Earth observation methods for sustainable Karoo Rangeland Management
| dc.contributor.advisor | Van Niekerk, Adriaan | en_ZA |
| dc.contributor.author | Harmse, Christiaan Johannes | en_ZA |
| dc.contributor.other | Stellenbosch University. Faculty of Arts and Social Sciences. Dept. of Geography and Environmental Studies. | en_ZA |
| dc.date.accessioned | 2024-02-19T07:47:23Z | |
| dc.date.accessioned | 2024-04-26T23:26:29Z | |
| dc.date.available | 2024-02-19T07:47:23Z | |
| dc.date.available | 2024-04-26T23:26:29Z | |
| dc.date.issued | 2023-12 | |
| dc.description | Thesis (DPhil)--Stellenbosch University, 2024. | en_ZA |
| dc.description.abstract | ENGLISH ABSTRACT: Rangelands, which comprise 25% of the earth's land surface, are under severe pressure due to the increasing global environmental problem of rangeland degradation. Ecological rangeland studies aim to determine the condition and productivity of rangelands and the severity of their degradation. In situ assessments are considered the most accurate way of monitoring rangeland degradation, but they are expensive and time-consuming, particularly when carried out over large areas. The Nama-Karoo biome in Southern Africa is primarily used for sheep and goat farming and is at risk of being overgrazed. Rangeland monitoring aims to determine whether grazing management strategies meet the goals of sustainable resource utilisation. Three experiments involving the combination of Earth observation technologies for rangeland monitoring were carried out in this research. First, a hypothesis that sheep graze more selectively under low stocking rates – potentially resulting in localised overgrazing – was tested. Livestock tracking, in situ observations, and Sentinel-2 imagery were used to make rangeland-scale observations of sheep grazing behaviour and vegetation conditions in the Nama-Karoo. The results showed that livestock congregates along drainage lines with deeper soil depth. There was a clear difference in the use of grazing areas among different stocking density classes. The conclusion was that spatial analyses of remotely sensed data can provide a landscape-scale overview of livestock movement patterns; and that high-resolution normalised difference vegetation index (NDVI) data can be used as a grazing management tool to determine the spatial variability of productive areas across the semiarid Upper Karoo rangelands and identify preferred grazing areas. Monitoring animal weight gain is expensive and often involves rounding up animals over large areas and long distances, leading to stress-related health problems and weight loss in animals. The second experiment evaluated remotely sensed vegetation indices for modelling sheep weight gain in semi-arid rangelands. The experiment also analysed the grazing behaviours in relation to time and location by using Sentinel-2 imagery and sheep movement data obtained from global positioning system (GPS) collar receivers. The results show that the average daily distance covered by sheep remained consistent throughout the year. The study successfully demonstrated the predictive capability of the NDVI in determining changes in the weight of sheep. The third experiment evaluated the effectiveness of multispectral (MS) and hyperspectral (HS) remotely sensed, unmanned aerial vehicle-(UAV)-based data and machine learning (random forest) methods to differentiate between 15 dominant Nama-Karoo plant species to aid ecological impact surveys. The results show that MS imagery was unsuitable as classification accuracies were generally low (37.5%). However, higher classification accuracies (>70.0%) were achieved when HS imagery was employed. Using in situ spectroscopic data collected with a fieldspectroradiometer, 12 key wavelengths were identified for discriminating among the dominant Karoo plant species with accuracies exceeding 90%. Reducing the dimensionality of the in situ spectroscopic dataset to the 12 key bands increased classification accuracies from 84.8% (all bands) to 91.7% (12 bands). Although classification accuracies were comparatively lower (76%) when HS remotely sensed imagery was used (instead of the in situ spectroscopic data), the results indicate that HS remote sensing imaging has the capability to effectively map indicator plant species in the Karoo region. The techniques developed in this research can be used to carry out satellite and UAV-based ecological assessments over large and inaccessible areas, assisting in managing the extensive Karoo rangelands more sustainably. | en_ZA |
| dc.description.abstract | AFRIKAANSE OPSOMMING: Weiveld wat 25% van die aarde se landoppervlak beslaan, is onder erge druk as gevolg van die toenemende wêreldwye omgewingsprobleem van agteruitgang van weivelde. Die doel van ekologiese weiveldstudies is om die toestand, produktiwiteit en mate van agteruitgang van weivelde te bepaal. Plaaslike assesserings word beskou as die mees akkurate manier om weiveldagteruitgang te monitor, maar dit is duur en tydrowend, veral wanneer dit oor groot gebiede uitgevoer word. Die Nama-Karoo-bioom in Suider Afrika word hoofsaaklik vir skaap- en bokboerdery gebruik, en loop die risiko om oorbewei te word. Weiveldmonitering beoog om vas te stel of weidingbestuurstrategieë aan die doelwitte vir volhoubare benutting van hulpbronne voldoen. Drie eksperimente wat die kombinasie van afstandwaarneming tegnologie vir weiveldmonitering behels, is in hierdie navorsing uitgevoer. Eerstens is 'n hipotese getoets dat skape meer selektief wei as die dieregetalle en weidingsdruk laag is wat moontlik tot gelokaliseerde oorbeweiding kan lei. Vee-opsporing, plaaslike waarnemings en Sentinel-2 satellietbeelde was gebruik om weiveld waarnemings van skaapweidingsgedrag en plantegroeitoestande in die Nama-Karoo te maak. Die resultate toon dat vee binne dreineringslyne met dieper gronde saamtrek. Daar was 'n duidelike verskil in die gebruik van weidingsgebiede onder verskillende vee beladings. Die gevolgtrekking is dat ruimtelike ontledings van afstandwaarnemings data 'n oorsig van veebewegingspatrone kan verskaf. Verder is gevind dat hoë-resolusie genormaliseerde verskil plantegroei-indeks data as 'n weidings bestuursinstrument gebruik kan word, veral om die veranderlikheid in produktiewe gebiede oor die semi-ariede Bo-Karoo weivelde te bepaal en voorkeur gebiede vir weiding te identifiseer. Die tweede eksperiment het afstandwaarneembare plantegroei-indekse vir die modellering van skaapgewigstoename in semi-ariede weivelde geëvalueer. Die monitering van gewigstoename van diere is duur en behels dikwels die aanjaag van diere oor groot gebiede en lang afstande wat dan stresverwante gesondheidsprobleme en gewigsverlies by diere kan veroorsaak. Die eksperiment het skape se weidingsgedrag in verhouding tot tyd en plek ontleed deur van Sentinel-2 beelde en skaap bewegingsdata, wat vanaf globale posisioneringstelsel (GPS) halsbande vir diere verkry is, gebruik te maak. Die resultate toon dat die gemiddelde daaglikse afstand wat deur skape afgelê is, konsekwent deur die jaar gebly het. Die studie het suksesvol die voorspellende vermoë van die genormaliseerde verskil plantegroei-indeks in die bepaling van veranderinge in die gewig van skape gedemonstreer. Die derde eksperiment het die doeltreffendheid van multispektrale (MS) en hiperspektrale (HS) afstandwaarmeembare, onbemande hommeltuig gebaseerde data en masjienleer (ewekansigewoud) metodes geëvalueer om tussen 15 dominante Nama-Karoo plantspesies te onderskei om ekologiese impakstudies te ondersteun. Die resultate toon dat MS-beelde nie geskik is nie, aangesien klassifikasie akkuraathede oor die algemeen laag was (37,5%). Hoër klassifikasie akkuraathede (>70,0%) was egter behaal toe HS-beelde gebruik was. Deur gebruik te maak van plaaslike spektroskopiese data wat met ‘n veld-spektroradiometer ingesamel is, is 12 sleutel golflengtes vir die onderskeiding tussen die dominante Karoo plantspesies geïdentifiseer en gebruik om klassifikasie akkuraathede van meer as 90% te behaal. Die vermindering van die dimensionaliteit van die spektroskopiese datastel na slegs 12 sleutelbande het klassifikasie akkuraathede van 84,8% (alle bande) na 91,7% (12 bande) verhoog. Alhoewel klassifikasie akkuraathede relatief laer (76%) was toe HS beelde (in plaas van die in situ spektroskopiese-data) gebruik is, dui die resultate daarop dat HS afstandwaarneming die vermoë het om indikator plantspesies in die Karoo-streek effektief te karteer. | af_ZA |
| dc.description.version | Doctoral | en_ZA |
| dc.format.extent | xix, 171 pages : illustrations | |
| dc.identifier.uri | https://scholar.sun.ac.za/handle/10019.1/130598 | |
| dc.language.iso | en | |
| dc.publisher | Stellenbosch : Stellenbosch University | en_ZA |
| dc.rights.holder | Stellenbosch University | en_ZA |
| dc.subject.lcsh | Rangelands -- Monitoring | en_ZA |
| dc.subject.lcsh | Remote-sensing images | en_ZA |
| dc.subject.lcsh | Sentinel-2 | en_ZA |
| dc.subject.lcsh | Hyperspectral imaging | en_ZA |
| dc.subject.lcsh | SLUSE model of natural resource management | en_ZA |
| dc.subject.lcsh | Multispectral imaging | en_ZA |
| dc.subject.lcsh | Range management | en_ZA |
| dc.subject.lcsh | Karoo (South Africa) | en_ZA |
| dc.subject.lcsh | Machine learning | |
| dc.subject.name | UCTD | |
| dc.title | Earth observation methods for sustainable Karoo Rangeland Management | en_ZA |
| dc.type | Thesis | en_ZA |
Files
Original bundle
1 - 1 of 1
Loading...
- Name:
- harmse_earth_2024.pdf
- Size:
- 4.65 MB
- Format:
- Adobe Portable Document Format
- Description: