Net Surface Shortwave Radiation Retrieval Using Random Forest Method With MODIS/AQUA Data

Self Cite

Net surface shortwave radiation (NSSR) is one of the most important fundamental parameters in various land processes. Benefiting from its efficient nonlinear fitting ability, machine learning algorithms have a great potential in the retrieval of NSSR. However, few studies have explored the level of accuracy that machine learning algorithms can reach for different land covers on the worldwide scale and what the optimal independent variables are in the machine learning-based NSSR model. To guide the use of machine learning algorithms correctly in the retrieval of NSSR, it is necessary to give a comprehensive analysis from algorithm complexity, accuracy, and other aspects. In this study, three classic machine learning algorithms, including Random Forest (RF), Artificial Neural Network (ANN), and Support Vector Regression (SVR), were built well to estimate instantaneous NSSR with optimal hyperparameters by elaborately selecting different independent variables, including top of atmosphere (TOA) channel spectral reflectance, geographic parameters, surface information, and atmosphere conditions. Global FLUXNET in situ measurements throughout 2014 were used to validate the accuracies of retrieved NSSR over various land cover types. The root mean square error (RMSE) is below 55 W/m2, and the distributions of error histogram are also similar. Approximately 50% of absolute error were within 25 W/m2. There was a performance difference of NSSR estimations in various surface types, and the performance of three machine learning methods in a specific surface type was also different. However, the RF method may be considered as the optimal methodology to retrieve NSSR from MODIS data, owing to its relatively better precision and concise hyperparameter-tuned process. The importance analysis of the proposed independent variables of NSSR retrieval shows that the introduction of geographic information can effectively reduce the error of NSSR retrieval, and surface information and atmosphere information are not necessary. It was also found that a combination of geographic information and blue band TOA reflectance already have a pretty good accuracy in NSSR retrieval, which implies there is a possibility to transfer our NSSR model to other satellite sensors, especially with insufficient channels. In a word, the NSSR model with machine learning algorithms would be an efficient, concise, and general method in the future.

Section: Accuracy Intercomparison Of Different Machine Learning Algorcontrasting

confidence: 56%

Benchmarking Machine Learning Algorithms for Instantaneous Net Surface Shortwave Radiation Retrieval Using Remote Sensing Data

Ying

2019

Self Cite

“…During the third step, a look-up-table (LUT) or machine learning algorithm was usually used to establish the relationship between VIs and LAI [44]. It has been reported that the machine learning model was more efficient and accurate than the LUT approach [45]. Thus, an RF model was used in the third step.…”

Section: Methodsmentioning

confidence: 99%

Optimal Integration of Optical and SAR Data for Improving Alfalfa Yield and Quality Traits Prediction: New Insights into Satellite-Based Forage Crop Monitoring

Chen,

Yu,

Cherney

et al. 2024

Global food security and nutrition is suffering from unprecedented challenges. To reach a world without hunger and malnutrition by implementing precision agriculture, satellite remote sensing plays an increasingly important role in field crop monitoring and management. Alfalfa, a global widely distributed forage crop, requires more attention to predict its yield and quality traits from satellite data since it supports the livestock industry. Meanwhile, there are some key issues that remain unknown regarding alfalfa remote sensing from optical and synthetic aperture radar (SAR) data. Using Sentinel-1 and Sentinel-2 satellite data, this study developed, compared, and further integrated new optical- and SAR-based satellite models for improving alfalfa yield and quality traits prediction, i.e., crude protein (CP), acid detergent fiber (ADF), neutral detergent fiber (NDF), and neutral detergent fiber digestibility (NDFD). Meanwhile, to better understand the physical mechanism of alfalfa optical remote sensing, a unified hybrid leaf area index (LAI) retrieval scheme was developed by coupling the PROSAIL radiative transfer model, spectral response function of the desired optical satellite, and a random forest (RF) model, denoted as a scalable optical satellite-based LAI retrieval framework. Compared to optical vegetation indices (VIs) that only capture canopy information, the results indicate that LAI had the highest correlation (r = 0.701) with alfalfa yield due to its capacity in delivering the vegetation structure characteristics. For alfalfa quality traits, optical chlorophyll VIs presented higher correlations than LAI. On the other hand, LAI did not provide a significant additional contribution for predicting alfalfa parameters in the RF developed optical prediction model using VIs as inputs. In addition, the optical-based model outperformed the SAR-based model for predicting alfalfa yield, CP, and NDFD, while the SAR-based model showed better performance for predicting ADF and NDF. The integration of optical and SAR data contributed to higher accuracy than either optical or SAR data separately. Compared to a traditional embedded integration approach, the combination of multisource heterogeneous optical and SAR satellites was optimized by multiple linear regression (yield: R2 = 0.846 and RMSE = 0.0354 kg/m2; CP: R2 = 0.636 and RMSE = 1.57%; ADF: R2 = 0.559 and RMSE = 1.926%; NDF: R2 = 0.58 and RMSE = 2.097%; NDFD: R2 = 0.679 and RMSE = 2.426%). Overall, this study provides new insights into forage crop yield prediction for large-scale fields using multisource heterogeneous satellites.

“…Note that MODIS band 6 was not used due to its poor quality [56]. Meanwhile, according to previous studies, surface shortwave radiation is closely related to the MODIS TOA reflectance from bands 1-7 [17,19,52], while the longwave radiation is mostly represented by the MODIS TOA radiance from bands 27-36 [53][54][55]; water vapor, which has a significant impact on solar radiation, is related to band 19 [19], and bands 21, 24, and 25 are affected by both solar illumination and the Earth's infrared emission [19]. Thus, MODIS bands 1-5, 7, 19, 21, 24, 25, and 27-36 were selected for modeling in this study.…”

Section: Methodsmentioning

confidence: 99%

“…In addition, the under or over-fitting issues which appear very often in machine learning can be solved by tuning some hyperparameters in RF (i.e., the number of trees in the forest, the maximum depth of each tree, minimum split sample number, and minimum sample leaves) [50]. RF has become a popular technique in remote sensing-related studies such as parameter estimation [17,50,52], parameter prediction [58], classification [59], variable importance determination [22], etc. The performance of RF is determined by the setting of the hyperparameters, such as "n-estimators", "max depth", "max features", "min samples split", "min sample leaves", etc.…”

Section: Modeling With Random Forestmentioning

confidence: 99%

“…As was done in the study of Wang et al [19], the viewing geometry (SZA, VZA, and RAA) and sky conditions (determined by the cloud mask) during each MODIS overpass time were taken as the key factors for case division and each of them divided into different bins, as shown in Table 5. Then, a group of conditional models determined by the combination of the above four factors was built by regressing the MODIS TOA data to the R n_daily as The cloud mask data obtained from MOD/MYD35 indicates the sky condition when the two MODIS sensors overpass, with "clear" defined as when the clear-sky confidence levels are "confident clear" or "probably clear" and "cloudy" defined as when the clear-sky confidence levels are "cloudy" or "uncertain clear" [23,52].…”

Section: Look-up Table (Lut-based) Modelmentioning

confidence: 99%

See 1 more Smart Citation

Estimation of the All-Wave All-Sky Land Surface Daily Net Radiation at Mid-Low Latitudes from MODIS Data Based on ERA5 Constraints

Jiang

Peng

et al. 2021

The surface all-wave net radiation (Rn) plays an important role in the energy and water cycles, and most studies of Rn estimations have been conducted using satellite data. As one of the most commonly used satellite data sets, Moderate Resolution Imaging Spectroradiometer (MODIS) data have not been widely used for radiation calculations at mid-low latitudes because of its very low revisit frequency. To improve the daily Rn estimation at mid-low latitudes with MODIS data, four models, including three models built with random forest (RF) and different temporal expansion models and one model built with the look-up-table (LUT) method, are used based on comprehensive in situ radiation measurements collected from 340 globally distributed sites, MODIS top-of-atmosphere (TOA) data, and the fifth generation of European Centre for Medium-Range Weather Forecasts Reanalysis 5 (ERA5) data from 2000 to 2017. After validation against the in situ measurements, it was found that the RF model based on the constraint of the daily Rn from ERA5 (an RF-based model with ERA5) performed the best among the four proposed models, with an overall validated root-mean-square error (RMSE) of 21.83 Wm−2, R2 of 0.89, and a bias of 0.2 Wm−2. It also had the best accuracy compared to four existing products (Global LAnd Surface Satellite Data (GLASS), Clouds and the Earth’s Radiant Energy System Edition 4A (CERES4A), ERA5, and FLUXCOM_RS) across various land cover types and different elevation zones. Further analyses illustrated the effectiveness of the model by introducing the daily Rn from ERA5 into a “black box” RF-based model for Rn estimation at the daily scale, which is used as a physical constraint when the available satellite observations are too limited to provide sufficient information (i.e., when the overpass time is less than twice per day) or the sky is overcast. Overall, the newly-proposed RF-based model with ERA5 in this study shows satisfactory performance and has strong potential to be used for long-term accurate daily Rn global mapping at finer spatial resolutions (e.g., 1 km) at mid-low latitudes.