You are here : University of Gloucestershire > Research > Research Repository

Creating a Rapid Identification Toolkit For Rural Ponds

Tools

Smith, Lucy (2023) Creating a Rapid Identification Toolkit For Rural Ponds. Masters thesis, University of Gloucestershire. doi:10.46289/7NH3YT56

Preview

Text (Thesis Final)
14033 Smith, Lucy (2023) MRes_Creating a Rapid Identification Toolkit For Rural Ponds.pdf - Accepted Version
Available under License All Rights Reserved.
Download (6MB) | Preview

Abstract

It is well known that rural ponds are a key biological resource especially in rural landscapes, providing biologically diverse hotspots and delivering ecosystem services such as flood water storage, water quality improvement and food resource. Over the last century pond numbers have decreased significantly and, although numbers appear to be increasing more recently, the lack of rural ponds is cause for concern. Due to their high contribution to regional biodiversity ponds should be included in aquatic conservation at a landscape scale. However, mapping of pond location at a country scale, i.e. the whole of Great Britain, is inconsistent and numbers can often be unreliable due to the use of estimates and scaling from small spatial areas which have been surveyed. This variability in pond mapping methods is often due to the time- consuming nature of accurately mapping ponds; this can impact the direction of conservation efforts as without the base knowledge of gaps in pond networks key conservation areas cannot be easily identified. To attempt to decrease the time and effort spent on mapping rural ponds, a rapid technique to identify ponds using remote sensing and image classification methods is explored in this research. A 23 km2 study site in Somerset, England was selected for this study. Manual digitising of pond location using QGIS software and in-field ecological surveys of pond condition were undertaken to provide a dataset with which to verify the remote sensing methods. Land use data was also obtained for the study area to determine if there was any relationship between land use type and the identification of ponds by remote sensing. Images acquired from six different satellite sensors were obtained; these had varying spatial resolutions, ranging from Landsat 8 with 30m resolution to World View 3 at 1.24 m. Two classification methods were performed on all six images: (1) an automated technique using the Normalised Difference Water Index (NDWI) and, (2) supervised image classification using the Semi-automated Classification Plugin (SCP). The pond location for each method was compared to the manual pond count to give a percentage of ponds correctly identified. Pond ecological condition and how this may affect identifiability of ponds when using these methods is also explored by comparing the remote sensing outputs to the in-field ecological surveys. The highest accuracy of the NDWI outputs was 28.3% of ponds accurately identified. The supervised classifications were consistently more accurate with the highest identifying 72.4% of all ponds in the study area. A potential relationship between ecological condition of ponds and how likely they are to be identified was found; if ponds are of poor ecological condition or worse they are less likely to be classified as ponds, although this was only tested with a small sample. Examining the influence of land use, it was found that ponds in improved grassland are more likely to be identified consistently than in any other land use type. This project provides insight into methods for mapping ponds at landscape level scale. It has found that the highest ranked method, when considering accuracy, cost and temporal resolution, is performing supervised classification on World View 3 imagery. This method can be used to identify areas of a landscape which are lacking in rural ponds or ponds of good ecological quality, as this process should be scalable to any rural landscape which is captured via remote sensing at a spatial resolution of 1.24 m or less.

Item Type:

Thesis (Masters)

Thesis Advisors: