wide Range

Abstract. Satellite and airborne remote sensing provide complementary capabilities for the observation of the sea ice cover. However, due to the differences in footprint sizes and noise levels of the measurement techniques, as well as sea ice's variability across scales, it is challenging to carry out inter-comparison or consistently study these observations. In this study we focus on the remote sensing of sea ice thickness parameters and carry out the following: (1) the analysis of variability and its statistical scaling for typical parameters and (2) the consistency study between airborne and satellite measurements. By using collocating data between Operation IceBridge and CryoSat-2 (CS-2) in the Arctic, we show that consistency exists between the variability in radar freeboard estimations, although CryoSat-2 has higher noise levels. Specifically, we notice that the noise levels vary among different CryoSat-2 products, and for the European Space Agency (ESA) CryoSat-2 freeboard product the noise levels are at about 14 and 20 cm for first-year ice (FYI) and multi-year ice (MYI), respectively. On the other hand, for Operation IceBridge and NASA's Ice, Cloud, and land Elevation Satellite (ICESat), it is shown that the variability in snow (or total) freeboard is quantitatively comparable despite more than a 5-year time difference between the two datasets. Furthermore, by using Operation IceBridge data, we also find widespread negative covariance between ice freeboard and snow depth, which only manifests on small spatial scales (40 m for first-year ice and about 80 to 120 m for multi-year ice). This statistical relationship highlights that the snow cover reduces the overall topography of the ice cover. Besides this, there is prevalent positive covariability between snow depth and snow freeboard across a wide range of spatial scales. The variability and consistency analysis calls for more process-oriented observations and modeling activities to elucidate key processes governing snow–ice interaction and sea ice variability on various spatial scales. The statistical results can also be utilized in improving both radar and laser altimetry as well as the validation of sea ice and snow prognostic models.

scholarly journals FIRE DETECTION BASED ON A SERIES OF THERMAL IMAGES AND POINT MEASUREMENTS: THE CASE STUDY OF COAL-WASTE DUMPS

2019 ◽  
Vol XLII-1/W2 ◽  
pp. 9-12 ◽  
Author(s):  
A. Abramowicz ◽  
R. Chybiorz
Keyword(s):  
Remote Sensing ◽  
Thermal Conditions ◽  
Fire Detection ◽  
Hard Coal ◽  
Coal Waste ◽  
Thermal Images ◽  
Mining Areas ◽  
Waste Dumps ◽  
Upper Silesian Coal Basin ◽  
Remote Sensing Methods

<p><strong>Abstract.</strong> The development of mining areas is always associated with a huge amount of unused gangue. It is treated as waste that should be disposed of. Most often it is done by creating waste dumps. Unfortunately, in the case of hard coal, such treatments can be dangerous, because the remaining coal can self-heat, what then leads to ignition and fire. These fires are uncontrolled and significantly affect the environment and the health of residents. Areas at risk of fire should be properly secured and its thermal conditions monitored. There are currently many methods available for measuring temperature. The latest and also the most popular are pyrometric and remote sensing methods. For the purposes of analysis, fieldwork on the selected coal-waste dump in the Upper Silesian Coal Basin was made using pyrometric (point measurements) and remote sensing methods (thermal images). A series of photos and measurements were compared over time and space to catch change trends. Differences in the received data were identified. Studies have shown that using only one method helps identify a fire, but does not provide enough information about its structure. Comprehensive monitoring is the best solution.</p>

scholarly journals Machine Learning-Based CYGNSS Soil Moisture Estimates over ISMN sites in CONUS

2020 ◽  
Vol 12 (7) ◽  
pp. 1168 ◽  
Author(s):  
Volkan Senyurek ◽  
Fangni Lei ◽  
Dylan Boyd ◽  
Mehmet Kurum ◽  
Ali Cafer Gurbuz ◽  
...  
Keyword(s):  
Machine Learning ◽  
Remote Sensing ◽  
Soil Moisture ◽  
Microwave Remote Sensing ◽  
Satellite System ◽  
Vital Role ◽  
Support Vector ◽  
Near Surface ◽  
Data Products ◽  
Global Navigation Satellite

Soil moisture (SM) derived from satellite-based remote sensing measurements plays a vital role for understanding Earth’s land and near-surface atmosphere interactions. Bistatic Global Navigation Satellite System (GNSS) Reflectometry (GNSS-R) has emerged in recent years as a new domain of microwave remote sensing with great potential for SM retrievals, particularly at high spatio-temporal resolutions. In this work, a machine learning (ML)-based framework is presented for obtaining SM data products over the International Soil Moisture Network (ISMN) sites in the Continental United States (CONUS) by leveraging spaceborne GNSS-R observations provided by NASA’s Cyclone GNSS (CYGNSS) constellation alongside remotely sensed geophysical data products. Three widely-used ML approaches—artificial neural network (ANN), random forest (RF), and support vector machine (SVM)—are compared and analyzed for the SM retrieval through utilizing multiple validation strategies. Specifically, using a 5-fold cross-validation method, overall RMSE values of 0.052, 0.061, and 0.065 cm3/cm3 are achieved for the RF, ANN, and SVM techniques, respectively. In addition, both a site-independent and a year-based validation techniques demonstrate satisfactory accuracy of the proposed ML model, suggesting that this SM approach can be generalized in space and time domains. Moreover, the achieved accuracy can be further improved when the model is trained and tested over individual SM networks as opposed to combining all available SM networks. Additionally, factors including soil type and land cover are analyzed with respect to their impacts on the accuracy of SM retrievals. Overall, the results demonstrated here indicate that the proposed technique can confidently provide SM estimates over lightly-vegetated areas with vegetation water content (VWC) less than 5 kg/m2 and relatively low spatial heterogeneity.

scholarly journals Spatiotemporal dynamics of snow cover based on multi-source remote sensing data in China

2016 ◽  
Vol 10 (5) ◽  
pp. 2453-2463 ◽  
Author(s):  
Xiaodong Huang ◽  
Jie Deng ◽  
Xiaofang Ma ◽  
Yunlong Wang ◽  
Qisheng Feng ◽  
...  
Keyword(s):  
Remote Sensing ◽  
Snow Cover ◽  
Snow Depth ◽  
Winter Season ◽  
Remote Sensing Data ◽  
Microwave Remote Sensing ◽  
The Tibetan Plateau ◽  
Optical Remote Sensing ◽  
The North ◽  
The Mean

Abstract. By combining optical remote sensing snow cover products with passive microwave remote sensing snow depth (SD) data, we produced a MODIS (Moderate Resolution Imaging Spectroradiometer) cloudless binary snow cover product and a 500 m snow depth product. The temporal and spatial variations of snow cover from December 2000 to November 2014 in China were analyzed. The results indicate that, over the past 14 years, (1) the mean snow-covered area (SCA) in China was 11.3 % annually and 27 % in the winter season, with the mean SCA decreasing in summer and winter seasons, increasing in spring and fall seasons, and not much change annually; (2) the snow-covered days (SCDs) showed an increase in winter, spring, and fall, and annually, whereas they showed a decrease in summer; (3) the average SD decreased in winter, summer, and fall, while it increased in spring and annually; (4) the spatial distributions of SD and SCD were highly correlated seasonally and annually; and (5) the regional differences in the variation of snow cover in China were significant. Overall, the SCD and SD increased significantly in south and northeast China, and decreased significantly in the north of Xinjiang province. The SCD and SD increased on the southwest edge and in the southeast part of the Tibetan Plateau, whereas it decreased in the north and northwest regions.

scholarly journals Evaluating remote sensing of deciduous forest phenology at multiple spatial scales using PhenoCam imagery

2014 ◽  
Vol 11 (2) ◽  
pp. 2305-2342 ◽  
Author(s):  
S. T. Klosterman ◽  
K. Hufkens ◽  
J. M. Gray ◽  
E. Melaas ◽  
O. Sonnentag ◽  
...  
Keyword(s):  
Remote Sensing ◽  
Time Series ◽  
Curve Fitting ◽  
Vegetation Index ◽  
Deciduous Forest ◽  
Spatial Scales ◽  
Visual Assessment ◽  
Late Spring ◽  
Statistical Uncertainty ◽  
Near Surface

Abstract. Plant phenology regulates ecosystem services at local and global scales and is a sensitive indicator of global change. Estimates of phenophase transition dates, such as the start of spring or end of autumn, can be derived from sensor-based time series data at the near-surface and remote scales, but must be interpreted in terms of biologically relevant events. We use the PhenoCam archive of digital repeat photography to implement a consistent protocol for visual assessment of canopy phenology at 13 temperate deciduous forest sites throughout eastern North America, as well as to perform digital image analysis for time series-based estimates of phenology dates. We then compare these near-surface results to remote sensing metrics of phenology at the landscape scale, derived from the Moderate Resolution Imaging Spectroradiometer (MODIS) and Advanced Very High Resolution Radiometer (AVHRR) sensors. We present a new type of curve fit, using a generalized sigmoid, to estimate phenology dates. We quantify the statistical uncertainty of phenophase transition dates estimated using this method and show that the generalized sigmoid results in less statistical uncertainty than other curve-fitting methods. Additionally, we find that dates derived from analysis of high-frequency PhenoCam imagery have smaller uncertainties than remote sensing metrics of phenology, and that dates derived from the remotely-sensed enhanced vegetation index (EVI) have smaller uncertainty than those derived from the normalized difference vegetation index (NDVI). Near-surface time series estimates for the start of spring are found to closely match visual assessment of leaf out, as well as remote sensing-derived estimates of the start of spring. However late spring and autumn phenology exhibit larger differences between near-surface and remote scales. Differences in late spring phenology between near-surface and remote scales are found to correlate with a landscape metric of deciduous forest cover. These results quantify the effect of landscape heterogeneity when aggregating to the coarser spatial scales of remote sensing, and demonstrate the importance of accurate curve fitting and vegetation index selection when analyzing and interpreting phenology time series.

scholarly journals Snow depth estimation and historical data reconstruction over China based on a random forest machine learning approach

2020 ◽  
Vol 14 (6) ◽  
pp. 1763-1778 ◽  
Author(s):  
Jianwei Yang ◽  
Lingmei Jiang ◽  
Kari Luojus ◽  
Jinmei Pan ◽  
Juha Lemmetyinen ◽  
...  
Keyword(s):  
Machine Learning ◽  
Random Forest ◽  
Snow Depth ◽  
Spatial Scales ◽  
Ground Truth ◽  
Depth Estimation ◽  
Predictor Variables ◽  
Science Data ◽  
Increasing Trend ◽  
Mean Square Errors

Abstract. We investigated the potential capability of the random forest (RF) machine learning (ML) model to estimate snow depth in this work. Four combinations composed of critical predictor variables were used to train the RF model. Then, we utilized three validation datasets from out-of-bag (OOB) samples, a temporal subset, and a spatiotemporal subset to verify the fitted RF algorithms. The results indicated the following: (1) the accuracy of the RF model is greatly influenced by geographic location, elevation, and land cover fractions; (2) however, the redundant predictor variables (if highly correlated) slightly affect the RF model; and (3) the fitted RF algorithms perform better on temporal than spatial scales, with unbiased root-mean-square errors (RMSEs) of ∼4.4 and ∼7.3 cm, respectively. Finally, we used the fitted RF2 algorithm to retrieve a consistent 32-year daily snow depth dataset from 1987 to 2018. This product was evaluated against the independent station observations during the period 1987–2018. The mean unbiased RMSE and bias were 7.1 and −0.05 cm, respectively, indicating better performance than that of the former snow depth dataset (8.4 and −1.20 cm) from the Environmental and Ecological Science Data Center for West China (WESTDC). Although the RF product was superior to the WESTDC dataset, it still underestimated deep snow cover (>20 cm), with biases of −10.4, −8.9, and −34.1 cm for northeast China (NEC), northern Xinjiang (XJ), and the Qinghai–Tibetan Plateau (QTP), respectively. Additionally, the long-term snow depth datasets (station observations, RF estimates, and WESTDC product) were analyzed in terms of temporal and spatial variations over China. On a temporal scale, the ground truth snow depth presented a significant increasing trend from 1987 to 2018, especially in NEC. However, the RF and WESTDC products displayed no significant changing trends except on the QTP. The WESTDC product presented a significant decreasing trend on the QTP, with a correlation coefficient of −0.55, whereas there were no significant trends for ground truth observations and the RF product. For the spatial characteristics, similar trend patterns were observed for RF and WESTDC products over China. These characteristics presented significant decreasing trends in most areas and a significant increasing trend in central NEC.

scholarly journals Variability Scaling and Consistency of Airborne and Satellite Altimetry Measurements of Arctic Sea Ice

2019 ◽  
Author(s):  
Shiming Xu ◽  
Lu Zhou ◽  
Bin Wang
Keyword(s):  
Remote Sensing ◽  
Sea Ice ◽  
Snow Depth ◽  
Spatial Scales ◽  
Ice Cover ◽  
Arctic Sea Ice ◽  
The Arctic ◽  
First Year ◽  
Noise Levels ◽  
Wide Range

Abstract. Satellite and airborne remote sensing provide complementary capabilities for the observation of the sea ice cover. However, due to the differences in footprint sizes and noise levels of the measurement techniques, as well as sea ice's variability across scales, it is challenging to carry out inter-comparison or consistency study of these observations. In this study we focus on the remote sensing of sea ice thickness parameters, and carry out: (1) the analysis of variability and its statistical scaling for typical parameters, and (2) the consistency study between airborne and satellite measurements. By using collocating data between Operation IceBridge and CryoSat-2 in the Arctic, we show that there exists consistency between the variability of radar freeboard estimations, although CryoSat-2 has higher noise levels. Specifically, we notice that the noise levels vary among different CryoSat-2 products, and for ESA CryoSat-2 freeboard product the noise levels are at about 14 and 20 cm for first-year and multiyear ice, respectively. On the other hand, for Operation IceBridge and ICESat, it is shown that the variability of snow (or total) freeboard is quantitatively comparable, despite over 5 years' the time difference between the two datasets. Furthermore, by using Operation IceBridge data, we also find wide-spread negative covariance between ice freeboard and snow depth, which only manifest at small spatial scales (40 m for first-year ice and about 80 to 120 m for MYI). This statistical relationship highlights that the snow cover reduces the overall topography of the ice cover. Besides, there is prevalent positive covariability between snow depth and snow freeboard across a wide range of spatial scales. The variability and consistency analysis calls for more process-oriented observations and modeling activities to elucidate key processes governing snow-ice interaction and sea ice variability on various spatial scales. The statistical results can also be utilized in improving both radar and laser altimetry, as well as the validation of sea ice and snow prognostic models.

Sign in / Sign up

Export Citation Format

Share Document