California’s grasslands have been dominated by naturalized exotic forage grasses for 250– 300 years, but nonetheless serve as quintessential examples of “working landscapes” that both support rich biodiversity and provide essential rangeland feed for California’s livestock industry. However, the more recent invasion of two noxious weedy grasses—Elymus caput-medusae and Aegilops triuncialis—into the naturalized forage grass-dominated community has reduced biodiversity, and harmed forage production and livestock welfare. Thus, both weed species are targets for control by for conservation professionals and ranchers alike. Critical information needed to support adaptive management includes information about how to optimize livestock grazing for weed control and how to assess the impacts of treatments over large areas and longer time. Remote sensing has offered promise as a tool for grassland assessments, but it has proven particularly challenging to distinguish among groups of similar grasses and some methods—such as use of hyperspectral imagery —are currently beyond the reach of most managers. To meet the pressing need for information to evaluate management strategies, we thus developed and tested a relatively simple phenologically-based remote sensing approach that used aerial color infrared imagery to detect and monitor weed cover dynamics in this system. To our knowledge, pot growing systems this is the first robust and cost-effective method to distinguish annual weed grasses within an annual grass community.
There is much potential application for this approach because it uses a relatively simple multi-temporal classification approach that is not species-specific, aerial imagery as its backbone, which is easily understood by land managers, and a methodology that is easily applied across a large range of remote sensing platforms, including emerging UAS/drone technologies. Moreover, the mapping scale is fine enough to provide early detection of new noxious weed infestations, which is critical for early eradication efforts that can prevent landscape-scale spread of invasives .We developed this mapping approach in partnership with private ranch managers and conservation professionals to quantify landscape-level effects of management actions on invasion in California grasslands. Previously, we developed means of using remote sensing to monitor field-by-field forage production values in order to support rancher decision-making about grazing regimes and to assess the effectiveness of restoration strategies. This partnership allowed us to select technologies and to develop mapping approaches with land managers in mind. For the base data, we chose aerial CIR imagery because of its long history of acquisition and use, and because it was already familiar to ranch managers. To allow the most effective comparisons, we standardized the imagery by using the red and infrared bands to calculate values of the Normalized Difference Vegetation Index , a classic index of plant greenness that is a good predictor of chlorophyll content in grasslands. Digitized imagery provided aspatial resolution that was fine enough to detect weed patches as small as 1 meter, which is appropriate for the clump size of this vegetation and essential for detecting patch edges and for tracking patch size and shape over time.
Even more importantly, aerial imagery could be acquired within precise windows of time at specific phenological transitions when spectral differences between the weed and forage grass groups were greatest. Moderate spatial resolution imagery from satellite platforms such as Landsat may be available for little or no charge, but the grain size is too coarse to detect the patch dynamics of interest and the schedule for image acquisition is usually too inflexible and infrequent. Finer-scale satellite imagery is now available from numerous commercial companies, but at present is too costly for many ranch managers; future market changes may increase use of such resources by the rangeland management community. The CIR-derived NDVI imagery well captured the biology of our study site, including the overall reduction in greenness during the 2009 drought. Consistent with our findings in 2004 when we first prototyped this approach, weed-dominated patches showed a distinctive signature of directional NDVI change from spring to summer that contrasted with forage-dominated patches in which change had the opposite sign . Thus, the distribution of cover dominated by either or both of two weedy annual grasses—Elymus caput-medusae and Aegilops triuncialis—was best identified with maximum likelihood supervised multi-temporal classifications that took advantage of the combined NDVI characteristics of the weed group in March and in May . Most importantly for studies of vegetation change, this mapping approach worked well in both normal and dry years. In semi-arid regions like California, drought is a frequent challenge managers must contend with, so weed assessment tools need to be robust to dry conditions. During both normal and low rainfall periods, kappa statistic values describing the accuracy of this classification were good to excellent .
The mean producer’s accuracy over time indicated that the maps missed only ~7% of weed-dominated cover. The mean user’s accuracy was a little lower , reflecting the primary emphasis on detecting weed cover , but still strong. It indicated that about 14% of the area identified as weed-dominated on the maps might instead be equal parts weeds and forage, or forage-dominated. Lastly, a key strength of the resulting maps was the new view they provided of weed patch dynamics over large areas at a fine-grained spatial scale. The mapping approach described here distinguished notably more patch detail than previously achieved elsewhere with multitemporal analysis . In use, we found the vegetation classification maps to be so consistent with the landscape that when we loaded maps onto our GPS units and walked through the grasslands monitoring our positions in real-time, the fine-grained map details nearly always matched our field observations. As a tool, these maps offer means of quantifying the persistence of, or fine-grained change in, vegetation types from year-to-year, which is extremely difficult without the assistance of remote sensing methods. The general approach we outline has potential to be applied to weed mapping efforts in many grassland systems.Using the multi-temporal classification described, we were able to produce annual distribution maps of cover dominated by goat grass and medusa head or by forage at the 1-m scale, evaluate change in cover type over time, and quantify vegetation cover differences among management units. For field use, the size and shape of patches made of numerous pixels aggregated together can be readily evaluated by eye. Researchers needing more quantitative measures of patch dimensions or their vector outlines could use additional landscape analysis tools , or object-oriented approaches . At our study site, the maps revealed clear effects of management on weed distribution, with fence-line contrasts across identical soils highlighting management influence. In management units with little to no grazing, weeds dominated a greater percentage of the area than in grazed units, the proportion of area covered by persistent weeds was 1.6–2.8 times greater, and the proportion of total weed cover that was stable from year-to-year was 1.3–1.4 times larger. In contrast, planting racks the area of beneficial forage-dominated cover was greater and more persistent ingrazed units. Although livestock may also spread weeds, these observations are consistent with indications that sheep-grazing can help control medusa head in some cases and with rancher observations that removal of thick weed thatch by livestock consumption or trampling is important in breaking up weed monocultures. Similarly, Mariotte and colleagues found by manipulating thatch quantities that increased thatch enhances the growth and/or seed production of invasive species in California grasslands, while harming the performance of natives. While our investigation does not delve into the specifics of grazing regime impact on these invasive weeds, our findings demonstrate the promise of this mapping tool in quantifying responses to management actions. In ecological terms, this vegetation analysis highlights the extent to which weed monocultures can form and persist in the absence of disturbance. For weeds such as goatgrass, aggregation in patches can reduce the negative impact of interspecific competition and increase seed output per individual.Fine-scale maps of vegetation cover provide essential data both for land managers and for biologists seeking to understand the complexity of plant communities. It is increasingly recognized that to understand community dynamics it is necessary to assess patterns and controllers across multiple scales.
The mapping methods developed here for aerial imagery can be extended to digital imagers on unmanned airborne systems to increase the potential for rapid, inexpensive, and user-driven assessment of grassland condition. Ultimately, these remote sensing technologies will enable more data-driven optimization of grassland management, paralleling the application of precision agriculture in crop management. For example, GPS-tracking of animal movement could be combined with fine-scale patch data to gain more insight about how specific grazing regimes and range resources drive vegetation change. Alternatively, such landscape analyses could be used to evaluate broader state-and-transition or gradient models of land cover change. Finally, understanding of vegetation patch dynamics and configuration can illuminate a broad range of spatially-explicit ecological processes .Understanding the evolutionary genetics of adaptation to human-mediated practices like small and large-scale production agriculture is critical to address global challenges including the security of food, fuel, bio-product, and fiber production. A central challenge in agriculture is to harness the genetic variation controlling key traits in crops to produce stable populations that can be planted, managed, and harvested effectively. Evolutionary models frame and explain the domestication, continued improvement, and management of cultivated plants. Examining these processes sheds light on the roles of selection and demography on genetic interactions of populations and species during adaptation. During domestication and crop improvement, individuals are selected for predictable traits. The means and variances of these traits in breeding lines over generations depend upon the relative roles of genetics and the environment in shaping variation and the number of alleles at loci governing these phenotypes. Likewise,the additive genetic variance associated with a given domestication trait may control how easy it is to fix a population for a trait value, particularly for traits that are vastly different from wild or weedy close relatives. Domestication is a selection process for adaptation to agro-ecological niches favorable for human use, harvest, consumption, and management. Historical gene flow between wild progenitors and domesticated plant populations ensures that cultivated varieties vary in their composition of domestication versus wild traits. Domesticated lines and wild relatives that can interbreed are common among plants and animals. Genome-wide studies of these interbreeding complexes help us understand how genetic introgression modulates adaptation and the maintenance of species boundaries in the face of gene flow. Although weedy rice physiologically and phenotypically resembles cultivated rice, it differs in several important weedy traits, including seed shattering habit, seed dormancy, protracted emergence, and the presence of red pigmentation in the seed pericarp in many cases. Shattering furthers propagation of the weed because seeds scatter in the field before cultivated rice is harvested. Variation in gene sequence and expression has been shown in many genes related to seed shattering, including qSH1, sh4, and SHAT1. The shattering trait in weedy rice has been shown to re-evolve after fixation of the non-shattering sh4 allele in its domesticated ancestors. Additionally, QTL analysis indicates that shattering has reemerged independently and is controlled by different genetic locations in weedy rice. Variable seed dormancy makes control of weedy rice by crop rotation difficult due to the ability of weedy rice to remain dormant for extended periodsin the field. Protracted emergence patterns make control by chemical means difficult because late or early emerging individuals can escape herbicide applications. Prolonged and highly variable emergence also makes control by non-chemical means, such as cover crops, difficult. Finally, weedy rice is commonly referred to as red rice when characterized by a red-pigmented pericarp. Contamination of commercialrice with pigmented red rice seed significantly lowers its commercial value. Most traits distinguishing crop from weedy forms are determined by recessive alleles of major-effect loci. A subsequent focus on the molecular evolution of genes important in de-domestication can guide our understanding of the tempo and process of evolution in weedy and feralized crop populations, but first we must examine the evolutionary origins and morphologies that characterize emergent weed populations. This knowledge informs agricultural management strategies that account for how weeds evolve and mitigate infestation. Understanding the genetic interplay underlying these processes will predict their directionality, identify traits for crop improvement in the face of new or changing environmental constraints, and outline ecosystem management strategies for sustainability. Cultivated Asian rice and its progenitor O. rufipogon are both diploid AA genome species,which facilitates introgression and the maintenance of hybrid feral forms .