Using machine learning algorithms to predict groundwater levels in Indonesian tropical peatlands.

Tropical peatlands play a vital role in the global carbon cycle as large carbon reservoirs and substantial carbon sinks. Indonesia possesses the largest share (65 %) of tropical peat carbon, equal to 57.4 Gt C. Human perturbations such as extensive logging, deforestation and canalization exacerbate water losses, especially during dry seasons, when low precipitation and high evapotranspiration rates combine with the increased drainage to lower groundwater levels. Drying and increasing temperatures of the surface peat exacerbate ignition and wildfire risks within the peat soils. As such, it is critically important to know how groundwater levels in peatlands are changing over space and time. In this study, a multilinear regression model as well as two machine learning algorithms, random forest and extreme gradient boosting, were used to model groundwater level over the study period (2010-12) within a peat dome impacted by drainage canals and multiple wildfires in Central Kalimantan, Indonesia. Although all three models performed well, based on overall fit, spatial modeling of groundwater level results revealed that extreme gradient boosting (R2 = 0.998, RMSE = 0.048 m) outperformed random forest (R2 = 0.997, RMSE = 0.054 m) and multilinear regression (R2 = 0.970, RMSE = 0.221 m) near drainage canals, which are key fire ignition risk locations in the peatlands. Our study also shows that, on average, elevation and precipitation are the most important parameters influencing groundwater level spatiotemporally.

Positive correlation between wood δ 15N and stream nitrate concentrations in two temperate deciduous forests.

A limitation to understanding drivers of long-term trends in terrestrial nitrogen (N) availability in forests and its subsequent influence on stream nitrate export is a general lack of integrated analyses using long-term data on terrestrial and aquatic N cycling at comparable spatial scales. Here we analyze relationships between stream nitrate concentrations and wood δ 15N records (n = 96 trees) across five neighboring headwater catchments in the Blue Ridge physiographic province and within a single catchment in the Appalachian Plateau physiographic province in the eastern United States. Climatic, acidic deposition, and forest disturbance datasets were developed to elucidate the influence of these factors on terrestrial N availability through time. We hypothesized that spatial and temporal variation of terrestrial N availability, for which tree-ring δ 15N records serve as a proxy, affects the variation of stream nitrate concentration across space and time. Across space at the Blue Ridge study sites, stream nitrate concentration increased linearly with increasing catchment mean wood δ 15N. Over time, stream nitrate concentrations decreased with decreasing wood δ 15N in five of the six catchments. Wood δ 15N showed a significant negative relationship with disturbance and acidic deposition. Disturbance likely exacerbated N limitation by inducing nitrate leaching and ultimately enhancing vegetative uptake. As observed elsewhere, lower rates of acidic deposition and subsequent deacidification of soils may increase terrestrial N availability. Despite the ephemeral modifications of terrestrial N availability by these two drivers and climate, long-term declines in terrestrial N availability were robust and have likely driven much of the declines in stream nitrate concentration throughout the central Appalachians.

Unprocessed Atmospheric Nitrate in Waters of the Northern Forest Region in the U.S. and Canada.

Little is known about the regional extent and variability of nitrate from atmospheric deposition that is transported to streams without biological processing in forests. We measured water chemistry and isotopic tracers (δ18O and δ15N) of nitrate sources across the Northern Forest Region of the U.S. and Canada and reanalyzed data from other studies to determine when, where, and how unprocessed atmospheric nitrate was transported in catchments. These inputs were more widespread and numerous than commonly recognized, but with high spatial and temporal variability. Only 6 of 32 streams had high fractions (>20%) of unprocessed atmospheric nitrate during baseflow. Seventeen had high fractions during stormflow or snowmelt, which corresponded to large fractions in near-surface soil waters or groundwaters, but not deep groundwater. The remaining 10 streams occasionally had some (<20%) unprocessed atmospheric nitrate during stormflow or baseflow. Large, sporadic events may continue to be cryptic due to atmospheric deposition variation among storms and a near complete lack of monitoring for these events. A general lack of observance may bias perceptions of occurrence; sustained monitoring of chronic nitrogen pollution effects on forests with nitrate source apportionments may offer insights needed to advance the science as well as assess regulatory and management schemes.

Surface water quality is improving due to declining atmospheric N deposition.

We evaluated long-term surface water nitrate and atmospheric nitrogen (N) deposition trends for a group of nine predominantly forested Appalachian Mountain watersheds during a recent multidecadal period (1986-2009) in which regional NOx emissions have been progressively reduced. Statistical analysis showed unexpected linear declines in both annual surface water nitrate-N concentrations (mean =46.4%) and yields (mean =47.7%) among the watersheds corresponding to comparable declines in annual wet N deposition (mean =34.4%) resulting from U.S. NOx emission control programs during the same time period. Nitrate-N concentration trends were robust across a large geographical region and appeared insensitive to watershed size across several orders of magnitude-suggesting that the improvements in water quality are probably propagated to surface and estuarine waters downstream. Surface waters are thus responding to declining atmospheric N deposition in much the same way they responded to declining sulfur deposition-although only one watershed showed a 1:1 relationship. Application of a kinetic N saturation model indicated that all nine forested watersheds are exhibiting signs of N saturation as evidenced by a limited, but variable, efficiency of demand for N. Further reductions in N deposition would be expected to produce additional reductions in streamwater N loads.

Variation in physicochemical responses to urbanization in streams between two Mid-Atlantic physiographic regions.

Urban development substantially alters the physicochemistry of streams, resulting in biodiversity and ecosystem function loss. However, interregional comparisons of physicochemical impact in urban streams suggest that geoclimatic heterogeneity may influence the extent of degradation. In the Mid-Atlantic United States, the adjacent Coastal Plain and Piedmont physiographic provinces possess distinctly different hydrogeomorphic properties that may influence how stream ecosystems respond to urbanization. Recent bioassessments have demonstrated that biotic sensitivity to urbanization is relatively acute in the Piedmont, suggesting that physicochemical change as a consequence of urbanization may be greater in that province. We compared hydrologic, chemical, and thermal characteristics of Mid-Atlantic Coastal Plain and Piedmont first- through fifth-order streams along gradients of impervious surface cover (ISC) at multiple spatial scales. Linear models were applied to test if conditions in rural streams and the degree of impact from ISC varied between provinces. Mean and maximum summer temperatures in Piedmont streams increased more per unit of ISC than in the Coastal Plain. Contrary to expectations, however, variables that quantified high-flow event frequency, magnitude and duration, exhibited significantly greater impact along the ISC gradient in the Coastal Plain. Most chemical changes associated with increasing ISC were similar in the two provinces, although the interregional chemical composition of rural streams differed substantially for most parameters. Our findings demonstrate consistent interregional heterogeneity in stream ecosystem responses to urbanization. Landscape-scale management decisions with stream ecosystem conservation, mitigation, or restoration as a goal must therefore carefully consider the geoclimatic context in order to maximize effectiveness.

Forest to reclaimed mine land use change leads to altered ecosystem structure and function.

The United States' use of coal results in many environmental alterations. In the Appalachian coal belt region, one widespread alteration is conversion of forest to reclaimed mineland. The goal of this study was to quantify the changes to ecosystem structure and function associated with a conversion from forest to reclaimed mine grassland by comparing a small watershed containing a 15-year-old reclaimed mine with a forested, reference watershed in western Maryland. Major differences were apparent between the two watersheds in terms of biogeochemistry. Total C, N, and P pools were all substantially lower at the mined site, mainly due to the removal of woody biomass but also, in the case of P, to reductions in soil pools. Mineral soil C, N, and P pools were 96%, 79%, and 69% of native soils, respectively. Although annual runoff from the watersheds was similar, the mined watershed exhibited taller, narrower storm peaks as a result of a higher soil bulk density and decreased infiltration rates. Stream export of N was much lower in the mined watershed due to lower net nitrification rates and nitrate concentrations in soil. However, stream export of sediment and P and summer stream temperature were much higher. Stream leaf decomposition was reduced and macroinvertebrate community structure was altered as a result of these changes to the stream environment. This land use change leads to substantial, long-term changes in ecosystem capital and function.

Contemporary trends in the acid-base status of two acid-sensitive streams in western Maryland.

Recovery of streamwater acid neutralizing capacity (ANC) resulting from declines in regional acid deposition was examined using contemporary (1990-2005) data from two long-term monitoring stations located on the Appalachian Plateau in western Maryland, U.S. Two computational methods were used to estimate daily, monthly, and annual fluxes and discharge-weighted concentrations of ANC, sulfate, nitrate, and base cations over the period of record, and two statistical methods were used to evaluate long-term trends in fluxes and concentrations. The methods used to estimate concentrations, as well as the statistical techniques, produced very similar results, underlining the robustness of the identified trends. We found clear evidence that streamwater sulfate concentrations have declined at an average rate of about 3 microeq L(-1) yr(-1) at the two sites due to a 34% reduction in wet atmospheric sulfur deposition. Trends in nitrate concentrations appear to be related to other watershed factors, especially forest disturbance. The best evidence of recovery is based on a doubling of ANC (from 21 to 42 microeq L(-1)) at the more acid-sensitive site over the 16-year period. A slowing, or possible reversal, in the sulfate, nitrate, and SBC trends is evident in our data and may portend a decline in the rate of--or end to--further recovery.

Analysis of trends in episodic acidification of streams in western Maryland.

In this study we report on changes in the magnitude and mechanisms of episodic acidification of a small acid-sensitive stream in western Maryland (U.S.) during the 1990s, a period in which wet sulfate deposition declined by 10-25% due to implementation of the Clean Air Act Amendments (CAAA) of 1990. We observed a relatively minor trend in the magnitude of episodic acidification over this period, as measured bytransient changes in acid neutralizing capacity (deltaANC) and minimum values of ANC (ANC(min)) during 22 events sampled prior to and following CAAA implementation. Any relationship to changes in atmospheric deposition appears to be confounded by large hydroclimatological variability between the two sampling periods. Nonetheless, results obtained prior to implementation of the CAAA indicated that the mechanism of episodic acidification was mostly attributable to flushing of accumulated sulfate from the watershed, whereas results obtained post-CAAA indicated domination by base cation dilution. This shift in the mechanism of episodic acidification is qualitatively consistent with hydrochemical theory, as well as with empirical results from surface waters in other regions where dramatic declines in sulfate deposition have taken place.

The influence of substrate quality and stream size on wood decomposition dynamics.

Woody materials decayed more rapidly in a first order stream than in larger streams in eastern Quebec, Canada. The rate of annual mass loss (k) was highest (k=1.20) for alder wood chips in a first order stream and lowest (k=0.04) for black spruce wood chips in a ninth order stream. Decay rates for woody materials in a first order stream were inversely related to their initial lignin to nitrogen ratios. In larger streams, decay rates of woody materials were inversely related to their initial lignin concentrations. A number of quantifiable relationships were found to exist between the initial lignin and nitrogen contents of woody materials and the nitrogen dynamics of decaying wood.