The application of remotely sensed data to a catchment-scale nutrient transport model
Nutrient transport models are being used increasingly as a tool for the research and management of nutrient enrichment (eutrophication) of freshwaters. Phosphorus is seen as the main cause of freshwater eutrophication. A nutrient transport model was acquired that could simulate the movement of phosphorus through a catchment. The SWAT model from the US Department of Agriculture, Agricultural Research Service appeared to suit the requirements of a catchment-scale, continual time model that was distributed in nature. It is based on physical processes in order that predictions could be made for land management practices or environmental conditions that had been absent in calibration processes.;Remote sensing technology has the potential to improve on estimates of distributed variables based on spot measurements and interpolative techniques. The initial intention of this project was to estimate several parameters from remote sensing images and use them as input to the chosen nutrient transport model. The SWAT model is only able to utilise mapped data for soil types and land cover. Whilst the latter can be extracted from various remote-sensing devices the former cannot. Synthetic aperture radar (SAR) has the potential to estimate several of the parameters considered influential to the movement of nutrients in a catchment. This study utilised five SAR images to investigate the potential of extracting: (i) land cover data, (ii) soil moisture, (iii) soil surface roughness, (iv) soil organic matter content (v) oilseed rape leaf area index and (vi) oilseed rape biomass. No significant relationships were found between any of the soil parameters and radar backscatter using linear regression. It is thought that this may be due to the excessive moisture levels at the time of sampling, but sampling intensity could also have been better. Likewise no significant relationships were found between the botanical parameters and radar backscatter. Wheat and oilseed rape characteristics were also collected and applied to the MIMICS model to assess the technology of radiative transfer models in the UK. There was a significant correlation between the backscatter values obtained through the MIMICS model and the backscatter from mature wheat to the SAR images to but not to green wheat or oilseed rape.;A land cover map was generated using a multi date composite of three of the SAR images. The images were acquired in May, July and August of 1999. Land-classes were assigned using supervised maximum likelihood estimation (MLE) and unsupervised training. Out of 11 classes of land cover found on the Stonton Brook, 11 were identified using the supervised training and MLE and only seven using the unsupervised training. The former method acquired a total accuracy of 46 % against the latter's 53%. On applying the classification schemes to a field boundary map the total accuracies improved to 58 % and 54 % respectively. Both maps were regarded as moderately accurate and both were used in the SWAT model.