Skip to main content
eScholarship
Open Access Publications from the University of California

Scholarly Works (20 results)

  • Thesis
  • Peer Reviewed
UC Davis Electronic Theses and Dissertations (2022)

Road facilitates invasive plant species' early establishment and spread, which can impact ecosystem services and cause economic loss. Monitoring invasive species populations along the road is important in roadside integrated vegetation management (IVM). A roadside weedy and invasive species map can assist in developing species population models and designing proper weed management strategies. However, the traditional road survey requires unrealistic human labor and time to map invasive species at large scales. A novel weed mapping system was developed to retrieve species location data by integrating Google Street View (GSV) imagery and object detection based on deep learning algorithms. The target species of this feasibility study was johnsongrass (Sorghum halepense). We trained the detection network, You Only Look Once (YOLOv2), with 911 johnsongrass roadside images retrieved from Google Street View. YOLOv2 is a fast and accurate deep neural network. The trained detection model could detect johnsongrass in GSV images, and output bounding boxes with the target species' confidence scores. The trained model was then applied to a large image dataset of ~270,000 images along 135,000 km of roads in California, Oregon, Washington, and Nevada. The network detected 2,031 new johnsongrass records along roads in these four states, and the location of each image was used to create a map of the johnsongrass population. Our current deep learning model has 85% recall, 73.9% precision, and 77.5% overall accuracy on the testing dataset, which included 2,040 images. The model also has a 30% false positive rate (FPR). Work is in progress to reduce the FPR. Using our novel AI-based method, the estimated cost of the weed survey in four states is $3,570, while the traditional road surveys with cars at the same scale will cost at least $63,558 without considering associated risks, such as car accidents. Besides that, traditional road surveys require six months, but the automated weed survey only requires a few days with a trained detection network. The automated mapping scheme can apply to other weedy and invasive species, and it is possible to map this weed (and others) on a much larger scale, which is the focus of our future work.

    Cover page: Integrating Deep Learning and Google Street View for Novel Weed Mapping
    • Article
    • Peer Reviewed
    UC Davis Previously Published Works (2021)

    Main conclusion

    Phenological isolation can potentially reduce seed output and may be exploited as a novel tool for ecological management of dioecious weeds. Dioecious plants may benefit from a maximized outcrossing and optimal sex-specific resource allocation; however, this breeding system may also be exploited for weed management. Seed production in dioecious species is contingent upon the co-occurrence and co-flowering of the two genders and can be further disturbed by flowering asynchrony. We explored dimorphism in secondary sex characters in Amaranthus palmeri, and tested if reproductive synchrony can be affected by water stress. We have used seeds of A. palmeri from California, Kansas and Texas, and studied secondary sex characters under natural conditions and in response to water stress. Seeds of A. palmeri from California (CA) and Kansas (KS) were cordially provided by Dr. Anil Shrestha (California State University, Fresno, California) and Dr. Dallas E. Peterson (Kansas State University, Manhattan, Kansas), respectively. Seeds of a third population were collected from mature plants (about 30 plants) from a set-aside field in College Station, Texas. A. palmeri showed no sexual dimorphism with regard to the timing of emergence, plant height, and relative growth rate. While the initiation of flowering occurred earlier in males than females, females preceded males in timing of anthesis. Water stress delayed anthesis in males to a greater extent than females increasing the anthesis mismatch between the two sexes by seven days. Our data provide the first evidence of environment-controlled flowering asynchrony in A. palmeri. From a practical point of view, phenological isolation can potentially reduce seed output and may be exploited as a novel tool for ecological management of dioecious weeds.
      Cover page: Sex dimorphism in dioecious Palmer amaranth (Amaranthus palmeri) in response to water stress
      • Article
      • Peer Reviewed
      UC Davis Previously Published Works (2017)
      Empirical evidence suggests that co-flowering species can facilitate each other through shared pollinators. However, the extent to which one co-flowering species can relieve pollination limitation of another while simultaneously competing for abiotic resource has rarely been examined. Using a deterministic model we explored the demographic outcome for one ("focal") species of its co-occurrence with a species that shares pollinators and competes for both pollinator visitation and abiotic resources. In this paper we showed how the overall impact can be positive or negative, depending on the balance between enhanced fertilization versus increased competition. Our model could predict the density of co-flowering species that will maximize the pollination rate of the focal species by attracting pollinators. Because that density will also give rise to competitive effects, a lower density of co-flowering species is required for optimizing the trade-off between enhanced fertilization and competition so as to give the maximum possible facilitation of reproduction in the focal species. Results were qualitatively different when we considered attractiveness of the co-flowering species, as opposed to its density, because attractiveness, unlike density, had no effect on competition for abiotic resources. Whereas unattractive neighbours would not bring in pollinators, very attractive neighbours would captivate pollinators, not sharing them with the focal species. Thus optimal benefit to the focal species came at intermediate levels of attractiveness in the co-flowering species. This intermediate level of attractiveness in co-flowering species simultaneously maximized pollination and overall facilitation of reproduction for the focal species. The likelihood of facilitation was predicted to decline with the selfing rate of the focal species, revealing an indirect cost for an inbreeding mating system. Whether a co-flowering species can be facilitative depends on the way pollinators respond to the plant density: only a Type III functional response for visitation rate can result in facilitation. Our model provided both a conceptual framework and precise quantitative measures for determining the impacts of a neighbouring co-flowering species on reproduction.
        Cover page: How to be a good neighbour: Facilitation and competition between two co-flowering species
        • Article
        • Peer Reviewed
        UC Davis Previously Published Works (2018)
        Time-to-event methods have been proposed in the agricultural sciences, as one of the most suitable options for the analysis of seed germination data. In contrast to traditional linear/nonlinear regression, time-to-event methods can easily account for all statistical peculiarities inherited in germination assays, such as censoring, and they can produce unbiased estimates of model parameters and their standard errors. So far, these methods have only been used in combination with empirical models of germination, which are lacking biological underpinnings. We bridge the gap between statistical requirements and biological understanding by developing a general method that formulates biologically-oriented hydro time (HT), thermal time (TT) and hydrothermal time (HTT) models into a time-to-event framework. HT, TT, and HTT models are widely used for describing seed germination and emergence of plants as affected by the environmental temperature and/or water potential. Owing to their simplicity and the direct biological interpretation of model parameters, these models have become one of the most common tools for both predicting germination as well as understanding the physiology of germination responses to environmental factors. However, these models are usually fitted by using nonlinear regression and, therefore, they fall short of statistical rigor when inference about model parameters is of interest. In this study, we develop HT-to-event, TT-to-event and HTT-to-event models and provide a readily available implementation relying on the package “drc” in the R statistical environment. Examples of usage are also provided and we highlight the possible advantages of this procedure, such as efficiency and flexibility.
          Cover page: Hydrothermal-time-to-event models for seed germination
          • Article
          • Peer Reviewed
          UC Davis Previously Published Works (2017)
          Increasing population has posed insurmountable challenges to agriculture in the provision of future food security, particularly in the Middle East and North Africa (MENA) region where biophysical conditions are not well-suited for agriculture. Iran, as a major agricultural country in the MENA region, has long been in the quest for food self-sufficiency, however, the capability of its land and water resources to realize this goal is largely unknown. Using very high-resolution spatial data sets, we evaluated the capacity of Iran's land for sustainable crop production based on the soil properties, topography, and climate conditions. We classified Iran's land suitability for cropping as (million ha): very good 0.4% (0.6), good 2.2% (3.6), medium 7.9% (12.8), poor 11.4% (18.5), very poor 6.3% (10.2), unsuitable 60.0% (97.4), and excluded areas 11.9% (19.3). In addition to overarching limitations caused by low precipitation, low soil organic carbon, steep slope, and high soil sodium content were the predominant soil and terrain factors limiting the agricultural land suitability in Iran. About 50% of the Iran's existing croplands are located in low-quality lands, representing an unsustainable practice. There is little room for cropland expansion to increase production but redistribution of cropland to more suitable areas may improve sustainability and reduce pressure on water resources, land, and ecosystem in Iran.
            Cover page: Iran’s Land Suitability for Agriculture
            • Article
            • Peer Reviewed
            UC Davis Previously Published Works (2018)
            [This corrects the article DOI: 10.1371/journal.pone.0195511.].
              Cover page: Correction: Surveying the spatial distribution of feral sorghum (Sorghum bicolor L.) and its sympatry with johnsongrass (S. halepense) in South Texas
              • Article
              • Peer Reviewed
              UC Davis Previously Published Works (2022)

              Background

              Optical sensing solutions are being developed and adopted to classify a wide range of biological objects, including crop seeds. Performance assessment of optical classification models remains both a priority and a challenge.

              Methods

              As training data, we acquired hyperspectral imaging data from 3646 individual tomato seeds (germination yes/no) from two tomato varieties. We performed three experimental data manipulations: (1) Object assignment error: effect of individual object in the training data being assigned to the wrong class. (2) Spectral repeatability: effect of introducing known ranges (0-10%) of stochastic noise to individual reflectance values. (3) Size of training data set: effect of reducing numbers of observations in training data. Effects of each of these experimental data manipulations were characterized and quantified based on classifications with two functions [linear discriminant analysis (LDA) and support vector machine (SVM)].

              Results

              For both classification functions, accuracy decreased linearly in response to introduction of object assignment error and to experimental reduction of spectral repeatability. We also demonstrated that experimental reduction of training data by 20% had negligible effect on classification accuracy. LDA and SVM classification algorithms were applied to independent validation seed samples. LDA-based classifications predicted seed germination with RMSE = 10.56 (variety 1) and 26.15 (variety 2), and SVM-based classifications predicted seed germination with RMSE = 10.44 (variety 1) and 12.58 (variety 2).

              Conclusion

              We believe this study represents the first, in which optical seed classification included both a thorough performance evaluation of two separate classification functions based on experimental data manipulations, and application of classification models to validation seed samples not included in training data. Proposed experimental data manipulations are discussed in broader contexts and general relevance, and they are suggested as methods for in-depth performance assessments of optical classification models.
                Cover page: Experimental data manipulations to assess performance of hyperspectral classification models of crop seeds and other objects
                • Article
                • Peer Reviewed
                UC Davis Previously Published Works (2016)
                Weed seeds may maintain their viability when passing through the digestive tract of cattle and can be therefore dispersed by animal movement or the application of manure. Whether different cattle types of the same species can cause differential weed seed fate is largely unknown to us particularly under non-grazed systems similar to Holstein-Friesian dairy farming. We investigated the effect on the seed survival of four weed species in the digestive tracts of four groups of Holstein cattle: lactating cows, feedlot male calves, dry cows and growing heifers. The weed species used were Cuscuta campestris, Polygonum aviculare, Rumex crispus and Sorghum halepense. Cattle excretion was sampled for recovery and viability of seeds at four 24 hourly intervals after seed intake. The highest seed recovery occurred two days after seed intake in all cattle groups. Averaged over weed species, dry and lactating cows had the lowest and highest seed recovery of 36.4% and 74.4% respectively. No significant differences were observed in seed recovery of the four weed species when their seeds were fed to dry cows. Based on a power model fitted to seed viability data, the estimated time to 50% viability loss after seed intake, over all cattle groups ranged from 65 h (R. crispus) to 76 h (P. aviculare). Recovered seeds from the dung of feedlot male calves showed the highest mortality among cattle groups. Significant correlation was found between seed viability and ruminal pH (r = 0.86; P<0.05). This study shows that management programs aiming to minimize weed infestation caused by livestock should account for the variation amongst cattle groups in seed persistence. Our findings can be used as a guideline for evaluating the potential risk of the spread of weeds via the application of cattle manure.
                  Cover page: Variation in Weed Seed Fate Fed to Different Holstein Cattle Groups
                  • Article
                  • Peer Reviewed
                  UC Davis Previously Published Works (2017)
                  The level of resistance and patterns of cross-resistance to clodinafop, sethoxydim, and pinoxaden were examined in 12 putative resistant and one susceptible populations of winter wild oat (Avena ludoviciana) collected from Fars Province, in the southwest of Iran. The responses of biomass and length of coleoptiles to the increasing dosages of the three herbicides were determined in both whole-plant and seed bioassays. In the whole-plant bioassay, all 12 putative resistant populations were found to be resistant to clodinafop with resistance ratios (R/S) ranging from 1.76 to >47.04. Most clodinafop-resistant populations exhibited low levels of cross-resistance to sethoxydim. Three highly sethoxydim-resistant populations, F2, S2, and ES4, were slightly resistant to clodinafop. Six populations (M1, M2, F2, S2, S4, and ES4) showed high cross-resistance to pinoxaden with R/S values as large as 10.73 to 40.29. A highly clodinafop-resistant population, M2, was more sensitive to pinoxaden than the susceptible population. The results of the seed bioassay resembled those obtained from the whole-plant experiment suggesting seed bioassay as an inexpensive, rapid method for screening-resistant genotypes.
                    Cover page: Cross-resistance patterns of winter wild oat (Avena ludoviciana) populations to ACCase inhibitor herbicides
                    • Article
                    • Peer Reviewed
                    UC Davis Previously Published Works (2018)
                    Sorghum (Sorghum bicolor) is an important grain and forage crop grown across the US. In some areas, sorghum can become feral along roadsides and other ruderal areas, as a result of seed spill during harvest or transport. In some of these situations, feral sorghum grows in or near established johnsongrass (S. halepense) populations. Johnsongrass, a wild relative of sorghum and an incredibly noxious weed, is capable of hybridizing with cultivated sorghum. Because commercial hybrid sorghum cultivars are produced with cytoplasmic male sterility, progeny of the hybrid crop which compose the founder feral populations also segregate for male sterility. Consequently, male sterility in feral sorghum may increase the risk of outcrossing with johnsongrass. Using field surveys and spatial modelling, the present study aimed at documenting the occurrence of feral sorghum and understanding the anthropogenic and environmental factors that influence its distribution. Further, this research documented the sympatry of feral sorghum and johnsongrass in the roadside habitat. A total of 2077 sites were visited during a systematic field survey conducted in fall 2014 in South Texas. Feral sorghum and johnsongrass were found in 360 and 939 sites, while the species co-existed at 48 sites (2.3% of all surveyed sites). The binary logistic analysis showed a significant association between the presence of feral sorghum and road type, road body-type, micro-topography of the sampling site, nearby land use, and the presence of johnsongrass, but no association with the distance to the nearest grain sorting facility. The probability of finding feral sorghum away from johnsongrass patches was generally higher than finding them co-occur in the same location. A probability map for spatial distribution of feral sorghum was developed using the nearby land use type and the regional habitat suitability for johnsongrass as two key predictors. Overall, results show that feral sorghum and johnsongrass co-occur at low frequencies in the roadside habitats of South Texas, but these low levels still present a significant opportunity for hybridization between the two species outside of cultivated fields.
                      Cover page: Surveying the spatial distribution of feral sorghum (Sorghum bicolor L.) and its sympatry with johnsongrass (S. halepense) in South Texas
                      Top