Remote Sensing of Mangroves

Table of contents

1 Introduction

2 Mangroves Ecology

3 Preliminary Thoughts

4 Hardware

5 Processing

6 Conclusion

References

Figures

1. Intoduction

The purpose of this work is to describe the application of remote sensing in the field of mangrove mapping. The question will be about the appropriate imagery and the diversity of processing methods. But first of all: What makes mangrove distinctive as a remote sensing target? Therefore, mangroves will be introduced with its ecological and physical properties. Furthermore, its significance for humans and the environment will be pointed out, which enforces the need to take care of this ecologically sensitive environment. This is why one should integrate the mangroves into coastal management processes. This requires knowledge about mangroves in their ecological habitats, from the micro-level, interrogated in the laboratory, up to a macro-level, observed from airborne or satellite platforms. Remote sensing will be introduced as an appropriate tool to map the extent and different habitats of mangroves and also to monitor dynamics over time.

In the past often conceived as wasteland, mangroves have come into the focus of environmental protection and scientific discussion more recently, because the decline around the world is remarkably. Mangroves once covered about ¾ of the tropical and subtropical coasts, but recently they have receded seriously. For instance, AKSORNKOAE (1993) estimates that Thailand’s loss over a period of 35 years reaches 50 %, of which 64 % can be attributed to the local shrimp industry. To get an impression of the overall extent of mangrove coverage, one can refer to figures of SAENGER ET AL. (1983) who estimates it to be 150,000 km² world wide at that time. The benefits for humans and the environment, drawn out of mangroves are various. According to STEVENSON (1996), the economic value of mangroves is about 100,000-277,000 US $ per km². They are used for poles, thatch, fuel and the bark can be used for tanning materials. In Malaysia, Bangladesh, Thailand and India, companies use mangroves to produce fuel wood, charcoal, wood pulp and timber in a organised commercial way (BLASKO ET AL. 1994). Besides, the harvesting is wide spread in parts of Africa, and Asia, and in Central America. Furthermore, mangroves are used because of its medical value and for fodder and honey production (BACONGUIS AND MAURICIO 1991). Also, mangroves serve as fishing areas, for recreation, wildlife reserves, human habitation and aquaculture (HAMILTON AND SNEDAKER 1984). Thus, it is also in the interest of the fishing industry to protect those areas which are important as nursery areas for a lot of commercial fish and crustacean species (ROBERTSON AND DUKE 1987). All in all, mangroves are very fragile ecosystems which are worth protecting for its economical and environmental value as well as to study in the local context of coastal management and on the global level in climate change response.

In the following chapter Mangroves Ecology, the mangroves are introduced as remotely sensed target. Afterwards, in the chapter Preliminary Thoughts the topics remote sensing and mangroves are brought together by introducing the physical properties and the surroundings, which are expected when mangroves are under observation from above. The chapter Hardware gives various applications as examples with their different imagery sources relating to their different project or study tasks. The subsequent chapter Processing treats different ways of how to deal with the obtained data, before the Conclusion will finalise and summarise this work.

2. Mangrove Ecology

Mangroves are often arranged as coasts along bands. Simplistically, they could be described as the interface between the ocean or a river on one side, and a tropical forest on the other, which is often the natural neighbour in this climate zone. However, due to the human impact and peculiar local climates, one has to also deal with pasture, coastal shrubs and urban areas. Pasture, for instance, can easily be mistaken with very sparse mangroves, while lush mangrove vegetation tends to be confused with the non-mangrove tropical forest. In addition, one has to take care of coastal-shrubs or, in general, non-mangrove vegetation within or nearby the labelled mangrove areas that yield similar spectral reflectance. A factor shaping mangrove stands is given by their local environmental properties. Coasts along mangrove bands are affected by sediment supply, which can be correlated with the strength and direction of the wind and the properties of the mud feeding sea, or relatively river streams. Thus, an increasing mud delivery can lead to mangrove progression (THOM 1984; AKSORNKOAE 1993), headed by the main pioneer species of Avicienna sp. and Sonneratia sp. (TOMLINSON 1994; LEE ET AL. 1996). A study at Pak Phanang in southeast Thailand (PANAPITUKKUL ET AL. 1997) found a progression of the mangrove forest edge of 38.6 meter per year on average over a time span of 28 years. This huge areal expansion is due to an inland erosion process and a sediment transport into the Bay of Pak Phanak. In general, bays are more attributable for mangrove progression than caps or neutral shore forms.

According to FROMARD ET AL. (2002), one can expect three main different structures of mangrove stands, which are displayed in Figure 1. (a) Different species zones are ordered regular and parallel to the coastline by a decrease in age towards the water line. The so-called “patch expansion” occurs where mud patches have resisted the erosion process and are thus the base for pioneer development. (c) The “arc shaped” colonisation is caused by both, the resistance of residual mud patches and a general sediment movement, in this case from east to west. The arrows describe the expansion direction of the species zonation. On the other side, the discrimination to the almost black-body water can be easily obtained by looking at a near- infrared channel. When the outline borders are identified, the outer areas can be masked out to avoid a bias in the mangrove analysis.

To tackle the above mentioned problems affecting the outlining process, it is useful to define the term mangrove. Mangrove communities occur in tropics and partly in sub-tropics around the world and can be seen as the most characteristic vegetation in those coastal zones (CHAPMAN 1969). One assumption for mangrove settling is a warm water surrounding which has to exceed 24°C in the warm months. This is given in the shallow coasts of the region about 30° north and south of the equator, but with a distinctive extension to the warm temperate zone of New Zealand at 38°3’ south (SPADLING 1997). The shore along bands of mangrove stands occur predominantly in intertidal regimes, and often at the interface of sea and fresh water supplying estuaries (MEZA DIAZ AND BLACKBURN 2001). The extent, delineation and the composition of mangrove communities depend on salinity, inundation regime, soil type, wave energy and temperature. In addition, aperiodic factors, such as cyclones and floods, are parameters (LUGO AND SNEDAKER 1974; HUTCHINGS AND SAENGER 1987). Examples for optimal conditions and thus the most extensive mangrove growth are forests in Ecuador, Thailand, Cameroon and Malaysia, where they can reach up to 45 m in height. Within critical conditions, mangroves reach heights of 1-2 m, such as on the arid coasts of the Arabian Gulf, the cooler areas of New Zealand, Rajasthan, Australia and South Madagascar (BLASCO ET AL. 1994). Mangroves belong to different genera and families and can be defined instead by its similarities in physiological, morphological and reproductive adaptations which enables a settling in those tidal, harsh and salty environment. All in all,about 80 plants of about 30 genera are estimated which means a relatively small diversity in relation to the richness of tropical ecosystems. Avicienna (family: Acanthaceae) and Rhizophora (Rizophoraceae) are the most dominant genera throughout the world. The further major genera are Laguncularia and Lumnitzera (both belonging to the famliy Combretaceae), Nypa (Arecaceae), Bruguiera, Ceriops, Kandelia, Rhizophora (all Rhizophoraceae) and Sonneratia (Sonneraticeae)(BLASCO ET AL. 1994).

3. Preliminary Thoughts

For mapping mangroves one has to consider some general challenges: Depending on the sensor, one has to deal with different resolutions and thus data volumes. For example, GAO (1998) revealed some results in which a better resolution did not lead to a higher accuracy, but actually confused the classification process (cp. GAO 1998). Furthermore, when pre- calculating the costs, one hast to take into account any radiometric correction process, which tends to be higher from airborne data than from satellite data. When merging imagery from different times, the growth of mangroves shouldn’t exceed the spatial resolution and also the seasonal effect has to be considered. In this case, a certain elbow-room can be thoroughly allowed, as GAO (1998) reveals in his study, in which different images are temporally separated by two years and furthermore by three months implying a seasonal impact. In this case of a mangrove mapping in Waitemata Harbour, Auckland, New Zealand, both sources could be filed in the winter season and the mangroves growth has not undergone a growth beyond the minimum resolution of 10 meters. This safety margin widens the source of imagery. Later on, the merging of different imagery will be a topic.

Before coming to the hardware requirements, the discussion concerns the general task of a project. What information do I want to get? Referring to the amount of publications, the overall goal is mostly to generate a map and outline the mangrove area, thus it is about at least two classes: mangrove and non-mangrove. Due to the distinctive loss of mangroves over the entire world, many studies were conducted to display progression and receding of mangrove areas over many years. In this respect, satellite imagery offers long-term coverage, which is attributable to this topic. For instance, the Landsat images are available from the early 1970s on and their quality might be sufficient for simplistic outlining purposes. More sophisticated studies separate the mangrove area into further classes, such as: stunted and lush, or different classes of height (often used for age detection). It is getting more complicated when it comes to a zonation analysis, which means a classification of the different species. The mapping of the species composition is meaningful for a better understanding of the ecological habitats of mangrove areas, because every specie can be linked with ecological values that describe their environment. In addition, in receding or progressing processes, the different species react individually to certain environmental changes (KOVACS ET AL. 2004) At this point, remote sensing techniques have now reached a limit. The problems might lie in the classification process rather than in the spatial resolution, which is already available in a sub-meter scale and probably steadily enhancing over future years. However, up to a certain level of detail, zonation analysis has been conducted successfully in various cases and mangroves can be separated and classified in their species sites. However, the success and level of detail also depends on the field site and on the budget for equipment as will be evaluated later on.

It is characteristic for the remote sensing of mangroves that the targeted areas are often inaccessible due to its soil humidity or vegetation density. HELD ET AL. (2000), for example, reported in their study of the mangrove ecosystem of the Daintree River, Queensland Australia, about inaccessible areas when taking preliminary ground measurements which, led to a gap in the spectral library. Furthermore, the tropical climate can be a challenge with long rain periods, dense cloud cover and high temperatures and humidity, which affect both the remote sensing process and the ground validation. The rain period in the tropics leads to a gap in the timeline of useable imagery of the optical sensors, which becomes especially problematic at responding to short term incidents such as strong winds, cyclones, surges etc.. An easing aspect might be the wide range of target properties, which is often given within relatively small areas. Due to mangrove sensitivity to a special ecological surrounding, (soil properties, salinity, tidal regime) the different types of mangrove, which will be for example classified by canopy density, height or species, might be arranged on small transect. The profile, given in Figure 2 illustrates two species, which are extremely different concerning canopy density and height, but, arranged within a cross section of a mangrove shoreline, lying next to each other. Another critical point typical for mangroves is the tidal level, which has to be taken into account because it strongly influences the soil’s backscatter and confuses comparative studies or merging operations.

4. Hardware

The methods for mapping mangroves are various and depend on the purpose of the study. Assessing the cost-effectiveness of them will mostly lead to the use of remotely sensed data, which offers some advantages over in-situ work (even though most remote sensing based operation requires some in situ-work as well). Depending on the density and the soil humidity, mangroves are often inaccessible for measuring operations or too wasteful concerning time and money efforts. Furthermore, the spatial extent of some operations disqualifies in-situ solutions from the beginning of the project planning. Another advantage comes up when working with temporal studies which go back over decades. From the 1970s on, satellite imagery with a broad coverage and steady temporal frequency is available and might be sufficient for long term studies.

The different approaches to obtain data cover a wide range of remote sensing techniques. According to the amount of written documentary, the most common and frequent platforms seem to be the Landsat and SPOT satellites. Figure 3, by GREEN ET AL. 1996, gives an overview of several works and their chosen sensors, which reveals a predominant occurrence of Landsat and SPOT systems. Here, it should be considered that aerial photography might be underrepresented due to a lack of publicity in this field (GREEN ET AL. 1996). When thinking about the choice of the data provider, it should be noted that there is no best-solution in general, but depends on the spatial extent of the target, the detail of information concerning temporal, spatial and spectral resolution that is required, and the available budget of money and time. Concerning these factors, the Landsat and SPOT systems can be a good all round compromise.

The downside of all optical scanning methods is the weather dependency. Landsat and SPOT, running the visible and the infra-red parts of the electromagnetic spectrum, suffer from the problem of cloud cover disturbance. The tropical conditions of mangrove areas are especially affected because of their high atmospheric energy, leading to unstable and less predictable weather conditions and a generally higher cloud production.

Moreover, rain periods deny the capturing for months and their temporal extent can cover in the worst case the whole year. Considering this challenge, a general trend of remote sensing develops the use of microwave technology, such as the active Synthetic Aperture Radar (SAR), an active radar system, which is not affected by cloud cover due to its use of longer, and thus more stable, wavelengths. But those efforts are still in an early experimental stage and are hardly documented up to now, meaning their use hasn’t been assessed critically. Some suchlike application and other solutions for the weather problem will be introduced later on.

As the Landsat and SPOT systems are the most popular systems for mapping mangroves, they deserve a closer look, as well as a comparison with each other. In some applications they are used in combination, but first of all, the differences shall be pointed out, because, in general, less sources means less costs and saves the incorporation effort as well. GAO (1998) worked in a comparative study with the SPOT XS and the Landsat TM systems to analyse the strengths and weaknesses of both systems and to assess the significance of spectral and spatial resolution. Gao (1998) mapped mangroves into lush and stunted categories and worked out their accuracy according to three different approaches. The first was generated by Landsat TM imagery with its 30 meter resolution. matic The second and third used SPOT XS imagery with resolution of 20 meters and 10 meters by the use of the panchroband (PAN). Even though the Landsat TM has a coarser resolution (30 meters), it revealed the more accurate results (95% and 87% for the both classes, in comparison to 77.5% and 67% for the SPOT XS 20 meters system and about 80% for the SPOT XS approach with 10 meters resolution). The reason might be the wider electromagnetic spectrum of the TM, which ranges from 0.45 to 2.35 µm in comparison to the SPOT XS system, which is restricted to a range of 0.50-0.98 µm. Thus, the additional wavelengths in the thermal infrared band leads to Landsat being the favourable system for this application, even though it has poorer spatial resolution. GAO (1998) also mentioned the problem, that a higher resolution demands a time-consuming processing and does not necessarily lead to a higher accuracy. Also more work on the training samples had to be done. However, one has to consider the task of GAO’s (1998) research, which was working out two density classes of the mangroves, beneath the general outlining of mangrove areas. Lush and stunted areas are internally relatively homogeneous and thus better suited to the TM, which does not offer the very high spatial resolution, but covers a wider spatial extent and supplies the mentioned spectral properties. More sophisticated studies with a detailed and specific mapping require a higher spatial resolution than the TM ones.

[...]