scholarly journals Intercomparison of Machine-Learning Methods for Estimating Surface Shortwave and Photosynthetically Active Radiation

2020 ◽  
Vol 12 (3) ◽  
pp. 372
Author(s):  
Meredith G. L. Brown ◽  
Sergii Skakun ◽  
Tao He ◽  
Shunlin Liang
Keyword(s):  
Machine Learning ◽  
Photosynthetically Active Radiation ◽  
Input Data ◽  
Surface Radiation ◽  
Shortwave Radiation ◽  
Learning Methods ◽  
Active Radiation ◽  
Machine Learning Methods ◽  
Robust Model ◽  
Data Requirements

Satellite-derived estimates of downward surface shortwave radiation (SSR) and photosynthetically active radiation (PAR) are a part of the surface radiation budget, an essential climate variable (ECV) required by climate and vegetation models. Ground measurements are insufficient for generating long-term, global measurements of surface radiation, primarily due to spatial limitations; however, remotely sensed Earth observations offer freely available, multi-day, global coverage of radiance that can be used to derive SSR and PAR estimates. Satellite-derived SSR and PAR estimates are generated by computing the radiative transfer inversion of top-of-atmosphere (TOA) measurements, and require ancillary data on the atmospheric condition. To reduce computational costs, often the radiative transfer calculations are done offline and large look-up tables (LUTs) are generated to derive estimates more quickly. Recently studies have begun exploring the use of machine-learning techniques, such as neural networks, to try to improve computational efficiency. Here, nine machine-learning methods were tested to model SSR and PAR using minimal input data from the Moderate Resolution Imaging Spectrometer (MODIS) observations at 1 km spatial resolution. The aim was to reduce the input data requirements to create the most robust model possible. The bootstrap aggregated decision tree (Bagged Tree), Gaussian Process Regression, and Neural Network yielded the best results with minimal training data requirements: an R 2 of 0.77, 0.78, and 0.78 respectively, a bias of 0 ± 6, 0 ± 6, and 0 ± 5 W / m 2 , and an RMSE of 140 ± 7, 135 ± 8, and 138 ± 7 W / m 2 , respectively, for all-sky condition total surface shortwave radiation and viewing angles less than 55°. Viewing angles above 55° were excluded because the residual analysis showed exponential error growth above 55°. A simple, robust model for estimating SSR and PAR using machine-learning methods is useful for a variety of climate system studies. Future studies may focus on developing high temporal resolution direct and diffuse estimates of SSR and PAR as most current models estimate only total SSR or PAR.

scholarly journals A New Set of MODIS Land Products (MCD18): Downward Shortwave Radiation and Photosynthetically Active Radiation

2020 ◽  
Vol 12 (1) ◽  
pp. 168 ◽  
Author(s):  
Dongdong Wang ◽  
Shunlin Liang ◽  
Yi Zhang ◽  
Xueyuan Gao ◽  
Meredith G. L. Brown ◽  
...  
Keyword(s):  
Spatial Resolution ◽  
Photosynthetically Active Radiation ◽  
Process Models ◽  
Surface Radiation ◽  
Shortwave Radiation ◽  
Glass Product ◽  
Ancillary Data ◽  
Diurnal Variability ◽  
Active Radiation ◽  
Downward Shortwave Radiation

Surface downward shortwave radiation (DSR) and photosynthetically active radiation (PAR), its visible component, are key parameters needed for many land process models and terrestrial applications. Most existing DSR and PAR products were developed for climate studies and therefore have coarse spatial resolutions, which cannot satisfy the requirements of many applications. This paper introduces a new global high-resolution product of DSR (MCD18A1) and PAR (MCD18A2) over land surfaces using the MODIS data. The current version is Collection 6.0 at the spatial resolution of 5 km and two temporal resolutions (instantaneous and three-hour). A look-up table (LUT) based retrieval approach was chosen as the main operational algorithm so as to generate the products from the MODIS top-of-atmosphere (TOA) reflectance and other ancillary data sets. The new MCD18 products are archived and distributed via NASA’s Land Processes Distributed Active Archive Center (LP DAAC). The products have been validated based on one year of ground radiation measurements at 33 Baseline Surface Radiation Network (BSRN) and 25 AmeriFlux stations. The instantaneous DSR has a bias of −15.4 W/m2 and root mean square error (RMSE) of 101.0 W/m2, while the instantaneous PAR has a bias of −0.6 W/m2 and RMSE of 45.7 W/m2. RMSE of daily DSR is 32.3 W/m2, and that of the daily PAR is 13.1 W/m2. The accuracy of the new MODIS daily DSR data is higher than the GLASS product and lower than the CERES product, while the latter incorporates additional geostationary data with better capturing DSR diurnal variability. MCD18 products are currently under reprocessing and the new version (Collection 6.1) will provide improved spatial resolution (1 km) and accuracy.

scholarly journals Estimating Surface Downward Longwave Radiation Using Machine Learning Methods

Atmosphere ◽  
2020 ◽  
Vol 11 (11) ◽  
pp. 1147 ◽  
Author(s):  
Chunjie Feng ◽  
Xiaotong Zhang ◽  
Yu Wei ◽  
Weiyu Zhang ◽  
Ning Hou ◽  
...  
Keyword(s):  
Machine Learning ◽  
Land Cover ◽  
Time Scale ◽  
Longwave Radiation ◽  
Surface Radiation ◽  
Gradient Boosting ◽  
Support Vector ◽  
Learning Methods ◽  
Machine Learning Methods ◽  
Downward Longwave Radiation

The downward longwave radiation (Ld, 4–100 μm) is a major component of research for the surface radiation energy budget and balance. In this study, we applied five machine learning methods, namely artificial neural network (ANN), support vector regression (SVR), gradient boosting regression tree (GBRT), random forest (RF), and multivariate adaptive regression spline (MARS), to estimate Ld using ground measurements collected from 27 Baseline Surface Radiation Network (BSRN) stations. Ld measurements in situ were used to validate the accuracy of Ld estimation models on daily and monthly time scales. A comparison of the results demonstrated that the estimates on the basis of the GBRT method had the highest accuracy, with an overall root-mean-square error (RMSE) of 17.50 W m−2 and an R value of 0.96 for the test dataset on a daily time scale. These values were 11.19 W m−2 and 0.98, respectively, on a monthly time scale. The effects of land cover and elevation were further studied to comprehensively evaluate the performance of each machine learning method. All machine learning methods achieved better results over the grass land cover type but relatively worse results over the tundra. GBRT, RF, and MARS methods were found to show good performance at both the high- and low-altitude sites.

scholarly journals Development Android Application for Visualization of Technical Characteristics Based on Augmented Reality and Machine Learning Methods

2019 ◽  
Vol 17 (3) ◽  
pp. 73-92
Author(s):  
I. S. Prokopyev ◽  
E. S. Shmakov
Keyword(s):  
Machine Learning ◽  
Input Data ◽  
Graphical Representation ◽  
Electronic Devices ◽  
Recognition System ◽  
Android Application ◽  
Greedy Search ◽  
Learning Methods ◽  
Machine Learning Methods ◽  
Device Type

Nowadays the number of electronic devices has increased as well as their complexity. At the same time only few people are reading user guides for these devices. Moreover, in certain cases even no guides are provided by the manufacturer. Therefore there is a large number of cases when people use such devices incorrectly. This issue could be possibly solved with a system automatically recognizing the type of the device. Such system could provide all necessary user information about that device. This article suggests one of possible implementations of such device type recognition system. It recognizes type of device in an unsupevised way and shows main characteristics and user guides for the recognized gadget. Our approach for constructing such system relies on machine learning methods since greedy search for an object pattern is not efficient, as it was found out by recent scholarly works. Moreover, automatic object patterns classifiers show higher performance in this task and allow to scale the system to various kinds of input data. The algorithms that we are using for object classification are based on feature extraction from an graphical representation of the object look. This representation is usually proposed in an digital photography format. We consider our study as the first work towards automated defining characteristics of a device based on its graphical representation.

Sign in / Sign up

Export Citation Format

Share Document