Remote sensing and spatial statistical techniques for modelling Ommatissus lybicus (Hemiptera: Tropiduchidae) habitat and population densities

Background

The date palm, Phoenix dactylifera Linnaeus, is an important economic resource in the Sultanate of Oman. Plant-parasitic nematodes, associated with date palm trees in Oman and in most other Arab countries, can reduce their economic yields (El-Juhany, 2010). A variety of insect pests can cause major damages to this crop through production losses and plant death (Abdullah, Lorca & Jansson, 2010; Al-Khatri, 2004; Blumberg, 2008; El-Shafie, 2012; Howard, 2001). One such species, Ommatissus lybicus de Bergevin 1930 (Hemiptera: Tropiduchidae), which is known more commonly as the Dubas bug (DB), has been identified as a major economic threat, and is presently affecting palm growth yield in Oman (Al-Yahyai, 2006). Indeed, the DB has been identified as one of the primary reasons for the decline in date production in Oman (Al-Yahyai & Al-Khanjari, 2008; Al-Zadjali, Abd-Allah & El-Haidari, 2006; Mamoon, Wright & Dobson, 2016). It is also a principal pest of date palms in many locations throughout the Middle East, East and North Africa, (Klein & Venezian, 1985; Mifsud et al., 2010). The DB is believed to have been introduced into the Tigris-Euphrates River Valley, from there it has spread to other zones in recent decades (Blumberg, 2008; El-Haidari, Mohammed & Daoud, 1968).

The DB is a sap feeding insect; both adults and nymphs suck the sap from date palms, thereby causing chlorosis (removal of photosynthetic cells and yellowing of fronds). Prolonged high infestation level will result in the flagging and destruction of palm plantations (Al-Khatri, 2004; Howard, 2001; Hussain, 1963; Mahmoudi et al., 2015; Mokhtar & Al Nabhani, 2010; Shah et al., 2013). There is also an indirect effect whereby honeydew secretions produced by the DB can promote the growth of black sooty mould on the foliage and consequently a reduction in the photosynthetic rates of date palms (Blumberg, 2008; Mokhtar & Al-Mjeini, 1999; Shah et al., 2012). Nymphs pass through five growth instars (Hussain, 1963; Shah et al., 2012), with adult female DB reaching 5–6 mm and the males 3–3.5 mm in length (Aldryhim, 2004; Mokhtar & Al Nabhani, 2010). Their colour is yellowish-green while the main distinguishing feature between males and females is the presence of spots on females; males have a more tapered abdomen and larger wings relative to the abdomen (Al-Azawi, 1986; Al-Mahmooli, Deadman & Al-Wahabi, 2005; Elwan & Al-Tamimi, 1999; Hussein & Ali, 1996; Jasim & Al-Zubaidy, 2010; Kaszab, Wittmer & Buttiker, 1979; Khalaf et al., 2012; Mokhtar & Al Nabhani, 2010; Thacker, Al-Mahmooli & Deadman, 2003).

Study area

The Sultanate of Oman, which covers an area of 309,500 km², extends from 16°40′N to 26°20′N, and 51°50′E to 59°40′E. It occupies the south-eastern corner of the Arabian Peninsula (Fig. 1). It has 3,165 km of coastline, extending from the Strait of Hormuz in the north to the border with the Republic of Yemen in the South. The coastline faces onto three different water bodies, namely the Arabian Sea, the Persian Gulf (also known as Arabian Gulf), and the Gulf of Oman.

To the west, Oman is bordered by the United Arab Emirates and the Kingdom of Saudi Arabia. Mountainous areas account for 15% of the land area, while desert plains and sandy areas cover 74%, agro-biodiversity areas cover 8%, and the coastal zone covers 3%, respectively (Luedeling & Buerkert, 2008). The location of Oman provides favourable conditions for agriculture, with land under agricultural use accounting for 8% of the territory and the economic output accounting for 14.6% of the GDP in 2008. According to the 2004–2005 soil survey conducted by the Ministry of Food and Agricultural (MFA), 22,230 km² (equivalent 2.223 million ha) is optimal for agricultural activity, which represents ∼7.5% of the country’s land area. Approximately 728.2 km² (∼72,820 ha) of the country is irrigated using the Falaj irrigation system, where local springs or wadis (streams) underflow areas are cultivated with palm trees, banana, limes, alfalfa, and vegetables (Gebauer et al., 2007).

Oman has an arid climate, receiving less than 100 mm of rainfall per year; however, the mountainous parts of the country receive higher precipitation levels (Kwarteng, Dorvlo & Vijaya Kumar, 2009). As the dependent variable, DB infestations occur where palm trees are concentrated; therefore, in this study we focused on northern Oman (26°50′N to 22°26′N, and 55°50′E to 59°50E) which experiences high infestations (Fig. 1) (Al-Kindi et al., 2017b).

Dubas bugs are active on leaflets, rachis, fruiting bunches, and spines during different stages of growth of date palm trees. These infestations are capable of causing up to 50% crop loss during a heavy infestation (Shah et al., 2013). Studies of insect pests of the date tree palm indicated more than 54 arthropods species insects connected with date plantations. Nevertheless, DB and red weevil (RPW) Rhynchophorus ferrugineus Oliver, and lesser moth, are considered major economically significant pests affecting growth and yield of date palm trees (Al-Zadjali, Abd-Allah & El-Haidari, 2006).

Biology and life history

The biology of this species has been investigated in a number of studies (Al-Azawi, 1986; Arbabtafti et al., 2014; Hussain, 1963; Jasim & Al-Zubaidy, 2010; Klein & Venezian, 1985; Payandeh & Dehghan, 2011; Shah et al., 2012). The DB produces two generations annually, including the spring and autumn generations (Blumberg, 2008; Hussain, 1963). In the spring generation, eggs start hatching from February to April, after which nymphs pass through five instars to become adults in approximately 6–7 weeks. The eggs aestivate during the hot season (i.e., summer) until the autumn generation, when they start hatching from late August to the last week of October. A nymph takes about 6 weeks to develop into an adult, which then lives for about 12 weeks. Females lay between 100 and 130 eggs (Elwan & Al-Tamimi, 1999; Mokhtar & Al Nabhani, 2010). The female DB lays her eggs by inserting them into holes in the tissue of the date palm frond at the end of each generation. The eggs remain dormant for about three months. When they hatch, the resulting nymphs remain on the fronds of the same tree, feeding on the sap, and defecating large amounts of honeydew, which eventually covers the palm fronds and promotes the growth of black sooty mould (Zamani, Aminaee & Khaniki, 2013).

In extreme cases, the sooty mould that develops from the honeydew secretions can block the stomata of the leaves, causing partial or complete suffocation of the date palm, which in turn reduces its yield. The honeydew secretion also makes the dates unpalatable (Aminaee, Zare & Assari, 2010; El-Juhany, 2010; Gassouma, 2004; Mamoon, Wright & Dobson, 2016). The eggs of DB are sensitive to temperature. In summer, the eggs can hatch within 18–21 days, but in winter they may take up to 170 days to hatch (Blumberg, 2008). The developmental time of DBs eggs has been studied at three different temperatures, 17.6, 27.5, and 32.4 °C in Oman (Al-Khatri, 2011). The results showed that a temperature of 27.5 °C appeared to be the optimal temperature for the biological activities of this species (Al-Khatri, 2011). At 35 °C, the biological processes of the pest are disrupted, thus leading to high mortality, particularly in females (Bagheri et al., 2016; Bedford et al., 2015).

Investigations into the population and the fluctuation in spatial distribution (Khalaf & Khudhair, 2015) of the two DB generations in the Bam region of Iran showed that this pest has an aggregated spatial distribution pattern (Payandeh, Kamali & Fathipour, 2010). Seasonal activities effected by climate change showed that nymphs were dynamic from April to May in the first generation and August to October in the second generation. In the first and second generations, the adults are active from May to June and from September to November, respectively. Earlier work (Blumberg, 2008) reported that temperature exposure below 0 °C adversely affects the ability of adults to survive. The DB avoids direct sunlight (Klein & Venezian, 1985; Shah et al., 2013), and it is spread via the transfer of infested offshoots as well as by wind or wind dust (Hassan, 2014; Jasim & Al-Zubaidy, 2010).

Biological control

Some research has also been conducted on the natural biological control of the DB, such as using predators and parasites. Early results showed a variety of natural predators that could be used as biological control agents, among these being Aprostocetus sp., Oligosita sp., and Runcinia sp. (Cammell & Knight, 1992). Furthermore, Hussain, 1963 reported that the eggs of the DB could be parasitised by a small Chalcidoid, while the nymphs and adults were preyed upon by the larvae of the lace wing (Chrysopa carnea Steph.). Hussain also stated that three adult species of Coccinellids prey on nymphs and adults of the DB. However, further study is needed to determine the distributions of these natural enemies in Oman and their effectiveness in controlling DB populations. Some measure of success was also achieved with pathogenic bacteria as microbiological control agents (Khudhair, Alrubeai & Khalaf, 2016), although their toxicological aspects require further research in order to assess the safety of their implementation at a large scale (Cannon, 1998).

Chemical control

Given the significant economic impact of this pest, research into effective management strategies demands high priority. Several insecticides have been evaluated for DB control in Oman since 1962 (Table 1) with Sumi-Alpha-5 EC being effective as a ground spray and Karate® 2 ULV, Trebon® 30 ULV, and Sumicombi 50® ULV achieving some measure of success as aerial sprays. Karate-Zeon® was also found to be very effective as it gave 100% reduction in numbers of DB instars and adults, between three and seven days after application. However, the use of this particular pesticide is restricted due to its side effects such as irritation (Al-Yahyai & Khan, 2015). In Israel, systemic carbamates such as aldicarb and butocarboxim have been successful, while in Iraq dichlorvos (DDVP) injected directly into the infected palms were also successful in suppressing the pest population (Blumberg, 2008). Nonetheless, this method of control is expensive with negative environmental impacts on non-target species (particularly natural enemies of DB) as well as on human health.

Research has shown that some pesticide residues persist in the fruit up to 60 days after application (Al-Samarrie & Akela, 2011). In addition, chemical control has achieved limited successes and DB continues to pose a major challenge to Omani agriculture. More information about the biological and chemical control and their impacts can be found in literature (Shifley et al., 2014; Thacker, Al-Mahmooli & Deadman, 2003).

Research opportunities

A number of opportunities exist for research into the biology and ecology of this species in order to gain a thorough understanding of its life cycle and its distribution. The climatic factors that influence its survival and distribution also merit study because such information may be useful in determining current and future potential distributions, particularly in light of climate change.

In a review of the effects of climate change on pest populations, an earlier report (Cammell & Knight, 1992) suggested that increases in mean global temperatures, as well as changes in rainfall and wind patterns, could have profound impacts on the population dynamics, abundance, and distribution of insect pests of agricultural crops. More recent research has supported this finding (Bale et al., 2002; Cannon, 1998; Cook, 2008; Shifley et al., 2014; Tobin, Parry & Aukema, 2014). A key issue in ecology and conservation is the mapping of pest species distributions as this information can be used to devise more effective management strategies for their control.

Mapping of DB infestations is important for developing predictive models that give the probability of occurrence, spatial distributions and densities under different environmental, meteorological, anthropogenic and resource availability conditions. Maps such as the DB hazard map can be used to devise an integrated palm management (IPM) plan, thus enhancing capacity and educating farmers on how to apply IPM for the control of this pest.

Mapping DBs are also beneficial in terms of better planning of date palm orchard locations, better design and management of farms, what cultivars to plant, distance between palms, irrigations, pesticides, fertilisations, etc. (Bouyer et al., 2010). There will also be savings in the cost of monitoring since RS based techniques developed as part of this study can provide a more efficient and cost-effective means for large scale monitoring of infestations and observation of stress levels on date palm trees.

The aim of this review is to highlight technological advances in the fields of RS (i.e., by aircraft or a satellite platform) and spatial statistical techniques that can be used to significantly enhance our ability to detect and characterise physical and biological stresses on several plant species. In particular, advanced RS and spatial statistical techniques need to be developed and implemented for the surveillance and control of DB adults and nymphs over large spatial scales. In turn, this will greatly assist plant protection service (PPS) projects, integrated pest management technology (IPMT) programs and farmers in protecting their palm tree orchards by adopting timely preventative actions.

Data requirements for crop management

It is important to collect data regarding crops and soil and to identify the changes that occur in the field to achieve precise crop management in the agricultural sector. Pinter et al. (2003) Data are needed on the conditions that are stable across seasons (e.g., crop type, soil fertility), differing during the seasons (e.g., pest attacks, water quality and quantity, nutrient contents, moisture, temperature), and on factors that contribute to crop yield variability (Hall, Skakun & Arsenault, 2006; Jadhav & Patil, 2014). This data is valuable for determining the unique phenological cycles of agricultural crops in different geographic regions (Jensen, 2000; Abdullah & Umer, 2004; Acharya & Thapa, 2015).

A good example of this are date palms. Typically, date palm trees are 7–10 m tall with crowns 2–4 m in diameter, and the trees are normally spaced 3–5 m apart. The understory of date palm plantations might include banana palms, mango trees, acacia bushes, vegetable crops, grain crops, forage crops. The reflectance characteristics of a date palm area are often driven by the density and health of the understory vegetation (Harris, 2003). It can be difficult to use small pixel data to study date palm areas with little or no understory vegetation because the small pixel effects may make it difficult to identify infestations (e.g., where date palms are infested between mountains and dry rivers) given the tree spacing and density of leaves and branches. Studies like Hussain (1963) and Mahmoudi et al. (2015) have revealed that heavy infestations occur mostly along valleys. Additionally, the characteristics of the understory vegetation may dominate the contribution of spectral responses rather than the tree vegetation themselves.

Optical remote sensing data

The vital feature of RS is the detection of radiant energy emitted by various objects. The energy detected might be in the form of acoustic energy (sound) or electromagnetic energy (visible light, infrared heat, ultraviolet, and microwaves). Remote sensing technology deployed from ground, air, or space-based platforms is capable of providing detailed spectral, spatial, and temporal information on vegetation health and is particularly useful for crop yield estimation applications (Justice et al., 2002).

Radar data

For many years, airborne technology has been employed in agricultural operations. Nevertheless, space-borne synthetic aperture radar (SAR) technology such as those of the Advanced Land Observing satellite; TerraSAR-X and Phased Array L-band have become available since the 2000s (Ortiz, Breidenbach & Kändler, 2013). Multiple radar sensors can work autonomously to detect solar radiation variation, but dissimilar optical sensors from which spectral reflectance measurements are taken are affected differently by variation in the solar emission. Radar technology has found limited applications in regional studies because of its high costs, the narrow swath widths, and limited extent of coverage (Feng et al., 2003).

The data can be extracted routinely by using the existing network of weather radars, and it can be used to alert growers that local crops are at heightened risk (Westbrook & Isard, 1999; Drake, 2002). Such information can then be used for fine tuning pest management practices such as pesticide applications, and could potentially reduce pesticide use by nearly 50% and lessen the overall impact of toxic chemicals on the environment (Dupont, Campanella & Seal, 2000), as well as on the natural enemies of these insect pests. Table 2 shows example applications of different remote sensing technologies used to detect change in vegetation.

Spectroscopic analysis

Fluorescence spectroscopy (FS) is a type of spectroscopic method by which fluorescence is measured of an object of interest following excitation by rays of light. Fluorescence has been used for vegetation research to monitor stress levels and physiological states in plants. There are two types of fluorescence. The first is blue-green fluorescence in the ∼400–600 nm range and the second type is chlorophyll fluorescence in the ∼650–800 nm range. Fluorescence spectroscopy can be used to monitor nutrient deficiencies, environmental conditions based on stress levels, infestations, and plant diseases. In fact, it can be used to monitor fruit quality, photosynthetic activity, tissue stress, and infestations in many types of crops (Karoui & Blecker, 2011; Tremblay, Wang & Cerovic, 2012).

Remote Sensing is a powerful technique for visualising, diagnosing, and quantifying plant responses to stress like temperature, drought, salinity, flooding, and mineral toxicity. Approaches can range from the use of simple combinations of thermal and reflectance sensor data to visible reflectance and fluorescence data. In particular, combined fluorescence reflectance and thermal imaging sensor data can be used for quick investigations of vegetation stress (Lenk et al., 2007).

Solar radiation and the humid-thermal ratio

Biological systems are highly dependent on two most important climatic factors, namely temperature and precipitation. Temperature is influenced by solar radiation and thermal emissions, while precipitation controls the dry or wet conditions (humidity) associated with plant growth. These factors are especially important in regions where extreme temperatures and humidity conditions are prevalent and large fluctuations exist throughout the seasons as such conditions can predispose plants to insect pests and diseases. In this regard, solar radiation models can be used to investigate insect infestations. Solar radiation models can be applied to calculate the potential solar radiation at a chosen location over a 12-month period.

The results from solar radiation studies can then be used to find correlations with different infestation levels to examine if solar radiation plays a determinant role in different infestation levels (see Fig. 2). Solar radiation can also be used to study the presence/absence and density of animals, plants diseases and infestations such as those caused by DB. More information on the theory and technical aspects of solar radiation models can be found in Bonan (1989), Dubayah & Rich (1995), Flint & Childs (1987), Geiger et al. (2002), Hetrick et al. (1993), Kumar, Skidmore & Knowles (1997), Kirkpatrick & Nunez (1980), Mazza et al. (2000), and Swift (1976).

The humid-thermal ratio (HTR) has successfully been used to develop and test relationships between different plant infestations levels in varied climate conditions in areas such as Australia, India, Europe, and North America. An HTR prototype has been developed to simulate ecological conditions appropriate for the establishment and spread of plant diseases in India (Jhorar et al., 1997). The HTR method has also been used to evaluate the risk of the establishment and spread of Karnal in wheat, grown under a variety of climatic conditions and in different areas (Mavi et al., 1992; Stansbury & Pretorius, 2001; Workneh et al., 2008). This method has potential value in researching insect pests and their associated diseases, which may allow for the prediction of occurrence and non-occurrence under specific combinations of climate and weather conditions.

Image processing for vegetation

In order to detect changes, important information must be provided including spatial distributions of change, change rates, change trajectories for different vegetation types, and assessment of the accuracy of the change detection results. The three main steps in implementing change detection are (1) image pre-processing, e.g., geometrical rectification (GR), image registration (IR), minimum noise fraction (MNF) analysis, radiometric, automorphic, and topographic correction (the latter is needed if the study area is close to mountains) (Bagheri et al., 2016; Bishop & Colby, 2002; Civco, 1989; Teillet, Guindon & Goodenough, 1982); (2) selection of optimal techniques to conduct the change detection analysis; and (3) accuracy assessments (Datt et al., 2003; Lu et al., 2004; Lunetta et al., 2006; Lyon et al., 1998; Song et al., 2001) (see Fig. 3).

Although the selection of appropriate change detection techniques is important for the accuracy of change results, in practice, it might not be easy to select a suitable algorithm for a specific change detection application. Some simple techniques can be used to provide change and non-change information (e.g., image differencing). Other techniques may be used to provide a complex matrix of change direction data such as that used for post-classification comparisons (Lu et al., 2004). This review provides examples of change detection methods that can be used to address DB infestations and their impacts on date palm trees.

Techniques and methods

Suitability model for detecting and investigating insect infestations

All of the methods used to study the relationships between dependent and independent variables discussed previously are traditional statistical methods, which sometimes might not reflect the complicated relationships between infestations and environmental factors. In particular, ecological and geographical environments represent complex systems in which individual elements interact to create complex behaviour, and consequently, complex methods such as ANN, Cellular Automata (CA), and multi-agent systems (MAS) may be better suited to study the relationships and conduct factor analyses in insect infestation or disease detection research and to perform spread simulations (De Smith, Goodchild & Longley, 2007).

Numerous suitability models have been proposed to identify locations that have a particular set of characteristics.

In Hernandez et al. (2006), the authors compared four different models (BIOCLIM, GAPP, DOMIN, and MAXENT) and found that MAXENT was most capable for producing useful results with small sample sizes and minimum species occurrences. These models can also be used to identify areas that are susceptible to risks such as insect infestations, based on conditions favoured by the species. For example, a relevant study (Drees et al., 2010) used the habitat suitability selection method to model potential conservation areas for a rare ground beetle species (using barcode index number or BIN). Specifically, they used five different data sets to identify several key habitat factors for Carabus variolosus stress levels. A model was developed in Bone, Dragicevic & Roberts (2005) by using fuzzy theory to identify areas of susceptibility to Dendroctonus ponderosae Hopkins in Canada. However, spatial data have unique characteristics that can impact the results of the model (Crooks & Castle, 2012).

Raster data models are often used for finding and rating suitable locations. The raster overlay results are formatted in a single layer of suitable versus unsuitable cells, rather than in a vector layer with many polygons and an attribute table, which contains the attribute values for each of the polygons. There are two ways to create raster suitability layers. The first approach is to query the individual sources to create the suitability layer. The query can be used to create a suitability layer with two values, ‘1’ for cells meeting all criteria of a suitable habitat, and ‘0’ for the others. Because the layer consists of only two values, one indicating suitable and the other unsuitable cells, they are called binary suitability layers. Binary processing however is not always necessary. Combined with other evaluation models, suitability mapping can be achieved by overlaying directly or by post processing the overlay results. Figure 4 shows a process that could be used to find suitable location conditions (habitat) for insects such as DB by using a raster method overlay.

The uncertainty that results from geo-processing operations, demonstrates that sophisticated spatial analysis cannot be achieved using traditional, deterministic geoprocessing methods alone (Goodchild & Glennon, 2010; Zhang & Goodchild, 2002). Fuzzy logic is a superset of Boolean logic and has the ability to handle uncertainty in data that arises from vagueness instead of randomness alone (Li et al., 2010).

Fuzzy logic can be utilised to extract information from high resolution RS data and combined with a raster-based spatial data to produce maps representing the spatial variation of vulnerability to pests across a landscape (Zhang & Foody, 2001). This method also allows for partial association with one or more classes, meaning that objects may be represented by a value based on a membership function between ‘0’ and ‘1’ (Li & Zhao, 2007). The membership function of an element x belonging to a fuzzy set A is computed by: (2) $${\text{μ}}_{\text{A}}:U\to [0,1]$$where U is the universal set of x. The concept of fuzzy sets has also been employed for defining the spatial and attributes characteristics of geographic objects (Burrough & Frank, 1996; Wang & Hall, 1996). The results of such analysis can be rendered directly into a decision framework via maps, tables, and charts. The results can also be used in further analyses or to provide additional understanding of the problem.

The challenge in any particular area of study is the geographical extent and the resolution of analysis, which is determined by the phenomenon being modelled. To achieve validity, researchers must ensure that they are using accurate and current data whenever possible. If the data are from one’s own organisation, one can rely on data quality controls that are in place. Data quality should be checked against alternate sources if possible in order to ensure it meets the requirements of the analysis. Assessing the quality of data will provide guidance to predicting what level of confidence can be attributed to the result of the modelling work.