Evaluation of machine learning models for rice dry biomass estimation and mapping using quad-source optical imagery

ABSTRACT

Several machine learning regression models have been advanced for the estimation of crop biophysical parameters with optical satellite imagery. However, literature on the comparative performances of such models is still limited in range and scope, especially under multiple data sources, despite the potential of multi-source imagery to improving crop monitoring in cloudy areas. To fill in this knowledge gap, this study explored the synergistic use of Landsat-8, Sentinel-2A, China’s environment and disaster monitoring and forecasting satellites (HJ-1 A and B) and Gaofen-1 (GF-1) data to evaluate four machine learning regression models that include Random Forest (RF), Support Vector Machine (SVM), k-Nearest Neighbor (k-NN), and Gradient Boosting Decision Tree (GBDT), for rice dry biomass estimation and mapping. Taking a major rice cultivation area in southeast China as case study during the 2016 and 2017 growing seasons, a cross-calibrated time series of the Enhanced Vegetation Index (EVI) was obtained from the quad-source optical imagery and on which the aforementioned models were applied, respectively. Results indicate that in the before rice heading scenario, the most accurate dry biomass estimates were obtained by the GBDT model (R² of 0.82 and RMSE of 191.8 g/m²) followed by the RF model (R² of 0.79 and RMSE of 197.8 g/m²). After heading, the k-NN model performed best (R² of 0.43 and RMSE of 452.1 g/m²) followed by the RF model (R² of 0.42 and RMSE of 464.7 g/m²). Whist the k-NN model performed least in the before heading scenario, SVM performed least in the after heading scenario. These findings may suggest that machine learning regression models based on an ensemble of decision trees (RF and GBDT) are more suitable for the estimation of rice dry biomass, at least with optical satellite imagery. Studies that would extend the evaluation of these machine learning models, to other parameters like leaf area index, and to microwave imagery, are hereby recommended.     

Measuring Earth orientation with the global positioning system

Summary A globally distributed network of high-precision receivers which obtain data from the full Global Positioning System (GPS) configuration of 18 or more satellites may soon become an efficient and economical method for the rapid determination of short-term variations in Earth orientation. A covariance analysis has been performed to evaluate the errors associated with GPS monitoring of Earth orientation. Earth orientation parameters were modeled either as constants over observing windows of various lengths, or as stochastic process-noise variables. The sensitivity of Earth orientation estimates to systematic errors in selected model parameters was also examined. GPS measurements appear to be highly competitive with those from other techniques, and have the potential to generate nearly continuous centimeter-level Earth orientation information to aid both spacecraft navigation and the study of high-frequency Earth orientation-related processes.     

Mid-Holocene evolution of a tidal basin in the western Netherlands: a model for future changes in the northern Netherlands under conditions of accelerated sea-level rise?

A scenario for the future development of the Dutch Wadden Sea is derived from an evolutionary model for tidal basins during a rise in sea level. The model is based on the evolution of the Atlantic/Subboreal Holland tidal basin, between 7000 BP and 3500 BP. It emphasizes the balance between the storage capacity created by a sea-level rise and the amount of sediment available.

If the rate of relative sea-level rise exceeds the rate of sediment supply, the innermost (central) portions of the basin will not receive sufficient sediment for an intertidal morphology to be preserved. Eventually, sand will be deposited only in tidal channels and in the flood-tidal delta through which the sediment is supplied, mud deposition will occur in the interchannel areas and salt marshes will disappear.     

变形模型可区分度量

在监测网设计阶段，变形体的变形特性一般不是完全清楚。因此，设计的监测方案应当能有效地和足够灵敏地区分几种可能的模型。本文研究了模型的可区分度量，所用的方法具有思路明确、清晰等特点，特别适用于在设计阶段设置指标。文中通过一个模拟网的分析和讨论，说明采用模型可区分度量作为设计准则的必要性。分析的结果还和Forstner法所得的结果进行初步的比较。     

An automatic condensation nucleus counter

Summary Samples of air are sucked into a tube, partly lined with damp blotting paper, and become saturated with water vapour. They are then allowed to expand and the cloud which is formed scatters light from a bulb at the side of the tube to a photomultiplier at one end of it. The sampling and expansion are controlled by solenoid valves and the apparatus works continuously, giving readings at intervals of one minute.     

Archean of Greenland and Fennoscandia

The North Atlantic craton in southern West Greenland mainly consists of a tectonic collage of Mesoarchean continental crustal terranes, which were amalgamated at c. 2.7 Ga and are currently exposed at mid-crustal amphibolite to granulite facies levels. Tonalitic orthogneisses predominate, intercalated with slightly older tholeiitic to andesitic metavolcanic rocks and associated gabbro-anorthosite intrusive complexes. The North Atlantic craton also contains enclaves of Eoarchean, c. 3.86-3.6 Ga orthogneisses and supracrustal rocks including the Isua greenstone （or supracrustal） belt. This is the oldest known assemblage of rocks deposited at the surface of the Earth, comprising mafic pillow lavas, banded iron formations and metasedimentary schists with local disseminated graphite of possible biogenic origin. Eoarchean rocks have not been found in Kola and Karelia in Fennoscandia where most rocks are 2.9-2.7 Ga tonalitic-trondhjemitic-granodioritic orthogneisses with intercalated coeval greenstone belts and amphibolites. Mesoarchean 3.0-3.2 Ga rocks are found in the eastern and western parts of the Karelian province. Subduction-related rocks like the Iringora supra-subduction type ophiolite and basalt-andesite-dacite-rhyolite series volcanic rocks in many greenstone belts, as well as eclogites are found in the Archean of Fennoscandia. A clear distinction between Greenland and Fennoscandia is the abundance of 2.75-2.65 Ga igneous rocks in Fennoscandia which indicates that these two cratons had a separate evolution during the Neoarchean.     

Trends in heat-related mortality in the United States, 1975–2004

This study addresses the long-term trends in heat-related mortality across 29 US metropolitan areas from 1975 to 2004 to discern the spatial patterns and temporal trends in heat vulnerability. Mortality data have been standardized to account for population trends, and seasonal and interannual variability. On days when a city experienced an “oppressive” air mass, mean anomalous mortality was calculated, along with the likelihood that oppressive days led to a mortality response at least one standard deviation above the baseline value. Results show a general decline in heat-related mortality from the 1970s to 1990s, after which the decline seems to have abated. The likelihood of oppressive days leading to significant increases in mortality has shown less of a decline. The number of oppressive days has stayed the same or increased at most metropolitan areas. With US homes near saturation in terms of air-conditioning availability, an aging population is still significantly vulnerable to heat events.