Deep learning-based water quality estimation and anomaly detection using Landsat-8/Sentinel-2 virtual constellation and cloud computing

ABSTRACT

Monitoring of inland water quality is of significant importance due to the increase in water quality related issues, especially within the Midwestern United States. Traditional monitoring techniques, although highly accurate, are vastly insufficient in terms of spatial and temporal coverage. Using a virtual constellation by harmonizing Landsat-8 and Sentinel-2 data a high temporal frequency dataset can be created at a relatively fine spatial scale. In this study, we apply a novel deep learning method for the estimation of blue-green algae (BGA), chlorophyll-α (Chl), fluorescent dissolved organic matter (fDOM), dissolved oxygen (DO), specific conductance (SC), and turbidity. The developed model is evaluated against previously studied machine learning methods and found to outperform multiple linear regression (MLR), support vector machine regression (SVR), and extreme learning machine regression (ELR) generating R² of 0.91 for BGA, 0.88, 0.89, 0.93, 0.87, and 0.84 for Chl, DO, SC, and turbidity respectfully. This model is then applied to all available data ranging from 2013–2018 and time series for each variable were generated for four selected waterbodies. We then use the Empirical Data Analytics (EDA) anomaly detection method on the time series to identify abnormal data points. Upon further analysis, the EDA method successfully identifies abnormal events in water quality. Our results also demonstrate strong correlation between non-optically active variables such as SC with Chl and fDOM. The framework developed in this study represents an efficient and accurate empirical method for inland water quality monitoring at the regional scale.     

Solar and cosmic X-rays above 7.7 keV

Since its launch on March 8, 1967, the OSO-III has continuously observed solar and cosmic X-rays over the 7.7–210 keV range. The sun emits many impulsive X-ray bursts having fluxes several orders of magnitude above the background level of 8 × 10^–9 ergs(cm²-sec)^–1 at 7.7 keV and characteristic times on the order of 5 min. Ninety-five such events having fluxes >3 × 10^–5 ergs(cm²-sec)^–1 were detected in the period from March 8 to June 15, 1967. The cosmic X-ray source Lupus XR-1 has been observed to have a power law spectral form and no significant time variations over a 40-day period. Upper limits have been obtained on the hard X-ray flux of the peculiar galaxy M 87.     

NEAR Infrared Spectrometer Photometry of Asteroid 433 Eros

We present near-infrared spectrometer (NIS) observations (0.8 to 2.4 μm) of the S-type asteroid 433 Eros obtained by the NEAR Shoemaker spacecraft and report results of our Hapke photometric model analysis of data obtained at phase angles ranging from 1.2° to 111.0° and at spatial resolutions of 1.25×2.5 to 2.75×5.5 km/spectrum. Our Hapke model fits successfully to the NEAR spectroscopic data for systematic color variations that accompany changing viewing and illumination geometry. Model parameters imply a geometric albedo at 0.946 μm of 0.27±0.04, which corresponds to a geometric albedo at 0.550 μm of 0.25±0.05. We find that Eros exhibits phase reddening of up to 10% across the phase angle range of 0-100°. We observe a 10% increase in the 1-μm band depth at high phase angles. In contrast, we observe only a 5% increase in continuum slope from 1.486 to 2.363 μm and essentially no difference in the 2-μm band depth at higher phase angles. These contrasting phase effects imply that there are phase-dependent differences in the parametric measurements of 1- and 2-μm band areas, and in their ratio. The Hapke model fits suggest that Eros exhibits a weaker opposition surge than either 951 Gaspra or 243 Ida (the only other S-type asteroids for which we possess disk-resolved photometric observations). On average, we find that Eros at 0.946 μm has a higher geometric albedo and a higher single-scatter albedo than Gaspra or Ida at 0.56 μm; however, Eros's single-particle phase function asymmetry and average surface macroscopic roughness parameters are intermediate between Gaspra and Ida. Only two of the five Hapke model parameters exhibit a notable wavelength dependence: (1) The single-scatter albedo mimics the spectrum of Eros, and (2) there is a decrease in angular width of the opposition surge with increasing wavelength from 0.8 to 1.7 μm. Such opposition surge behavior is not adequately modeled with our shadow-hiding Hapke model, consistent with coherent backscattering phenomena near zero phase.     

Taking the Pulse of Mountains: Ecosystem Responses to Climatic Variability

An integrated program of ecosystem modeling and field studies in the mountains of the Pacific Northwest (U.S.A.) has quantified many of the ecological processes affected by climatic variability. Paleoecological and contemporary ecological data in forest ecosystems provided model parameterization and validation at broad spatial and temporal scales for tree growth, tree regeneration and treeline movement. For subalpine tree species, winter precipitation has a strong negative correlation with growth; this relationship is stronger at higher elevations and west-side sites (which have more precipitation). Temperature affects tree growth at some locations with respect to length of growing season (spring) and severity of drought at drier sites (summer). Furthermore, variable but predictable climate-growth relationships across elevation gradients suggest that tree species respond differently to climate at different locations, making a uniform response of these species to future climatic change unlikely. Multi-decadal variability in climate also affects ecosystem processes. Mountain hemlock growth at high-elevation sites is negatively correlated with winter snow depth and positively correlated with the winter Pacific Decadal Oscillation (PDO) index. At low elevations, the reverse is true. Glacier mass balance and fire severity are also linked to PDO. Rapid establishment of trees in subalpine ecosystems during this century is increasing forest cover and reducing meadow cover at many subalpine locations in the western U.S.A. and precipitation (snow depth) is a critical variable regulating conifer expansion. Lastly, modeling potential future ecosystem conditions suggests that increased climatic variability will result in increasing forest fire size and frequency, and reduced net primary productivity in drier, east-side forest ecosystems. As additional empirical data and modeling output become available, we will improve our ability to predict the effects of climatic change across a broad range of climates and mountain ecosystems in the northwestern U.S.A.     

Fluid and Solute Transport from a Conduit to the Matrix in a Carbonate Aquifer System

Within carbonate systems, the working hypothesis suggests that when a conduit is flooded fluid and solute migrate from the conduit into the matrix. This flux of fluid and solute into the matrix creates a reservoir that can be slowly released once the flooding recedes. Although hypothesized, these fluxes have never been measured. To quantify the distance that a fluid and solute would move into a matrix, the fluxes of fluid and solute from a conduit into a matrix were simulated for nine different carbonate aquifer systems. Two independent numerical approaches were used to simulate (1) fluid flux into the matrix and (2) solute flux into the matrix during a flooding event. When flooding occurs within the conduit, the volume of water transported into and stored in the matrix with a high porosity and high hydraulic conductivity (Floridan Aquifer) was less than 0.34 m³ along a 1 m length of conduit, resulting in a penetration depth of 7.2×10⁻² m into the matrix. In a low porosity and low hydraulic conductivity matrix (Ozark Plateau), the volume of water transported into and stored in the matrix was less than 6.85×10⁻⁵ m³ along a 1 m length of conduit, resulting in a penetration depth of 2.0×10⁻⁴ m into the matrix. Simulated solute flow shows that less than 0.1% of the solute moves in to the matrix. The two approaches demonstrate that during high flow conditions fluid and solute are forced through the conduits, with very little moving into the carbonate matrix. Once the fluid and solute enter a conduit and are moving, they will remain in the conduit until they are discharged at an outlet. Thus, a carbonate matrix does not become a reservoir for solute and fluid during a high-flow event.     

Range-wide population structure of shortnose sturgeonAcipenser brevirostrum based on sequence analysis of the mitochondrial DNA control region

Riverine populations of shortnose sturgeon (Acipenser brevirostrum) once occurred in rivers and estuaries along the east coast of North America from the St. John River, New Brunswick, to the St. Johns River, Florida. Within this range, 19 population segments were identified by the U.S. Federal Shortnose Sturgeon Recovery Team; empirical data supporting this structure is limited. We obtained samples from 11 (12 including a small sample from the Cape Fear River, North Carolina) of these population segments and used PCR and direct sequence analysis of 440 base pairs of the mitochondrial DNA (mtDNA) control region to define the coast-wide genetic population structure of shortnose sturgeon. Collections from most population segments exhibited significant differences in haplotype frequencies with their nearest neighbors, including from the Ogeechee and Savannah Rivers, Georgia (despite the known movement of hatchery-reared offspring from the Savannah into the Ogeechee River). Collections from the Chesapeake Bay and Delaware River exhibited similar haplotype frequencies, suggesting that specimens collected in the Chesapeake Bay had dispersed from the Delaware River. Collections from the Kennebec River and Androscoggin River within a hypothesized single population segment did not exhibit significant differentiation of mtDNA haplotype frequencies. Haplotype frequencies were almost identical between collections from above and below the Holyoke Dam on the Connecticut River, indicating that these aggregations should be managed as a single unit. Our results support the population segment status afforded to shortnose sturgeon in at least the following 9 systems; St. John River, Kennebec-Androscoggin Rivers, upper-lower Connecticut River, Hudson River, Delaware River-Chesapeake Bay, Pee Dee River, Cooper River, Savannah River, and Ogeechee-Altamaha Rivers.     

The application of spatial data processing to land use conflict resolution in the lal lal water catchment of the Ballarat growth corridor: a case study from chapter one of the Buninyong planning scheme

Land use conflict within the Lal Lal Water Catchment is between landowners, the Shire, the Water Board and those directing regional pressures for development. It is exacerbated by inconsistency in building permit appraisal outcomes. The deliberations involved would clearly have benefited from access to spatial data sets through use of a Geographical Information System (GIS) during applications appraisal, in that inconsistencies are most obvious when mapped in relation to the criteria applied. GIS methodologies would ensure that information of a high standard was produced. Three strategic points are identified for the implementation of GIS: at the planning scheme formulation stage; the permit application stage; and the formal conflict resolution stage. Reference to GIS at the first two stages presupposes the existence of a local government area (LGA) spatial data base and a GIS/LIS that can output overlay maps. Application of GIS at the third stage is somewhat ‘after the event’. This study suggests that without such a regional review of the spatial patterns of permit appraisal criteria (including overland flow modelling) each appeal or negotiation is likely to yield an outcome that is inconsistent in relation to other cases. An LGA planning office maintaining a GIS with all layers relating to these criteria will have assembled its spatial data according to the requirements of its planning scheme and will make most use of it in a GIS at the permit appraisal stage.     

Development of lava tubes in the light of observations at Mauna Ulu,Kilauea Volcano,Hawaii

During the 1969–1974 Mauna Ulu eruption on Kilauea's upper east rift zone, lava tubes were observed to develop by four principal processes: (1) flat, rooted crusts grew across streams within confined channels; (2) overflows and spatter accreted to levees to build arched roofs across streams; (3) plates of solidified crust floating downstream coalesced to form a roof; and (4) pahoehoe lobes progressively extended, fed by networks of distributaries beneath a solidified crust. Still another tube-forming process operated when pahoehoe entered the ocean; large waves would abruptly chill a crust across the entire surface of a molten stream crossing through the surf zone. These littoral lava tubes formed abruptly, in contrast to subaerial tubes, which formed gradually. All tube-forming processes were favored by low to moderate volume-rates of flow for sustained periods of time. Tubes thereby became ubiquitous within the pahoehoe flows and distributed a very large proportionof the lava that was produced during this prolonged eruption. Tubes transport lava efficiently. Once formed, the roofs of tubes insulate the active streams within, allowing the lava to retain its fluidity for a longer time than if exposed directly to ambient air temperature. Thus the flows can travel greater distances and spread over wider areas. Even though supply rates during most of 1970–1974 were moderate, ranging from 1 to 5 m³/s, large tube systems conducted lava as far as the coast, 12–13 km distant, where they fed extensive pahoehoe fields on the coastal flats. Some flows entered the sea to build lava deltas and add new land to the island. The largest and most efficient tubes developed during periods of sustained extrusion, when new lava was being supplied at nearly constant rates. Tubes can play a major role in building volcanic edifices with gentle slopes because they can deliver a substantial fraction of lava erupted at low to moderate rates to sites far down the flank of a volcano. We conclude, therefore, that the tendency of active pahoehoe flows to form lava tubes is a significant factor in producing the common shield morphology of basaltic volcanoes.     

Atmospheric CO2 variations at Barrow,Alaska, 1973–1982

The first 10 years (1973–1982) of atmospheric CO₂ measurements at Barrow, Alaska, by the NOAA/GMCC program are described. The paper updates and extends the Barrow CO₂ record presented in Tellus (1982). The data are given in final form, based on recent calibrations of the Scripps Institution of Oceanography, with selected values identified as representative of large, spacescale conditions. Analyses of the data show: (1) a long-term CO₂ average increase of 1.3 ppm per year, but with large year-to-year variations in that growth rate; (2) a suggestion, not statistically significant, of a secular increase in the amplitude of the annual cycle, presumably a reflection of global-scale biospheric variability; and (3) good absolute agreement between the Barrow results and those from four neighboring high latitude sites between 50 and 82°N.