Refine
Has Fulltext
- yes (5)
Is part of the Bibliography
- yes (5)
Year of publication
- 2023 (5) (remove)
Document Type
- Doctoral Thesis (5)
Language
- English (5)
Keywords
- Datenfusion (2)
- Ernteertrag (2)
- Landwirtschaft / Nachhaltigkeit (2)
- Winterweizen (2)
- Accurate crop monitoring (1)
- Baumhöhle (1)
- Bestäubung (1)
- Biene (1)
- Biene <Gattung> (1)
- Bienenkrankheiten (1)
Institute
Sonstige beteiligte Institutionen
Accurate crop monitoring in response to climate change at a regional or field scale
plays a significant role in developing agricultural policies, improving food security,
forecasting, and analysing global trade trends. Climate change is expected to
significantly impact agriculture, with shifts in temperature, precipitation patterns, and
extreme weather events negatively affecting crop yields, soil fertility, water availability,
biodiversity, and crop growing conditions. Remote sensing (RS) can provide valuable
information combined with crop growth models (CGMs) for yield assessment by
monitoring crop development, detecting crop changes, and assessing the impact of
climate change on crop yields. This dissertation aims to investigate the potential of RS
data on modelling long-term crop yields of winter wheat (WW) and oil seed rape (OSR)
for the Free State of Bavaria (70,550 km2
), Germany. The first chapter of the dissertation
describes the reasons favouring the importance of accurate crop yield predictions for
achieving sustainability in agriculture. Chapter second explores the accuracy
assessment of the synthetic RS data by fusing NDVIs of two high spatial resolution data
(high pair) (Landsat (30 m, 16-days; L) and Sentinel-2 (10 m, 5–6 days; S), with four low
spatial resolution data (low pair) (MOD13Q1 (250 m, 16-days), MCD43A4 (500 m, one
day), MOD09GQ (250 m, one-day), and MOD09Q1 (250 m, 8-days)) using the spatial
and temporal adaptive reflectance fusion model (STARFM), which fills regions' cloud
or shadow gaps without losing spatial information. The chapter finds that both L-MOD13Q1 (R2 = 0.62, RMSE = 0.11) and S-MOD13Q1 (R2 = 0.68, RMSE = 0.13) are more
suitable for agricultural monitoring than the other synthetic products fused. Chapter
third explores the ability of the synthetic spatiotemporal datasets (obtained in chapter
2) to accurately map and monitor crop yields of WW and OSR at a regional scale. The
chapter investigates and discusses the optimal spatial (10 m, 30 m, or 250 m), temporal
(8 or 16-day) and CGMs (World Food Studies (WOFOST), and the semi-empiric light
use efficiency approach (LUE)) for accurate crop yield estimations of both crop types.
Chapter third observes that the observations of high temporal resolution (8-day)
products of both S-MOD13Q1 and L-MOD13Q1 play a significant role in accurately
measuring the yield of WW and OSR. The chapter investigates that the simple light use
efficiency (LUE) model (R2 = 0.77 and relative RMSE (RRMSE) = 8.17%) that required fewer input parameters to simulate crop yield is highly accurate, reliable, and more
precise than the complex WOFOST model (R2 = 0.66 and RRMSE = 11.35%) with higher
input parameters. Chapter four researches the relationship of spatiotemporal fusion
modelling using STRAFM on crop yield prediction for WW and OSR using the LUE
model for Bavaria from 2001 to 2019. The chapter states the high positive correlation
coefficient (R) = 0.81 and R = 0.77 between the yearly R2 of synthetic accuracy and
modelled yield accuracy for WW and OSR from 2001 to 2019, respectively. The chapter
analyses the impact of climate variables on crop yield predictions by observing an
increase in R2
(0.79 (WW)/0.86 (OSR)) and a decrease in RMSE (4.51/2.57 dt/ha) when
the climate effect is included in the model. The fifth chapter suggests that the coupling
of the LUE model to the random forest (RF) model can further reduce the relative root
mean square error (RRMSE) from -8% (WW) and -1.6% (OSR) and increase the R2 by
14.3% (for both WW and OSR), compared to results just relying on LUE. The same
chapter concludes that satellite-based crop biomass, solar radiation, and temperature
are the most influential variables in the yield prediction of both crop types. Chapter six
attempts to discuss both pros and cons of RS technology while analysing the impact of
land use diversity on crop-modelled biomass of WW and OSR. The chapter finds that
the modelled biomass of both crops is positively impacted by land use diversity to the
radius of 450 (Shannon Diversity Index ~0.75) and 1050 m (~0.75), respectively. The
chapter also discusses the future implications by stating that including some dependent
factors (such as the management practices used, soil health, pest management, and
pollinators) could improve the relationship of RS-modelled crop yields with
biodiversity. Lastly, chapter seven discusses testing the scope of new sensors such as
unmanned aerial vehicles, hyperspectral sensors, or Sentinel-1 SAR in RS for achieving
accurate crop yield predictions for precision farming. In addition, the chapter highlights
the significance of artificial intelligence (AI) or deep learning (DL) in obtaining higher
crop yield accuracies.
Accurate crop monitoring in response to climate change at a regional or field scale plays a significant role in developing agricultural policies, improving food security, forecasting, and analysing global trade trends. Climate change is expected to significantly impact agriculture, with shifts in temperature, precipitation patterns, and extreme weather events negatively affecting crop yields, soil fertility, water availability, biodiversity, and crop growing conditions. Remote sensing (RS) can provide valuable information combined with crop growth models (CGMs) for yield assessment by monitoring crop development, detecting crop changes, and assessing the impact of climate change on crop yields. This dissertation aims to investigate the potential of RS data on modelling long-term crop yields of winter wheat (WW) and oil seed rape (OSR) for the Free State of Bavaria (70,550 km2), Germany. The first chapter of the dissertation describes the reasons favouring the importance of accurate crop yield predictions for achieving sustainability in agriculture. Chapter second explores the accuracy assessment of the synthetic RS data by fusing NDVIs of two high spatial resolution data (high pair) (Landsat (30 m, 16-days; L) and Sentinel-2 (10 m, 5–6 days; S), with four low spatial resolution data (low pair) (MOD13Q1 (250 m, 16-days), MCD43A4 (500 m, one day), MOD09GQ (250 m, one-day), and MOD09Q1 (250 m, 8-days)) using the spatial and temporal adaptive reflectance fusion model (STARFM), which fills regions' cloud or shadow gaps without losing spatial information. The chapter finds that both L-MOD13Q1 (R2 = 0.62, RMSE = 0.11) and S-MOD13Q1 (R2 = 0.68, RMSE = 0.13) are more suitable for agricultural monitoring than the other synthetic products fused. Chapter third explores the ability of the synthetic spatiotemporal datasets (obtained in chapter 2) to accurately map and monitor crop yields of WW and OSR at a regional scale. The chapter investigates and discusses the optimal spatial (10 m, 30 m, or 250 m), temporal (8 or 16-day) and CGMs (World Food Studies (WOFOST), and the semi-empiric light use efficiency approach (LUE)) for accurate crop yield estimations of both crop types. Chapter third observes that the observations of high temporal resolution (8-day) products of both S-MOD13Q1 and L-MOD13Q1 play a significant role in accurately measuring the yield of WW and OSR. The chapter investigates that the simple light use efficiency (LUE) model (R2 = 0.77 and relative RMSE (RRMSE) = 8.17%) that required fewer input parameters to simulate crop yield is highly accurate, reliable, and more precise than the complex WOFOST model (R2 = 0.66 and RRMSE = 11.35%) with higher input parameters. Chapter four researches the relationship of spatiotemporal fusion modelling using STRAFM on crop yield prediction for WW and OSR using the LUE model for Bavaria from 2001 to 2019. The chapter states the high positive correlation coefficient (R) = 0.81 and R = 0.77 between the yearly R2 of synthetic accuracy and modelled yield accuracy for WW and OSR from 2001 to 2019, respectively. The chapter analyses the impact of climate variables on crop yield predictions by observing an increase in R2 (0.79 (WW)/0.86 (OSR)) and a decrease in RMSE (4.51/2.57 dt/ha) when the climate effect is included in the model. The fifth chapter suggests that the coupling of the LUE model to the random forest (RF) model can further reduce the relative root mean square error (RRMSE) from -8% (WW) and -1.6% (OSR) and increase the R2 by 14.3% (for both WW and OSR), compared to results just relying on LUE. The same chapter concludes that satellite-based crop biomass, solar radiation, and temperature are the most influential variables in the yield prediction of both crop types. Chapter six attempts to discuss both pros and cons of RS technology while analysing the impact of land use diversity on crop-modelled biomass of WW and OSR. The chapter finds that the modelled biomass of both crops is positively impacted by land use diversity to the radius of 450 (Shannon Diversity Index ~0.75) and 1050 m (~0.75), respectively. The chapter also discusses the future implications by stating that including some dependent factors (such as the management practices used, soil health, pest management, and pollinators) could improve the relationship of RS-modelled crop yields with biodiversity. Lastly, chapter seven discusses testing the scope of new sensors such as unmanned aerial vehicles, hyperspectral sensors, or Sentinel-1 SAR in RS for achieving accurate crop yield predictions for precision farming. In addition, the chapter highlights the significance of artificial intelligence (AI) or deep learning (DL) in obtaining higher crop yield accuracies.
The western honeybee (Apis mellifera) is widely known as the honey producer and pollinator managed by beekeepers but neglected as a wild bee species. Central European honeybee populations have been anthropogenically disturbed since about 1850 through introgression and moderate artificial selection but have never been truly domesticated due to a lack of mating control. While their decline in the wild was historically attributed to the scarcity of nesting cavities, a contemporary view considers the invasion of the parasitic mite Varroa destructor in the 1970s as the major driver. However, there are no longitudinal population data available that could substantiate either claim. Based on the insight that introduced European honeybees form viable wild populations in eastern North America and reports on the occurrence of wild-living colonies from various European countries, we systematically studied the ecology of wild-living honeybees in Germany. First, we investigated whether wild-living honeybees colonising German forests form a self-sustaining population. Second, we asked how the parasite burden of wild-living colonies relates to that of managed colonies. And third, we explored whether the winter mortality of wild-living colonies is associated with parasite burden, nest depredation, or the lack of resources on the landscape scale.
Between 2017 and 2021, we monitored listed trees with black woodpecker cavities for honeybees in the managed forests of three study regions (Swabian Alb, counties Coburg and Lichtenfels, county Weilheim-Schongau). Continuity of occupation was determined using microsatellite genetic markers. Wild-living colonies predictably colonised forests in summer, when about 10% of all cavities were occupied. The annual colony survival rate and colony lifespan (based on N=112 colonies) were 10.6% and 0.6 years, with 90% of colonies surviving summer (July–September), 16% surviving winter (September–April), and 72% surviving spring (April–July). The average maximum and minimum colony densities were 0.23 (July) and 0.02 (April) colonies per km^2. During the (re-)colonisation of forests in spring, swarms preferred cavities that had already been occupied by other honeybee colonies. We estimate the net reproductive rate of the population to be R0= 0.318, meaning that it is currently not self-sustaining but maintained by the annual immigration of swarms from managed hives. The wild-living colonies are feral in a behavioural sense.
We compared the occurrence of 18 microparasites among feral colonies (N=64) and managed colonies (N=74) using qPCR. Samples were collected in four regions (the three regions mentioned above and the city of Munich) in July 2020; they consisted of 20 workers per colony captured at flight entrances. We distinguished five colony types representing differences in colony age and management histories. Besides strong regional variation, feral colonies consistently hosted fewer microparasite taxa (median: 5, range 1–8) than managed colonies (median: 6, range 4–9) and had different parasite communities. Microparasites that were notably less prevalent among feral colonies were Trypanosomatidae, Chronic bee paralysis virus, and Deformed wing viruses A and B. In the comparison of five colony types, parasite burden was lowest in newly founded feral colonies, intermediate in overwintered feral colonies and managed nucleus colonies, and highest in overwintered managed colonies and hived swarms. This suggests that the natural mode of colony reproduction by swarming, which creates pauses in brood production, and well-dispersed nests, which reduce horizontal transmission, explain the reduced parasite burden in feral compared to managed colonies.
To explore the roles of three potential drivers of feral colony winter mortality, we combined colony observations gathered during the monitoring study with data on colony-level parasite burden, observations and experiments on nest depredation, and landscape analyses. There was no evidence for an effect of summertime parasite burden on subsequent winter mortality: colonies that died (N=57) did not have a higher parasite burden than colonies that survived (N=10). Camera traps (N=15) installed on cavity trees revealed that honeybee nests are visited by a range of vertebrate species throughout the winter at rates of up to 10 visits per week. Four woodpecker species, great tits, and pine martens acted as true nest depredators. The winter survival rate of colonies whose nest entrances were protected by screens of wire mesh (N=32) was 50% higher than that of colonies with unmanipulated entrances (N=40). Analyses of land cover maps revealed that the landscapes surrounding surviving colonies (N=19) contained on average 6.4 percentage points more resource-rich cropland than landscapes surrounding dying colonies (N=94).
We estimate that tens of thousands of swarms escape from apiaries each year to occupy black woodpecker cavities and other hollow spaces in Germany and that feral colonies make up about 5% of the regional honeybee populations. They are unlikely to contribute disproportionately to the spread of bee diseases. Instead, by spatially complementing managed colonies, they contribute to the pollination of wild plants in forests. Honeybees occupying tree cavities likely have various effects on forest communities by acting as nest site competitors or prey, and by accumulating biomass in tree holes. Nest depredation (a consequence of a lack of well-protected nest sites) and food resource limitation seem to be more important than parasites in hampering feral colony survival. The outstanding question is how environmental and intrinsic factors interact in preventing population establishment. Nest boxes with movable frames could be used to better study the environmental drivers of feral colonies’ mortality. Pairs of wild (self-sustaining) and managed populations known to exist outside Europe could provide answers to whether modern apiculture creates honeybee populations maladapted to life in the wild. In Europe, large continuous forests might represent evolutionary refuges for wild honeybees.
The original habitat of native European honey bees (\(Apis\) \(mellifera\)) is forest, but currently there is a lack of data about the occurrence of wild honey bee populations in Europe. Prior to being kept by humans in hives, honey bees nested as wild species in hollow trees in temperate forests. However, in the 20th century, intensification of silviculture and agriculture with accompanying losses of nesting sites and depletion of food resources caused population declines in Europe. When the varroa mite (Varroa destructor), an invasive ectoparasite from Asia, was introduced in the late 1970s, wild honey bees were thought to be eradicated in Europe. Nevertheless, sporadic, mostly anecdotal, reports from ornithologists or forest ecologists indicated that honey bee colonies still occupy European forest areas. In my thesis I hypothesize that near-natural deciduous forests may provide sufficient large networks of nesting sites representing refugia for wild-living honey bees. Using two special search techniques, i.e. the tracking of flight routes of honey bee foragers (the “beelining” method) and the inspection of known cavity trees, I collected for the first time data on the occurrence and density of wild-living honey bees in forest areas in Germany (CHAPTER 3). I found wild-living honey bee colonies in the Hainich national park at low densities in two succeeding years. In another forest region, I checked known habitat trees containing black woodpecker cavities for occupation by wild-living honey bee colonies. It turned out that honey bees regularly use these cavities and occur in similar densities in both studied forest regions, independent of the applied detection method. Extrapolating these densities to all German forest areas, I estimate several thousand wild-living colonies in Germany that potentially interact in different ways with the forest environment. I conclude that honey bees regularly colonize forest areas in Germany and that networks of mapped woodpecker cavities offer unique possibilities to study the ecology of wild-living honey bees over several years.
While their population status is ambiguous and the density of colonies low, the fact that honey bees can still be found in forests poses questions about food supply in forest environments. Consequently, I investigated the suitability of woodlands as a honey bee foraging habitat (CHAPTER 4). As their native habitat, forests are assumed to provide important pollen and nectar sources for honey bee colonies. However, resource supply might be spatially and temporally restricted and landscape-scale studies in European forest regions are lacking. Therefore, I set up twelve honey bee colonies in observation hives at locations with varying degree of forest cover. Capitalizing on the unique communication behaviour, the waggle dance, I examined the foraging distances and habitat preferences of honey bees over almost an entire foraging season. Moreover, by connecting this decoded dance information with colony weight recordings, I could draw conclusions about the contribution of the different habitat types to honey yield. Foraging distances generally increased with the amount of forest in the surrounding landscape. Yet, forest cover did not have an effect on colony weight. Compared to expectations based on the proportions of different habitats in the surroundings, colonies foraged more frequently in cropland and grasslands than in deciduous and coniferous forests, especially in late summer when pollen foraging in the forest is most difficult. In contrast, colonies used forests for nectar/honeydew foraging in early summer during times of colony weight gain emphasizing forests as a temporarily significant source of carbohydrates. Importantly, my study shows that the ecological and economic value of managed forest as habitat for honey bees and other wild pollinators can be significantly increased by the continuous provision of floral resources, especially for pollen foraging.
The density of these wild-living honey bee colonies and their survival is driven by several factors that vary locally, making it crucial to compare results in different regions. Therefore, I investigated a wild-living honey bee population in Galicia in north-western Spain, where colonies were observed to reside in hollow electric poles (CHAPTER 5). The observed colony density only in these poles was almost twice as high as in German forest areas, suggesting generally more suitable resource conditions for the bees in Galicia. Based on morphometric analyses of their wing venation patterns, I assigned the colonies to the native evolutionary lineage (M-lineage) where the particularly threatened subspecies \(Apis\) \(mellifera\) \(iberiensis\) also belongs to. Averaged over two consecutive years, almost half of the colonies survived winter (23 out of 52). Interestingly, semi-natural areas both increased abundance and subsequent colony survival. Colonies surrounded by more semi-natural habitat (and therefore less intensive cropland) had an elevated overwintering probability, indicating that colonies need a certain amount of semi-natural habitat in the landscape to survive. Due to their ease of access these power poles in Galicia are, ideally suited to assess the population demography of wild-living Galician honey bee colonies through a long-term monitoring.
In a nutshell, my thesis indicates that honey bees in Europe always existed in the wild. I performed the first survey of wild-living bee density yet done in Germany and Spain. My thesis identifies the landscape as a major factor that compromises winter survival and reports the first data on overwintering rates of wild-living honey bees in Europe. Besides, I established methods to efficiently detect wild-living honey bees in different habitat. While colonies can be found all over Europe, their survival and viability depend on unpolluted, flower rich habitats. The protection of near-natural habitat and of nesting sites is of paramount importance for the conservation of wild-living honey bees in Europe.
Chapter I – Introduction
Global trade of beans of the cacao tree (Theobroma cacao), of which chocolate is produced, contributes to the livelihoods of millions of smallholder farmers. The understorey tree is native to South America but is nowadays cultivated in many tropical regions. In Peru, a South American country with a particularly high cacao diversity, it is common to find the tree cultivated alongside non-crop trees that provide shade, in so-called agroforestry systems. Because of the small scale and low management intensity of such systems, agroforestry is one of the most wildlife-friendly land-use types, harbouring the potential for species conservation. Studying wildlife-friendly land-use is of special importance for species conservation in biodiversity-rich tropical regions such as Peru, where agricultural expansion and intensification are threatening biodiversity. Moreover, there is a growing body of evidence that shows co-occurrence of high biodiversity levels and high yield in wildlife-friendly cacao farming. Yet studies are restricted to non-native cacao countries, and since patterns might be different among continents, it is important to improve knowledge on wildlife-friendly agroforestry in native countries.
Because studies of wildlife-friendly cultivation processes are still largely lacking for South America, we set out to study multiple aspects of cacao productivity in agroforests in Peru, part of cacao´s region of origin. The natural pollination process of cacao, which is critically understudied, was investigated by trapping flower visitors and studying pollen deposition from macrophotographs (Chapter II). Next, we excluded birds, bats, ants and flying insects and squirrels from cacao trees in a full-factorial field experiment and quantified these animals´ contribution to cacao fruit set, fruit loss and yield (Chapter III). Lastly, we aimed to assess whether fruit quantity and quality of native cacao increases through manually supplementing pollen (Chapter II and IV), and whether microclimatic conditions and the genetic background of the studied varieties limit fruit set (Chapter IV).
Chapter II – Cacao flower visitation: Low pollen deposition, low fruit set and dominance of herbivores
Given the importance of cacao pollination for the global chocolate production, it is remarkable that fruit set limitations are still understudied. Knowledge on flower visitation and the effect of landscape context and local management are lacking, especially in the crop’s region of origin. Moreover, the role of pollen deposition in limiting fruit set as well as the benefits of hand pollination in native cacao are unknown. In this chapter, we aimed to close the current knowledge gaps on cacao pollination biology and sampled flower visitors in 20 Peruvian agroforests with native cacao, along gradients of shade cover and forest distance. We also assessed pollen quantities and compared fruit set between manually and naturally pollinated flowers. We found that herbivores were the most abundant flower visitors in both northern and southern Peru, but we could not conclude which insects are effective cacao pollinators. Fruit set was remarkably low (2%) but improved to 7% due to pollen supplementation. Other factors such as a lack of effective pollinators, genetic pollen incompatibility or resource unavailability could be causing fruit set limitations. We conclude that revealing those causes and the effective pollinators of cacao will be key to improve pollination services in cacao.
Chapter III – Quantifying services and disservices provided by insects and vertebrates in cacao agroforestry landscapes
Pollination and pest control, two ecosystem services that support cacao yield, are provided by insects and vertebrates. However, animals also generate disservices, and their combined contribution is still unclear. Therefore, we excluded flying insects, ants, birds and bats, and as a side effect also squirrels from cacao trees and we assessed fruit set, fruit loss and final yield. Local management and landscape context can influence animal occurrence in cacao agroforestry landscapes; therefore, shade cover and forest distance were included in the analyses. Flying insects benefitted cacao fruit set, with largest gains in agroforests with intermediate shade cover. Birds and bats were also associated with improved fruit set rates and with a 114% increase in yield, potentially due to pest control services provided by these animals. The role of ants was complicated: these insects had a positive effect on yield, but only close to forest. We also evidenced disservices generated by ants and squirrels, causing 7% and 10% of harvest loss, respectively. Even though the benefits provided by animals outweighed the disservices, trade-offs between services and disservices still should be integrated in cacao agroforestry management.
Chapter IV – Cross-pollination improves fruit set and yield quality of Peruvian native cacao
Because yields of the cacao tree are restricted by pollination, hand pollination has been proposed to improve yield quantity and potentially, also quality. However, low self- and cross-compatibility of native cacao, and abiotic conditions could cancel out hand pollination benefits. Yet, the impact of genetic constraints and abiotic conditions on fruit set have not been assessed in native cacao so far. To increase our understanding of the factors that limit fruit set in native cacao, we compared manual self- and cross-pollination with five native genotypes selected for their sensorial quality and simultaneously tested for effects of soil water content, temperature, and relative air humidity. We also compared quality traits between manually and naturally pollinated fruits. Success rates of self-pollination were low (0.5%), but increased three- to eightfold due to cross-pollination, depending on the genotype of the pollen donor. Fruit set was also affected by the interaction between relative air humidity and temperature, and we found heavier and more premium seeds in fruits resulting from manual than natural pollination. Together, these findings show that reproductive traits of native cacao are constrained by genetic compatibility and abiotic conditions. We argue that because of the high costs of hand pollination, natural cross-pollination with native pollen donors should be promoted so that quality improvements can result in optimal economic gains for smallholder farmers.
Chapter V – Discussion
In this thesis, we demonstrated that the presence of flying insects, ants and vertebrates, local and landscape management practices, and pollen supplementation interactively affected cacao yield, at different stages of the development from flower to fruit. First, we showed that fruit set improved by intermediate shade levels and flower visitation by flying insects. Because the effective cacao pollinators remain unknown, we recommend shade cover management to safeguard fruit set rates. The importance of integrating trade-offs in wildlife-friendly management was highlighted by lower harvest losses due to ants and squirrels than the yield benefits provided by birds and bats. The maintenance of forest in the landscape might further promote occurrence of beneficial animals, because in proximity to forest, ants were positively associated with cacao yields. Therefore, an integrated wildlife-friendly farming approach in which shade cover is managed and forest is maintained or restored to optimize ecosystem service provision, while minimizing fruit loss, might benefit yields of native cacao. Finally, manual cross-pollination with native genotypes could be recommended, due to improved yield quantity and quality. However, large costs associated with hand pollination might cancel out these benefits. Instead, we argue that in an integrated management, natural cross-pollination should be promoted by employing compatible genotypes in order to improve yield quantity and quality of native cacao.