Spatio-temporal analyses of fruit fly populations in selected areas of the Western Cape

Date
2019-02
Journal Title
Journal ISSN
Volume Title
Publisher
Stellenbosch : Stellenbosch University
Abstract
ENGLISH ABSTRACT: The aim of this study was to investigate the spatio-temporal distribution of Ceratitis capitata (Wiedemann) (Diptera: Tephritidae) populations in heterogeneous fruit producing environments in Western Cape, South Africa, using geospatial analyses and machine learning (ML) techniques. A small scale study was conducted at orchard level on the Welgevallen experimental farm in Stellenbosch, South Africa, investigating the spatial patterns and associations of C. capitata and Ceratitis quilicii females, another important fruit fly pest in the Western Cape. The females of both species had aggregated spatial patterns, but their temporal patterns differed, with C. capitata aggregating significantly more towards the end of the season while C. quilicii aggregated significantly towards the beginning of the season. Ceratitis capitata and C. quilicii females were spatially associated, most prominently in home gardens, natural vegetation, citrus and nectarines. A geographical database was developed, incorporating existing area-wide trap monitoring data for C. capitata populations in the Elgin/Grabouw, Villiersdorp, Vyeboom (EGVV) region, Western Cape, an area currently under Sterile Insect Technique (SIT) management, was used to develop a geographical database with the aim to investigate the area-wide spatiotemporal distribution of C. capitata. There were no definitive spatial distribution pattern of C. capitata across all seasons. However, through visual analyses of spatial maps, a southeast/north-west split was observed where traps in the south-eastern parts of the study area showed higher catches and traps in the north-western parts showed lower catches. The results suggested a relationship between the geographic characteristics of EGVV and the abundance and distribution of C. capitata populations. The relationship between the geographic characteristics of the study area (EGVV) and the spatio-temporal distribution of C. capitata were further investigated using ML techniques. Monthly and seasonal long-term C. capitata spatio-temporal distributions were quantified into hot-and cold spots (HCSs), using spatial analyses tools. HCSs were then related to a set of geographic variables, using the random forest (RF) ML classification algorithm to determine the main drivers of the HCSs for C. capitata in the EGVV region. Spatial analyses showed that hot spots were concentrated in the hotter and drier areas, while cold spots were concentrated in the colder and wetter areas. The RF results indicated that rainfall was the most important driver of the HCSs in the EGVV region. To test the robustness of the RF algorithm for the purpose of explaining C. capitata HCSs in a heterogeneous fruit producing environment, the sample size and the variability in the geographic variables were increased by combining data from two regions: the EGVV and the Warmbokkeveld (WB), another fruit producing region under SIT. RF model accuracies from the combined dataset were not significantly lower than those of the individual regions. The drivers of C. capitata spatial distribution were different between regions, but distance to urban areas in the early fruiting season emerged as a strong driver in all scenarios. The findings showed that RF is a useful tool for investigating the spatio-temporal distribution of area-wide tephritid fruit fly trapping data, and that it can handle complex classification problems. It was evident from this study that the spatio-temporal distribution of C. capitata populations are driven by area-specific geographic variables. The area-specific RF models provided invaluable information, which could be used to improve the planning and implementation of area-wide C. capitata management programmes in heterogeneous agricultural landscapes. This study is relevant to the integrated management of fruit flies and potentially other insect pest species, on a local and regional scale. The framework which was developed will allow for the integration of a variety of data and the resultant analyses are relevant at an orchard and regional level. The information will assist efficient decision making by farmers and managers of area-wide integrated pest management programmes.
AFRIKAANSE OPSOMMING: Die doel van hierdie studie was om die tydruimtelike verspreiding van die Mediterreense vrugtevlieg populasies, Ceratitis capitata (Wiedemann) (Diptera: Tephritidae), te ondersoek in heterogene vrugteproduserende omgewings in die Wes-Kaap, Suid-Afrika, deur gebruik te maak van georuimtelike analises en masjienleer (ML) tegnieke. ’n Kleinskaalse studie is eers gedoen op boordvlak op die Welgevallen proefplaas in Stellenbosch, Suid-Afrika. Hierdie studie het beoog om die ruimtelike-patrone en - ooreenstemming van C. capitata en Ceratitis quilicii te ondersoek. Ceratitis quilicii is ook een van die vrugtevlieë in die Wes-Kaap wat ekonomies belangrik is. Resultate het getoon dat die wyfies van beide spesies versamelde ruimtelike-patrone vertoon, maar dat hulle tydspatrone verskil. Ceratitis capitata het beduidend aan die einde van die seisoen byeengekom, terwyl C. quilicii beduidend aan die begin van die seisoen byeengekom het. Beide spesies se ruimtelike-patrone het ooreengestem, maar die mees prominente ooreenstemming was in huistuine, natuurlike plantegroei, sitrus en nektariens. Na die bogenoemde studie is ’n geograpfiesedatabasis geskep met die beskikbare areawye lokvaldata vir C. capitata van die Elgin/Grabouw, Villiersdorp, Vyeboom (EGVV) area in die Wes-Kaap. Ceratitis. capitata word tans in EGVV bestuur deur gebruik te maak van die Steriele Insek Tegniek (SIT). Die doel van die databasis is om die area-wye tydruimtelike verspreiding van C. capitata te ondersoek. Resultate het getoon dat daar geen definitiewe ruimtelike verspreidingspatrone voorgekom het vir die tydperk van al die seisoene wat geanaliseer is nie. Nietemin het die visuele bestudering van kaarte getoon dat daar ‘n suidoos/noord-wes verdeling in lokval-tellings voorgekom het. Lokvalle met hoë tellings was meer verpreid in die suid-oostelike gedeeltes van die studie-area, terwyl die lokvalle met lae tellings meer verpreid was in die noord-westelike gedeeltes van die studie-area. Die resultate stel voor dat daar ’n verhouding bestaan tussen die geografiese karaktereienskappe van EGVV en die oorvloedige voorkoms en verspreiding van C. capitata-populasies. Hierna is die verhouding tussen die geografiese karaktereienskappe van EGVV en die tydruimtelike verspreiding van C. capitata verder ondersoek deur gebruik te maak van ML tegnieke. Die maandelikse en seisoenale langtermyn C. capitata verspreiding is gekwantifiseer in warm- en koue kolle (WKKe) deur gebruik te maak van ruimtelike analises. Daarna is ooreenkomstes gevind tussen die WKKe en ’n reeks geografiese veranderlikes deur gebruik te maak van die “random forest” (RF) ML klassifikasie algoritme, met die doel om die hoofdrywers van die WKKe vir C. capitata in die EGVV-area te bepaal. Die ruimtelike analises het gevind dat warm kolle in die warmer en droër areas gekonsentreerd was, terwyl die koue kolle meer in die kouer en natter areas gekonsentreerd was. Die RF se resultate het getoon dat reënval die belangrikste drywer, vir die WKKe in die EGVV-streek is. Om die robuustheid van die RF algoritme te toets met die doel om die C. capitata WKKe in heterogene vrugproduserende omgewings te verduidelik, is die steekproef vergroot en die variasie van die geografiese veranderlikes verhoog deur die data van twee streke te kombineer: die EGVV en die Warmbokkeveld (WB), nog ’n vrugproduserende area onder die SIT. Die resultate het getoon dat die RF model van die gekombineerde datastel se akkuraatheid nie beduidend verskil van die akkuraatheid van die individuele areas nie. Die drywers van die ruimtelike verspreiding van C. capitata het verskil tussen EGVV en WB, maar afstand vanaf stedelike areas gedurende die vroeë vrugteseisoen het as ’n sterk drywer in beide areas na vore gekom. Die bevindings het gewys dat RF ’n handige hulpbron is om die tydruimtelike verspreding van area-wye vrugtevlieg-lokval data te ondersoek en dat die RF algoritme komplekse klassifikasie probleme kan hanteer. Dit het duidelik getoon dat die tydruimtelike verspreiding van C. capitata-populasies gedryf word deur areaspesifieke geografiese faktore. Die area-spesifieke RF modelle het waardevolle inligting verskaf wat aangewend kan word om die beplanning en implimentering van area-wye C. capitata bestuursprogramme in heterogene landbou-landskappe te verbeter. Hierdie studie is relevant tot die geïntegreerde bestuur van vrugtevlieë asook ander potensiële insek-peste op ’n plaaslike- en streeksvlak. Die raamwerk wat ontwikkel is skep die platvorm om ’n verkeidenheid van data-soorte te integreer en te analiseer op boord- en streeksvlak. Boere en bestuurders van area-wye geïntegreerde pes-bestuursprogramme kan die ontginde inligting van die studie gebruik as hulpmiddel in hulle besluitnemingsprosesse.
Description
Thesis (PhDConsEcol)--Stellenbosch University, 2019.
Keywords
Geographic information systems, Fruit-flies -- Distribution -- South Africa -- Western Cape, Fruit -- Diseases and pests, UCTD, Geospatial data -- Analysis
Citation