RESUMEN
Robust quantification of vegetative biomass using satellite imagery using one or more forms of machine learning (ML) has hitherto been hindered by the extent and quality of training data. Here, we showcase how ML predictive demonstrably improves when additional training data is used. We collated field datasets of pasture biomass obtained via destructive sampling, 'C-Dax' reflective measurements and rising plate meters (RPM) from ten livestock farms across four States in Australia. Remotely sensed data from the Sentinel-2 constellation was used to retrieve aboveground biomass using a novel machine learning paradigm hereafter termed "SPECTRA-FOR" (Spectral Pasture Estimation using Combined Techniques of Random-forest Algorithm for Features Optimisation and Retrieval). Using this framework, we show that the low temporal resolution of Sentinel-2 in high latitude regions with persistent cloud cover leads to extensive gaps between cloud-free images, hindering model performance and, thus, contemporaneous ability to forecast real-time pasture biomass. By leveraging the spectral consistency between Sentinel-2 and Planet Lab SuperDove to overcome this limitation, we used ten spectral bands of Sentinel-2, four bands of Sentinel-2 as a proxy for pre-2022 SuperDove (referred to as synthetic SuperDove or SSD), and the actual SuperDove (ASD), given that SuperDove imagery has a higher resolution and more frequent passage compared with Sentinel-2. Using their respective bands as input features to SPECRA-FOR, model performance for the ten bands of Sentinel-2 were R2 = 0.87, root mean squared error (RMSE) of 439 kg DM/ha and mean absolute error (MAE) of 255 kg DM/ha, while that for SSD increased to an R2 of 0.92, RMSE of 346 kg DM/ha and MAE = 208 kg DM/ha. The study revealed the importance of robust data mining, imagery harmonisation and model validation for accurate real-time modelling of pasture biomass with ML.
Asunto(s)
Aprendizaje Automático , Imágenes Satelitales , Imágenes Satelitales/métodos , Biomasa , Granjas , AustraliaRESUMEN
During the 2019-2020 Australian bushfire season, large expanses (~47%) of agricultural and forested land in the Upper Murray River catchment of southeastern (SE) Australia were burned. Storm activity and rainfall following the fires increased sediment loads in rivers, resulting in localized fish kills and widespread water-quality deterioration. We collected water samples from the headwaters of the Murray River for sediment and contaminant analysis and assessed changes in water quality using long-term monitoring data. A robust runoff routing model was used to estimate the effect of fire on sediment loads in the Murray River. Peak turbidity in the Murray River reached values of up to 4200 nephelometric turbidity units (NTU), shown as pitch-black water coming down the river. The increase in suspended solids was accompanied by elevated nutrient concentrations during post-bushfire runoff events. The model simulations demonstrated that the sediment load could be five times greater in the first year after a bushfire than in the prefire condition. It was estimated that Lake Hume, a large reservoir downstream from fire-affected areas, would receive a maximum of 600 000 metric tonnes of sediment per month in the period immediately following the bushfire, depending on rainfall. Total zinc, arsenic, chromium, nickel, copper, and lead concentrations were above the 99% toxicant default guideline values (DGVs) for freshwater ecosystems. It is also likely that increased nutrient loads in Lake Hume will have ongoing implications for algal dynamics, in both the lake and the Murray River downstream. Information from this study provides a valuable basis for future research to support bushfire-related policy developments in fire-prone catchments and the mitigation of postfire water quality and aquatic ecosystem impacts. Integr Environ Assess Manag 2021;17:1203-1214. © 2021 Commonwealth of Australia. Integrated Environmental Assessment and Management © 2021 Society of Environmental Toxicology & Chemistry (SETAC).
Asunto(s)
Ecosistema , Sedimentos Geológicos , Animales , Australia , Monitoreo del Ambiente , RíosRESUMEN
Numerosos actores socioeconómicos y políticos utilizan las estimaciones de la superficie cultivada para planificar, reducir la incertidumbere o mejorar la asiganción de recursos. Para resultar confiables y útiles, las estimaciones deben basarse en una metodología debidamente documentada, reproducible para el espacio y en el tiempo, idependiente del observador, evaluable de manera cuantitativa. ¿En qué medida se satisfacen los criterio anteriores en Argentina?. Más allá de su utilidad, la información disponible incorpora fuentes de incertidumbre que afectan seriamente las estimaciones. Estas icluyen las dificulatdes para referir las estimaciones a un área determinada, las posibilidades de los informantes de integrar la información local, la ausencia de protocolos claros y diferencias asociadas a la heterogeneidad de formación, motivación y compromiso de los informantes. La comparación de las estimaciones en dos agencias independientes para una año particular arroja, para las mismas áreas, diferencias de hasta el 24 por ciento en el área sembrada con trigo. Esta diferencia es muy superior a las variaciones interanuales que pretenden detectarse. El análisis multiespectral y multitemporal de imágenes satelitales permite discriminar tipos de cobertura del suelo sobre la base de su comportamiento fenológico. La combinación de información satelital provenientes de sensores con distinta resolución espacial ofrece enormes posibilidades para descripción de los tipos de cobertura del suelo y la estimación de superficies agrícolas. En tal sentido se presenta una propuesta operativa, basada en el uso de imágenes Landsat TM, SAC-C y AVHRR/NOAA, para la evolución regional de la superficie cultivada en Mercosur