Using geographic information systems and remote sensing technique to classify land cover types and predict grassland bird abundance and distribution in Nairobi National Park, Kenya
Frank Juma Ong'ondo , Shrinidhi Ambinakudige , Philista Adhiambo Malaki , Peter Njoroge , Hafez Ahmad
{"title":"Using geographic information systems and remote sensing technique to classify land cover types and predict grassland bird abundance and distribution in Nairobi National Park, Kenya","authors":"Frank Juma Ong'ondo , Shrinidhi Ambinakudige , Philista Adhiambo Malaki , Peter Njoroge , Hafez Ahmad","doi":"10.1016/j.ijgeop.2025.02.003","DOIUrl":null,"url":null,"abstract":"<div><div>Accurate and high-resolution mapping of land cover is essential for modeling species response, guiding habitat management practices, and prioritizing conservation efforts, especially in restricted and remote areas. Geographic information systems (GIS) and remote sensing (RS) techniques offer potential solutions. This study assessed the utility of GIS and RS techniques to map and predict grassland bird species in Nairobi National Park (NNP), Kenya. We utilized Sentinel-2B median imagery, which was accessible through Google Earth Engine (GEE), for January 2022 to classify six land cover classes: forest, shrubland, woodland, grassland, water, and bare soil. Grassland bird data were extracted from Kenya Bird Map (KBM) website for the period between 2015 and 2022, using full protocol card records. We hypothesized that grassland and shrubland would cover a larger portion of NNP and that grassland birds would respond positively to grassland, shrubland and woodland. We tested the second hypothesis using KBM data. Training samples for various land cover types were collected and used to train a Random Forest (RF) classifier on Sentinel-2B imagery. Model accuracy was evaluated with a confusion matrix, showing an overall accuracy of 99.93% and a Kappa statistic of 0.9989. Land cover composition indicated that grassland had the highest composition (44.9%), while water had the least (0.003%). Woodland, shrubland, forest and bare soil comprised 33.7%, 15.4%, 5.9%, and 0.2%, respectively. Logistic regression results showed that grassland birds responded positively to grassland and shrubland but tended to avoid woodland and bare soil. These findings demonstrate that land cover maps derived from GIS and RS techniques are fundamental tools for studying the abundance and distribution of grassland bird species, especially in remote areas. These tools are also essential for conservation and habitat management.</div></div>","PeriodicalId":36117,"journal":{"name":"International Journal of Geoheritage and Parks","volume":"13 1","pages":"Pages 92-101"},"PeriodicalIF":0.0000,"publicationDate":"2025-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"International Journal of Geoheritage and Parks","FirstCategoryId":"1085","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S2577444125000085","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"Social Sciences","Score":null,"Total":0}
引用次数: 0
Abstract
Accurate and high-resolution mapping of land cover is essential for modeling species response, guiding habitat management practices, and prioritizing conservation efforts, especially in restricted and remote areas. Geographic information systems (GIS) and remote sensing (RS) techniques offer potential solutions. This study assessed the utility of GIS and RS techniques to map and predict grassland bird species in Nairobi National Park (NNP), Kenya. We utilized Sentinel-2B median imagery, which was accessible through Google Earth Engine (GEE), for January 2022 to classify six land cover classes: forest, shrubland, woodland, grassland, water, and bare soil. Grassland bird data were extracted from Kenya Bird Map (KBM) website for the period between 2015 and 2022, using full protocol card records. We hypothesized that grassland and shrubland would cover a larger portion of NNP and that grassland birds would respond positively to grassland, shrubland and woodland. We tested the second hypothesis using KBM data. Training samples for various land cover types were collected and used to train a Random Forest (RF) classifier on Sentinel-2B imagery. Model accuracy was evaluated with a confusion matrix, showing an overall accuracy of 99.93% and a Kappa statistic of 0.9989. Land cover composition indicated that grassland had the highest composition (44.9%), while water had the least (0.003%). Woodland, shrubland, forest and bare soil comprised 33.7%, 15.4%, 5.9%, and 0.2%, respectively. Logistic regression results showed that grassland birds responded positively to grassland and shrubland but tended to avoid woodland and bare soil. These findings demonstrate that land cover maps derived from GIS and RS techniques are fundamental tools for studying the abundance and distribution of grassland bird species, especially in remote areas. These tools are also essential for conservation and habitat management.