Identifying clusters of precipitation for the Brazilian Legal Amazon based on magnitude of trends and its correlation with sea surface temperature.

Prioritizing watershed management interventions relies on delineating homogeneous precipitation regions. In this study, we identify these regions in the Brazilian Legal Amazon based on the magnitude of Sen's Slope trends using annual precipitation data from September to August, employing the Google Earth Engine platform. Utilizing the silhouette method, we determine four distinct clusters representing zones of homogeneous precipitation patterns. Cluster 0 exhibits a significant median increase in precipitation of 3.20 mm year-1 over the period from 1981 to 2020. Cluster 1 shows a notable increase of 8.13 mm year-1, while Clusters 2 and 3 demonstrate reductions in precipitation of - 1.61 mm year-1 and - 3.87 mm year-1, respectively, all statistically significant. Notably, the region known as the arc of deforestation falls within Cluster 2, indicating a concerning trend of reduced precipitation. Additionally, our analysis reveals significant correlations between Sea Surface Temperature (SST) in various oceanic regions and precipitation patterns over the Brazilian Legal Amazon. Particularly noteworthy is the strong positive correlation with SST in the South Atlantic, while negative correlations are observed with SST in the South Pacific and North Atlantic. These findings provide valuable insights for enhancing climate adaptation strategies in the Brazilian Legal Amazon region.

Transformation of Brazil's biomes: The dynamics and fate of agriculture and pasture expansion into native vegetation.

Land use and cover change (LUCC) in Brazil encompass a complex interplay of diverse factors across different biomes. Understanding these dynamics is crucial for informed decision-making and sustainable land management. In this study, we comprehensively analyzed LUCC patterns and drivers using 30 m resolution MapBiomas Collection 6.0 data (1985-2020). By mapping deforestation of primary and secondary natural vegetation, natural vegetation regeneration, and transitions between pasture, soybean, agriculture, and irrigation, we shed light on the intricate nature of LUCC in Brazil. Our findings highlight significant and increasing trends of deforestation in primary vegetation in the country. Simultaneously, the Atlantic Forest, Caatinga, Pampa, and other regions of the Cerrado have experienced intensification processes. Notably, the pasture area in Brazil reached its peak in 2006 and has since witnessed a gradual replacement by soybean and other crops. While pasture-driven deforestation persists in most biomes, the net pasture area has only increased in the Amazon and Pantanal, decreasing in other biomes due to the conversion of pasturelands to intensive cropping in other regions. Our analysis further reveals that primary and secondary vegetation deforestation accounts for a substantial portion of overall forest loss, with 72 % and 17 %, respectively. Of the cleared areas, 48 % were in pasture, 9 % in soybean cultivation, and 16 % in other agricultural uses in 2020. Additionally, we observed a lower rate of deforestation in the Atlantic Forest, a biome that has been significantly influenced by anthropogenic activities since 1986. This holistic quantification of LUCC dynamics provides a solid foundation for understanding the impacts of these changes on local to continental-scale land-atmosphere interactions. By unraveling the complex nature of LUCC in Brazil, this study aims to contribute to the development of effective strategies for sustainable land management and decision-making processes.

Land use and land cover changes and their impacts on surface-atmosphere interactions in Brazil: A systematic review.

Major land use and land cover changes (LULCC) have taken place in Brazil, including large scale conversion of forest to agriculture. LULCC alters surface-atmosphere interactions, changing the timing and magnitude of energy fluxes, impacting the partitioning of available energy, and therefore the climate and water balance. The objective of this work was to provide a detailed analysis of how LULCC has affected surface-atmosphere interactions over the Brazilian territory, particularly focusing on impacts on precipitation (P), evapotranspiration (ET), and atmospheric humidity (h). Our systematic review yielded 61 studies, with the Amazon being the most studied biome followed by the Cerrado. P was the most analyzed variable, followed by ET. Few papers analyzed LULCC impacts on h. For the Amazon biome, decreasing dry season P and in annual ET were reported. In the Cerrado biome, decreasing P in the wet and dry seasons and decreasing dry season ET were the most common result. For the Atlantic Forest biome, increasing annual P and increasing wet season ET, likely due to reforestation, were reported. Few studies documented LULCC impacts on surface-atmosphere interactions over the Brazilian biomes Caatinga, Pantanal and Pampa. Therefore, new research is needed to assess impacts of LULCC on these biomes, including assessments of atmospheric moisture recycling, and interactions of LULCC with global climate and climate extremes including droughts.

A novel ensemble computational intelligence approach for the spatial prediction of land subsidence susceptibility.

Land subsidence (LS) is a significant problem that can cause loss of life, damage property, and disrupt local economies. The Semnan Plain is an important part of Iran, where LS is a major problem for sustainable development and management. The plain represents the changes occurring in 40% of the country. We introduce a novel-ensemble intelligence approach (called ANN-bagging) that uses bagging as a meta- or ensemble-classifier of an artificial neural network (ANN) to predict LS spatially on the Semnan Plain in Semnan Province, Iran. The ensemble model's goodness-of-fit (to training data) and prediction accuracy (of the validation data) are compared to benchmarks set by ANN-bagging. A total of 96 locations of LS and 12 LS conditioning factors (LSCFs) were collected. Each feature in the LS inventory map (LSIM) was randomly assigned to one of four groups or folds, each comprising 25% of cases. The novel ensemble model was trained using 75% (3 folds) and validated with the remaining 25% (1 fold) in a four-fold cross-validation (CV) system, which is used to control for the effects of the random selection of the training and validation datasets. LSCFs for LS prediction were selected using the information-gain ratio and multi-collinearity test methods. Factor significance was evaluated using a random forest (RF) model. Groundwater drawdown, land use and land cover, elevation, and lithology were the most important LSCFs. Using the k-fold CV approaches, twelve LS susceptibility maps (LSSMs) were prepared as each fold employed all three models (ANN-bagging, ANN, and bagging). The LS susceptibility mapping showed that between 5.7% and 12.6% of the plain had very high LS susceptibility. All three models produced LS susceptibility maps with acceptable prediction accuracies and goodness-of-fits, but the best maps were produced by the ANN-bagging ensemble method. Overall, LS risk was highest in agricultural areas with high groundwater drawdown in the flat lowlands on quaternary sediments (Qcf). Groundwater extraction rates should be monitored and potentially limited in regions of severe or high LS susceptibility. This investigation details a novel methodology that can help environmental planners and policy makers to mitigate LS to help achieve sustainability.

Machine learning approaches for spatial modeling of agricultural droughts in the south-east region of Queensland Australia.

A quantitative understanding of the hydro-environmental factors that influence the occurrence of agricultural drought events would enable more strategic climate change adaptation and drought management plans. Practical drought hazard mapping remains challenging due to possible exclusion of the most pertinent drought drivers, and to the use of inadequate predictive models that cannot describe drought adequately. This research aims to develop new approaches to map agricultural drought hazard with state-of-the-art machine learning models, including classification and regression trees (CART), boosted regression trees (BRT), random forests (RF), multivariate adaptive regression splines (MARS), flexible discriminant analysis (FDA) and support vector machines (SVM). Hydro-environmental datasets were used to calculate the relative departure of soil moisture (RDSM) for eight severe droughts for drought-prone southeast Queensland, Australia, over the period 1994-2013. RDSM was then used to generate an agricultural drought inventory map. Eight hydro-environmental factors were used as potential predictors of drought. The goodness-of-fit and predictive performance of all models were evaluated using different threshold-dependent and threshold-independent methods, including the true skill statistic (TSS), Efficiency (E), F-score, and the area under the receiver operating characteristic curve (AUC-ROC). The RF model (AUC-ROC = 97.7%, TSS = 0.873, E = 0.929, F-score = 0.898) yielded the highest accuracy, while the FDA model (with AUC-ROC = 73.9%, TSS = 0.424, E = 0.719, F-score = 0.512) showed the worst performance. The plant available water holding capacity (PAWC), mean annual precipitation, and clay content were the most important variables to be used for predicting the agricultural drought. About 21.2% of the area is in high or very high drought risk classes, and therefore, warrant drought and environmental protection policies. Importantly, the models do not require data on the precipitation anomaly for any given drought year; the spatial patterns in AGH were consistent for all drought events, despite very different spatial patterns in precipitation anomaly among events. Such machine-learning approaches are able to construct an overall risk map, thus assisting in the adoption of a robust drought contingency planning measure not only for this area, but also, in other regions where drought presents a pressing challenge, including its influence on key practical dimensions of social, environmental and economic sustainability.

Modelling Runoff and Sediment Loads in a Developing Coastal Watershed of the US-Mexico Border.

Urbanization can increase sheet, rill, gully, and channel erosion. We quantified the sediment budget of the Los Laureles Canyon watershed (LLCW), which is a mixed rural-urbanizing catchment in Northwestern Mexico, using the AnnAGNPS model and field measurements of channel geometry. The model was calibrated with five years of observed runoff and sediment loads and used to evaluate sediment reduction under a mitigation scenario involving paving roads in hotspots of erosion. Calibrated runoff and sediment load had a mean-percent-bias of 28.4 and - 8.1, and root-mean-square errors of 85% and 41% of the mean, respectively. Suspended sediment concentration (SSC) collected at different locations during one storm-event correlated with modeled SSC at those locations, which suggests that the model represented spatial variation in sediment production. Simulated gully erosion represents 16%-37% of hillslope sediment production, and 50% of the hillslope sediment load is produced by only 23% of the watershed area. The model identifies priority locations for sediment control measures, and can be used to identify tradeoffs between sediment control and runoff production. Paving roads in priority areas would reduce total sediment yield by 30%, but may increase peak discharge moderately (1.6%-21%) at the outlet.

Land subsidence modelling using tree-based machine learning algorithms.

Land subsidence (LS) is among the most critical environmental problems, affecting both agricultural sustainability and urban infrastructure. Existing methods often use either simple regression models or complex hydraulic models to explain and predict LS. There are few studies that identify the risk factors and predict the risk of LS using machine learning models. This study compares four tree-based machine learning models for land subsidence hazard modelling at a study area in Hamadan plain (Iran). The study also analyzes the importance of six risk factors including topography (elevation, slope), geomorphology (distance from stream, drainage density), hydrology (groundwater drawdown) and lithology on LS. Thematic layers of each variable related to the LS phenomenon are prepared and utilized as the inputs to the four tree-based machine learning models, including the Rule-Based Decision Tree (RBDT), Boosted Regression Trees (BRT), Classification And Regression Tree (CART), and the Random Forest (RF) algorithms to produce a consolidated LS hazard map. The accuracy of the generated maps is then evaluated using the area under the receiver operating characteristic curve (AUC) and the True Skill Statistics (TSS). The RF approach had the lowest predictive error for mapping the LS hazard (i.e., AUC 96.7% for training, AUC 93.8% for validation, TSS 0.912 for training, TSS 0.904 for validation) followed by BRT. Groundwater drawdown was seen to be the most influential factor that contributed to land subsidence in the present study area, followed by lithology and distance from the stream network.

Land subsidence hazard modeling: Machine learning to identify predictors and the role of human activities.

Land subsidence caused by land use change and overexploitation of groundwater is an example of mismanagement of natural resources, yet subsidence remains difficult to predict. In this study, the relationship between land subsidence features and geo-environmental factors is investigated by comparing two machine learning algorithms (MLA): maximum entropy (MaxEnt) and genetic algorithm rule-set production (GARP) algorithms in the Kashmar Region, Iran. Land subsidence features (N = 79) were mapped using field surveys. Land use, lithology, the distance from traditional groundwater abstraction systems (Qanats), from afforestation projects, from neighboring faults, and the drawdown of groundwater level (DGL) (1991-2016) were used as predictive variables. Jackknife resampling showed that DGL, distance from afforestation projects, and distance from Qanat systems are major factors influencing land subsidence, with geology and faults being less important. The GARP algorithm outperformed the MaxEnt algorithm for all performance metrics. The performance of both models, as measured by the area under the receiver-operator characteristic curve (AUROC), decreased from 88.9-94.4% to 82.5-90.3% when DGL was excluded as a predictor, though the performance of GARP was still good to excellent even without DGL. MLAs produced maps of subsidence risk with acceptable accuracy, both with and without data on groundwater drawdown, suggesting that MLAs can usefully inform efforts to manage subsidence in data-scarce regions, though the highest accuracy requires data on changes in groundwater level.

Stream channel erosion in a rapidly urbanizing region of the US-Mexico border: documenting the importance of channel hardpoints with Structure-from-Motion photogrammetry.

Urbanization can lead to accelerated stream channel erosion, especially in areas experiencing rapid population growth, unregulated urban development on erodible soils, and variable enforcement of environmental regulations. A combination of field surveys and Structure-from-Motion (SfM) photogrammetry techniques was used to document spatial patterns in stream channel geometry in a rapidly urbanizing watershed, Los Laureles Canyon (LLCW), in Tijuana, Mexico. Ground-based SfM photogrammetry was used to map channel dimensions with 1 to 2 cm vertical mean error for four stream reaches (100-300 m long) that were highly variable and difficult to survey with a differential GPS. Regional channel geometry curves for LLCW had statistically larger slopes and intercepts compared with regional curves developed for comparable, undisturbed reference channels. Cross-sectional areas of channels downstream of hardpoints, such as concrete reaches or culverts, were up to 64 times greater than reference channels, with enlargement persisting, in some cases, up to 230 m downstream. Percentage impervious cover was not a good predictor of channel enlargement. Proximity to upstream hardpoint, and lack of riparian and bank vegetation paired with highly erodible bed and bank materials may account for the instability of the highly enlarged and unstable cross-sections. Channel erosion due to urbanization accounts for approximately 25-40% of the total sediment budget for the watershed, and channel erosion downstream of hardpoints accounts for one-third of all channel erosion. Channels downstream of hardpoints should be stabilized to prevent increased inputs of sediment to the Tijuana Estuary and local hazards near the structures, especially in areas with urban settlements near the stream channel.

Measuring ephemeral gully erosion rates and topographical thresholds in an urban watershed using unmanned aerial systems and structure from motion photogrammetric techniques.

Both rural and urban development can lead to accelerated gully erosion. Quantifying gully erosion is challenging in environments where gullies are rapidly repaired, and in urban areas where microtopographic complexity complicates the delineation of contributing areas. This study used unmanned aerial vehicles (UAVs) and Structure-from-Motion (SfM) photogrammetric techniques to quantify gully erosion in the Los Laureles Canyon watershed, a rapidly urbanizing watershed in Tijuana, Mexico. Following a storm event, the gully network extent was mapped using an orthomosaic (0.038 m pixel size); the local slope and watershed area contributing to each gully head were mapped with a Digital Surface Model (0.3 m pixel size). Gullies formed almost exclusively on unpaved roads which had erodible soils and concentrated flow. Management practices (e.g. road maintenance that fill gullies after large storms) contributed to total sediment production at the watershed scale. Sediment production from gully erosion was higher and threshold values of slope and drainage area for gully incision were lower than ephemeral gullies reported for agricultural settings. This indicates high vulnerability of unpaved roads to gully erosion which is consistent with high soil erodibility and low critical shear stress measured in the laboratory with a mini jet-erosion-test device. Future studies that evaluate effects of different soil types on gully erosion rates for unpaved roads, as well as those that model effects of management practices such as road paving and their impact on runoff, soil erosion, and sediment loads are needed to advance sediment management and planning in urban watersheds.

Modelling Ephemeral Gully Erosion from Unpaved Urban Roads: Equifinality and Implications for Scenario Analysis.

Modelling gully erosion in urban areas is challenging due to difficulties with equifinality and parameter identification, which complicates quantification of management impacts on runoff and sediment production. We calibrated a model (AnnAGNPS) of an ephemeral gully network that formed on unpaved roads following a storm event in an urban watershed (0.2 km2) in Tijuana, Mexico. Latin hypercube sampling was used to create 500 parameter ensembles. Modelled sediment load was most sensitive to the Soil Conservation Service (SCS) curve number, tillage depth (Td), and critical shear stress (τc). Twenty-one parameter ensembles gave acceptable error (behavioural models), though changes in parameters governing runoff generation (SCS curve number, Manning's n) were compensated by changes in parameters describing soil properties (TD, τc, resulting in uncertainty in the optimal parameter values. The most suitable parameter combinations or "behavioural models" were used to evaluate uncertainty under management scenarios. Paving the roads increased runoff by 146-227%, increased peak discharge by 178-575%, and decreased sediment load by 90-94% depending on the ensemble. The method can be used in other watersheds to simulate runoff and gully erosion, to quantify the uncertainty of model-estimated impacts of management activities on runoff and erosion, and to suggest critical field measurements to reduce uncertainties in complex urban environments.

Oiling accelerates loss of salt marshes, southeastern Louisiana.

The 2010 BP Deepwater Horizon (DWH) oil spill damaged thousands of km2 of intertidal marsh along shorelines that had been experiencing elevated rates of erosion for decades. Yet, the contribution of marsh oiling to landscape-scale degradation and subsequent land loss has been difficult to quantify. Here, we applied advanced remote sensing techniques to map changes in marsh land cover and open water before and after oiling. We segmented the marsh shorelines into non-oiled and oiled reaches and calculated the land loss rates for each 10% increase in oil cover (e.g. 0% to >70%), to determine if land loss rates for each reach oiling category were significantly different before and after oiling. Finally, we calculated background land-loss rates to separate natural and oil-related erosion and land loss. Oiling caused significant increases in land losses, particularly along reaches of heavy oiling (>20% oil cover). For reaches with ≥20% oiling, land loss rates increased abruptly during the 2010-2013 period, and the loss rates during this period are significantly different from both the pre-oiling (p < 0.0001) and 2013-2016 post-oiling periods (p < 0.0001). The pre-oiling and 2013-2016 post-oiling periods exhibit no significant differences in land loss rates across oiled and non-oiled reaches (p = 0.557). We conclude that oiling increased land loss by more than 50%, but that land loss rates returned to background levels within 3-6 years after oiling, suggesting that oiling results in a large but temporary increase in land loss rates along the shoreline.

Rapid increase in copper concentrations in a new marina, San Diego Bay.

Concentrations of copper in water rose rapidly following the introduction of boats to a new marina in San Diego Bay. Two months after the marina reached half its capacity, a majority of water samples exceeded chronic and acute criteria for dissolved copper, and copper concentrations in several samples exceeded the highest concentrations observed in another marina that has been listed as an impaired water body. A box model suggested that a small fraction of the leached copper was sequestered in sediment. Copper concentrations in water entering the marina from the bay was more than half the chronic concentration limit, so only 50% of marina boat capacity could be accommodated without exceeding the chronic criterion more than 50% of the time. Copper concentrations in water may increase rapidly following boat introduction in small marinas, but could return to pre-introduction levels by controlling boat numbers or reducing use of copper-based paints.

Soil salinity and exchangeable cations in a wastewater irrigated area, India.

The salinity and cation composition of water and soil were documented in a large (98 km(2)) wastewater-irrigated area (WIA) downstream of Hyderabad, India. The wastewater, which flows in a river that passes through the city, had a high to very high salinity hazard (EC = 1.1-3.0 dS m(-1)) that increased with distance from the city. The EC of soil irrigated by wastewater sampled within 8 km of the city was 6.2 to 8.4 times the EC of soil irrigated by uncontaminated groundwater. Between 57 to 100% of soil samples in the upper 10 cm within 8 km of the city exceeded the salinity tolerance of rice (Oryza sativa L.). Soil salinity fell rapidly after 8 km downstream and changed most in the upper 0 to 5 cm of the soil, indicating retention of cations in the upper soil horizon. The effect of wastewater irrigation on soil exchangeable cations was most evident for Na(+) (Exch-Na) near the city (<8 km downstream), where Exch-Na averaged 20 to 22 times the Exch-Na in soils irrigated by groundwater outside the WIA. Exchangeable Mg(+) and K(+) correlated with clay percentage, though both still had higher concentrations near the city controlling for clay content. Near the city, where salinity and Exch-Na concentrations were highest, farmers had replaced rice with para grass [Brachiaria mutica (Forsk.)], which has higher salinity tolerance and expanding demand as a fodder crop. Salinity may constrain rice production in wastewater-irrigated areas of India and elsewhere.

The rate and extent of deforestation in watersheds of the southwestern Amazon basin.

The rate and extent of deforestation determine the timing and magnitude of disturbance to both terrestrial and aquatic ecosystems. Rapid change can lead to transient impacts to hydrology and biogeochemistry, while complete and permanent conversion to other land uses can lead to chronic changes. A large population of watershed boundaries (N=4788) and a time series of Landsat TM imagery (1975-1999) in the southwestern Amazon Basin showed that even small watersheds (2.5-15 km2) were deforested relatively slowly over 7-21 years. Less than 1% of all small watersheds were more than 50% cleared in a single year, and clearing rates averaged 5.6%/yr during active clearing. A large proportion (26%) of the small watersheds had a cumulative deforestation extent of more than 75%. The cumulative deforestation extent was highly spatially autocorrelated up to a 100-150 km lag due to the geometry of the agricultural zone and road network, so watersheds as large as approximately 40000 km2 were more than 50% deforested by 1999. The rate of deforestation had minimal spatial autocorrelation beyond a lag of approximately 30 km, and the mean rate decreased rapidly with increasing area. Approximately 85% of the cleared area remained in pasture, so deforestation in watersheds of Rondônia was a relatively slow, permanent, and complete transition to pasture, rather than a rapid, transient, and partial cutting with regrowth. Given the observed landcover transitions, the regional stream biogeochemical response is likely to resemble the chronic changes observed in streams draining established pastures, rather than a temporary pulse from slash-and-burn.