Proceedings p186-256

Proceedings of the 10th International Conference of AARSE, October 2014 186

MULTI-TEMPORAL ANALYSIS OF THE MIKEA DRY FOREST LANDSCAPE (SOUTHWEST MADAGASCAR)

R., Ravonjimalala(1,2); S., Razanaka(1);D., Herv(3); J., Bogaert(4); H., Randriambanona(1); M., Paegelow(5); S., Rakotondraompiana(2)

1- Centre National de Recherche sur lEnvironnement (CNRE), Antananarivo, Madagascar, R. Ravonjimalala [email protected]

2- Laboratoire de Gophysique de lEnvironnement et Tldtection. Institut & Observatoire de Gophysique dAntananarivo (IOGA), Universit dAntananarivo. Antananarivo, Madagascar

3- UMR 220 GRED, Institut de Recherche pour le Dveloppement. Montpellier, France

4- Unit Biodiversit et paysage, Universit de Lige /Gembloux Agro-Bio Tech Gembloux, Belgium

5- UMR 5602 CNRS, Maison de la Recherche, 5, all. A. Machado, Universit Jean Jaurs, F-31058 Toulouse cedex 9, France

KEY WORDS: landscape, dry forest, spatial analysis, fragmentation, Madagascar.

ABSTRACT

Deforestation and fragmentation of the forest landscape are among the major problems for the

conservation of ecosystems and the biodiversity. Human activities such as cultivation are among

causes of land use change, landscape fragmentation and forest habitat, due to slash and burning. The

study was undertaken in western part of Analamisampy village (Southwestern part of Madagascar),

and is about 167,835 hectares in size. The main land cover type this area is dry forest. Multi-spectral

SPOT images acquired in 2008 and 2013 were used to create a land cover map and accuracy

assessment was done using the field data. Using the land cover map, forest and Deforestation map

was developed. In 2008 to 2013 the deforestation rate was 1.6%, different from 6.7% reported by a

previous study undertaken for a period from 2008 and 2013. Spatial observation shows that

deforestation in southwestern region of Madagascar is mainly caused by fragmentation. The

deforestation mechanism in rainforest area is different as the deforestation front is there almost

linear. Two fragmentation mechanisms are also observed: circular and rectangular fragmentation

process. Further research will focus on the spatial modeling of deforestation process in the dry

forest.

INTRODUCTION

The Food and Agriculture Organization (FAO) estimated that the rate of deforestation as about 13

million ha.yr-1 (FAO, 2010). In most cases, deforestation through a process of habitat destruction is a

major driver of species extinction (Sala et al. 2000; Foley et al. 2005). It causes both a reduction of

forest area and a shift on landscape configuration (Skole & Tucker, 1993) which contributes to the

degradation of habitat. The effects of habitat loss and fragmentation on populations occur

concurrently (Fahrig 1997; Grez, 2005). Forest landscape fragmentation in turn affects the ecological

conditions of remaining forests and their biodiversity (Krummel et al., 1987). Tropical dry forests are

mailto:[email protected]


one of the most threatened ecosystems in the world (Hoekstra et al., 2005). The practices of slash

and burn farming, fuel wood extraction and charcoal production cause a very fragmented landscape

characterized by a mosaic of crops (corn, cassava, cotton, and legume grains), young fallows, old

fallows or secondary forests and natural dry forests. However, deforestation has significantly

reduced, after the implementation of protected area in 2007, where slash and burn farming and

charcoal production were forbidden.

The ecological consequences of forest fragmentation may vary depending on the spatial pattern

or the spatial configuration of the landscape (Armenteras et al., 2003; Ite & Adams, 1998). The

temporal assessment of forest changes based on satellite imagery and related to fragmentation

analysis becomes a valuable tool to evaluate the degree of threat to ecosystems (Armenteras et al.,

2003; Franklin, 2001; Imbernon and Branthomme, 2001; Luque, 2000; Sader et al., 2001). Several

studies have been conducted to follow the evolution of rain forest degradation in tropical area

(Imbernon and Branthomme, 2001; Sader et al., 2001; Skole and Tucker, 1993; Steininger et al., 2001;

Turner and Corlett, 1996) mainly in the Amazon forest (Jorge and Garcia, 1997; Laurance, 1999;

Laurance et al., 2000; Pedlowski et al., 1997; .Ranta et al., 1998; Sierra, 2000).

In Madagascar, a long-term analysis of spatial patterns of deforestation and fragmentation of

forest habitats across the landscape was conducted in the Mikea forest situated in the southwestern

part of the Island. The goal of this study is to describe the spatio-temporal evolution of deforestation

and forest fragmentation across the landscape. Land cover changes and changes in the spatial

configuration of the Mikea forest over time are studied using satellite imagery.

MATERIAL AND METHODS

The study area

Mikea forest is located in south-western part of Madagascar, 35 km north of the regional capital,

Toliara city. The study area is between Antampimbato village (2231'14.52"S 4327'35.53"E) and

Analamisampy village (2229'8.08"S 4339'13.13"E). This area is characterized by a semi-arid climate

according to Thornthwaite classification (Hoerner, 1986), with annual rainfall between 600 mm to

800 mm and one rainy season between January to March. Along a 10 km East-West transect, from

Ampasikibo village towards the Salary bay, the gradient of vegetal formations reveals the climatic

gradient to dryer conditions: savanna, crop and fallow, forest regrowth, dense and deciduous dry

forest, and finally scrublands on white sands of Mozambique canal seaside. We study the western

part of the Analasampy community, which is concerned by forest degradation.

Slash and burn farming is practiced on the dense and deciduous dry forest (Commiphora

grandifolia - Adansoniafony - Euphorbia laro), mainly on red sand (ergi). Leprun (1998) described four

types of dune sand formations from the coast to the East: the most recent are the white sandy beige

limestone of marine origin, followed by the non-calcareous clear red sands with low rates of clay

(5%), then the red siliceous sands (ergi) with clay content between 5 and 10% of wind origin; the

oldest are red sands (ergi) probably of river-wind origin, with clay content between 10 and 15%

Remote sensing data


Two SPOT 5 scenes use, acquired on 19th March 2008 and 14th May 2013. The SPOT 5 images have

3 spectral channels XS1, XS2, XS3 with 10 m spatial resolution and panchromatic channel (PAN) with

2.5 m spatial resolution. The multispectral scene was pan-sharpened to 2.5m using the PAN band

(Christophe et al., 2008). Training sets were collected from the field using the GPS. A first set was

used for 2013 classification and another set for validation. Images from Google Earth were also used

to create additional test data sets for validation or accuracy assessment.

Image processing and classification

Atmospheric corrections of images were done using QUAC method (Bernstein et al., 2005).

Images are also ortho-rectified using SRTM DEM data (Quarteroni, 2000). A supervised object-

oriented classification algorithm Support Vector machine (SVM) (Weston and Watkins, 1998) was

chosen for image classification.

Analysis of deforestation and fragmentation

Land cover maps were analyzed using ARCGIS software and FRAGSTAS modules for spatial indices

calculations (McGarigal et al., 2002). These indices will be used to characterize landscape

fragmentation pattern (Forman, 1995; Haines-Young and Chopping, 1996; Trani and Giles, 1999;

McGarigal et al., 2002).

NP is the number of patches of forest class and is determined for both 2008 and 2013 maps. The

more the forest is divided into small blocks, the higher is the index value. It provides information on

the fragmentation of forest class during the 5 years period comprised between these two dates.

The total area (TAj) occupied by the class j is calculated according to the equation (1) (Turner and

Ruscher, 1988; Trani and Giles, 1999):

Where aij is the area of the i-th spot in class j. The average value of patches area of class j is .

Largest Patch Index, or Dominance Dj(a), indicates the proportion of area occupied by the dominant

SPOT in the class j. It is calculated using the equation (2) (McGarigal et al., 2002; Lfman and Kouki,

2003):

Where amax,j is the area of the greatest spot on class j. . The more the value of

dominance, the less the class is fragmented.

The Shannons diversity areas index, denoted as Hj(a) is calculated using the equation (3) (McGarigal

and Marks, 1995).


This index measures the relative diversity of spots in the class j. The value Hj (a) depends on the

number of spots (NP), and their relative proportions aij/TAj.

It is equal to 0 when the class is made of only one spot and its value will increase with the number of

patches and evenness between areas patches of class (McGarigal and Marks, 1995).

The shape index IFij of the i-th spot in class j is calculated by equation (4) where Pij is the total

perimeter of the class j (Patton, 1975)

RESULTS AND DISCUSSION

Four land cover classes were classified: dense forest, secondary forest (old fallow), forest

regrowth (young fallow) and crop.

Figure 1 shows the landscape dynamics between 2008 and 2013. Forests occupy 53.30% of area in

2008 and 50.27% in 2013. Forest loss between 2008 and 2013 is about 1.45% per annum. Mikea

forest decrease is continuing despite its Protected Area status, but deforestation is significantly

reduced, from 6.7% to 1.45%.


Figure 1: Forest maps of eastern Analamisampy in 2008 (left) and 2013 (right)


Figure 2: Forest fragmentation of eastern Analamisampy: rectangular patterns in forest line,

circular patterns beside

Values of spatial indices for Mikea forest fragmentation are presented in table I.

Table I: Spatial Index for dense forests (2008 and 2013)

Spatial index 2008 2013

NP Patch number 569 765

TA Total forest area 5419 4999

LPI Largest patch index 47.65 42.31

SI Shape index 8.17 5.88

SDI Shannon diversity

index 1.08 1.07

The general structure of the forest fragmented landscape is characterized by many small patches

which sizes are generally less than 100 km . This shows a strong fragmentation of the forest cover of


the study area. Forests are particularly vulnerable to human activities when the NP index is high

(Saunders et al., 1991). From the fragmentation indices (Table I), we detect changes in the spatial

structure of the forest landscape between 2008 and 2013. Fragmentation is quantified by the

increase in spots number and the decrease in average areas of dense forest. The values of mean area

considerably drop (9.52 ha to 6.53 ha). Dense dry forest class is not dominant. . It is higher for the dry

forest class in 2008 (47.65%) and that of the dry forest in 2013 (42.31%). This shows that the dry

dense forest in 2008 is as strong as the fragmented dry forest in 2013.

Another observation is the evolution of the index forms IFj. We note in the forest class where the

number of plots increase that the value of the shape index decreased from 8.17 in 2008 to 5.88 in

2013. The index is high compared with the value of the regular shape of a circle or square (1 / 4) of

a spot (Krummel et al. 1987). The shape ratio we obtained in this study confirms the complexity of

the forms of dry forest deforestation.

CONCLUSION

Deforestation of the forest is reduced through the implementation of protected area status in

2007. Deforestation and logging have shared a sub-division of the remaining smaller plot forest

areas, although the impacts of deforestation were statistically significant. The cause of deforestation

in this region is not really linear. The figures showed fragmentation that is circularly or sometimes

rectangular

The number of spots between 2008 and 2013 increased from 569 to 765, reflecting a fragmented

with a fragmentation of the initial spots. Fragmentation was confirmed by the decrease in average

areas of dense forest, from 9.52 ha to 6.53 ha, while the dominance values decreased from 47.65%

to 42.31%. We also note that the value of the shape is a high index. Then, a complexity of the shape

of the dry dense forest is observed. Future work is to model the mechanism of deforestation in the

context of dry forest.

ACKNOWLEDGEMENTS

Funding for this research was provided by FPPSM (Forests, parks and poverty in South

Madagascar) project of French Foreign Office Ministry. R. Ravonjimalala beneficied a doctoral bourse

from RAMI () program of the Agence Universitaire de la Francophonie (AUF). Special thanks to the

SEAS-OI station (La Runion) for providing satellite data.

REFERENCES

Armenteras, D., Gast, F., Villareal, H., 2003. Andean forest fragmentation and the representativeness of protected natural areas in the eastern Andes, Colombia. Biological Conservation 113, 245256.

Bernstein, L. S., Adler-Golden, S. M., Sundberg, R. L., 2005. Validation of the Quick Atmospheric Correction (QUAC) algorithm for VNIR-SWIR multi- and hyperspectral imagery. SPIE Proceedings, Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XI. Vol. 5806, pp. 668-678.


Fahrig, L. 1997. Relative effects of habitat loss and fragmentation on species extinction. Journal of Wildlife Management 61, 603610.

Foley, .J.A., DeFries, R., Asner, G.P., Barford, C., Bonan, G., Carpenter, S.R., Chapin, F.S.,Coe, M.T., Daily, G.C., Gibbs, H.K., Helkowski, J.H., Holloway, T., Howard, E.A.,Kucharik, C.J., Monfreda, C., Patz, J.A., Prentice, I.C., Ramankutty, N., Snyder, P.K., 2005. Global consequences of land use. Science 309, 570574.

Food and Agriculture Organization (FAO) (2010). Global forest resources assessment 2010. Food and Agriculture Organization of the United Nations, Rome, 2010.

Forman, R.T.T., 1995. Land mosaics: The ecology of landscapes and regions. Cambridge University Press, United Kingdom. Pp 652

Franklin, S., 2001.Remote Sensing for Sustainable ForestManagement.Lewis Publishers, USA.

Grez, A., 2005. The importance of small fragments of Maulino forest for the conservation of epigeous beetles. In:Smith-Ramirez, C., Armesto, J.J., Valdovinos, C. (Eds.), Historia,biodiversidad y ecologa de los bosquescosteros de Chile.Editorial Universitaria. Santiago, Chile, pp. 565572

Haines-Young, R., and Chopping M., 1996.Quantifying landscape structure: A review of landscape indices and their application to forested landscapes. Prog. Phys. Geog. 20, 418 445.

Hoekstra, J.M., Boucher, T.M., Ricketts, T.H., Roberts, C., 2005. Confronting a biome crisis: global disparities of habitat loss and protection. Ecol. Lett. 8, 2329.

Hoerner, J.M., 1986. "Gographie rgionale du Sud-Ouest de Madagascar", Antanarivo, pp.189

Imbernon, J., Branthomme, A., 2001. Characterization of landscape patterns of deforestation in tropical rain forests. International Journal of Remote Sensing 22, 17531765.

Ite, U.E., Adams, W.M., 1998. Forest conversion, conservation and forestry in Cross River State, Nigeria. Applied Geography 18, 301314.

Jorge, L.A.B., Garcia, G.J., 1997. A study of habitat fragmentation in Southern Brazil using remote sensing and geographic information systems (GIS). Forest Ecology & Management 98, 3547.

Krummel, J.R., Gardner, R.H., Sugihara, G., ONeill, R.V., Coleman, P.R., 1987. Landscape patterns in a

disturbed environment. Oikos 48 (3), 321-324.

Laurance, W.F., 1999. Reflections on the tropical deforestation crisis. Biological Conservation 91, 109117.

Laurance, W.F., Vasconcelos, H.L., Lovejoy, T.E., 2000. Forest loss and fragmentation in the Amazon: implications for wildlife conservation. Oryx 34, 3945.

Lfman, S., and Kouki, J. 2003. Scale and dynamics of a transformingforest landscape.Forest Ecol. Manag. 175, 247252.

Leprun, J.C., 1998. Compte rendu de mission Madagascar (projetGEREM), Anatananarivo, ORSTOM-CNRE.

Luque, S., 2000. Evaluating temporal changes using Multi-spectral Scanner and Thematic Mapper data on the landscape of a natural reserve: the New Jersey Pine Barrens, a case study. International Journal of Remote Sensing 21, 25892611.


McGarigal, K. & Marks, B.J., 1995. FRAGSTATS: Spatial pattern analysis program for quantifying

structure. Department of Agriculture, General Technical Report, PNW-GTR-351, Pacific

Northwest Research Station, Oregon, USA, 122 p.

Mcgarigal, K., Cushman, S.A., Neel, M.C., Ene, E., 2002. Fragstats: Spatial Pattern Analysis Program for Categorical Maps.Retrieved January 20, 2003, from University of Massachusetts, Landscape Ecology Program Web site: www.umass.edu/landeco/research/fragstats/fragstats.html(accessed 28June2014).

Millington, A.C., Velez-Liendo, X.M., Bradley, A.V., 2003. Scale dependence in multitemporal mapping of forest fragmentation in Bolivia: implications for explaining temporal trends in landscape ecology and applications to biodiversity conservation. Photogrammetry & Remote Sensing 57, 289299.

Patton, D.R., 1975. A diversity index for quantifying habitat edge. Wildlife Society Bulletin 3,171-

173.

Pedlowski, M., Dale, V.H., Matricardi, E.A.T., Pereira da Silva Filho,E., 1997. Patterns and impacts of deforestation in Rondoma, Brazil. Landscape and Urban Planning 38, 149157.

Quarteroni, A., Sacco, R., Salari,F., 2000. Mthode numrique pour le calcul scientifique. Programme MATLAB. Collection IRIS. Pp. 247-402

Ranta, P., Blom, T., Niemela, J., Joensuu, E., Siitonen, M., 1998. The fragmented Atlantic rain forest of Brazil: size, shape and distribution of forest fragments. Biodiversity and Conservation 7, 385403.

Rey-Benayas, J.M., Pope, K., 1995. Landscape ecology and diversity patterns in the seasonal tropics from Landsat TM imagery. Ecological Applications 5, 386394.

Sader, S.A., Hepinstall, D.J.H., Coan, M., Soza, C., 2001.Forestchange monitoring of a remote biosphere reserve. International Journal of Remote Sensing 22, 19371950.

Sala, O.E., Chapin III, F.S., Armesto, J.J., Berlow, E., Bloomfield, Dirzo, R., Huber-Sanwald, E., Huenneke, L.F., Jackson, R., Kinzig, A., Leemans, R., Lodge, D.M.,Mooney, H.A., Oesterheld, M., Proff, N.L., Sykes, M.T., Walker, B.H., Walker, M.,Wall, D.H., 2000. Global biodiversity scenarios for the year 2100. Science 287, 17701774.

Saunders, D.A., Hobbs, R.J. & Margules, C.R., 1991. Biological consequences of ecosystem

fragmentation: a review. Conservation Biology 5,1832.

Sierra, R., 2000. Dynamics and patterns of deforestation in the western Amazon: the Napo deforestation front, 19861996. Applied Geography 20, 116.

Skole, D., Tucker, C., 1993. Tropical deforestation and habitat fragmentation in the Amazon: satellite data from 1978 to 1988. Science 260, 19051910.

Steininger, M., Tucker, C., Ersts, P., Killeen, T., Villegas, Z., Hecht,S., 2001. Clearance and fragmentation of tropical deciduous forest in the Tierras Bajas, Santa Cruz, Bolivia. Conservation Biology 15, 856866.

Trani, M.K., and Giles, R.H., 1999. An analysis of deforestation: Metrics used to describe pattern change. Forest Ecol. Manag. 114, 459 470.

Turner, I.M., Corlett, T., 1996.The conversion value of small, isolated fragments of lowland tropical rain forest. Trends in Ecology & Evolution 11, 330333.

http://www.umass.edu/landeco/research/fragstats/fragstats.html


Turner, M.G., and Ruscher, C.L., 1988. Changes in landscape patterns in Georgia, USA.Landsc. Ecol. 1, 241251.

Weston, J., and Watkins, C., 1998. Multi-class support vector machines. Technical report CSD-TR-98-04, Department of Computer Science, Royal Holloway, University of London, Royaume-Uni.


IDENTIFYING THE BEST SEASON FOR MAPPING EVERGREEN SWAMP AND MANGROVE SPECIES USING LEAF-LEVEL SPECTRA IN AN ESTUARINE SYSTEM IN KWAZULU-NATAL, SOUTH AFRICA

Heidi van Deventer 1, Moses Azong Cho1, 2, Onisimo Mutanga2, Laven Naidoo1, Nontembeko Dudeni-Tlhone 1

1. Council for Scientific & Industrial Research (CSIR), South Africa, Heidi van Deventer [email protected]

2. University of KwaZulu-Natal, South Africa

KEYWORDS: Species discrimination, Swamp forests, Mangrove forests, Leaf spectroscopy, Random

Forest

ABSTRACT

Swamp and mangrove forests are some of the most threatened forest types in the world. In

Africa, these forests are essential in providing food, construction material and medicine to people.

These forest types have not sufficiently been mapped and changes in the extent or quality of these

habitats can therefore not be effectively monitored. Compared to traditional surveying methods,

remote sensing can be used to map these inaccessible areas over regional extents. This study

investigated which season would provide the best discrimination of six evergreen tree species,

associated with swamp (Ficus Trichopoda), mangrove (Avicennia marina, Bruguiera gymnorrhiza,

Hibiscus tiliaceus), wetlands in adjacent woodlands (Syzygium cordatum) and coastal floodplain

systems (Ficus sycomorus), using leaf-level hyperspectral data. Leaf spectra were collected from 113

trees for the winter, spring, summer and autumn months between the years of 2011-2012 in the

subtropical estuarine system of the uMfolozi, uMsunduzi and St Lucia Rivers, on the east coast of

KwaZulu-Natal, South Africa. The classification accuracy for each season was evaluated in the WEKA

software using the Random Forest classification algorithm. When the data was upscaled to canopy-

level, the results showed that all four seasons produced overall accuracies of > 90%. Spring, summer

and autumn produced the highest overall accuracy of 94.7%, whereas the overall accuracy for winter

was 89.5%. The results of the leaf-level analysis showed a decrease in accuracy of between 4 11%

for the four seasons. Similar to other studies, our results showed that the simulated object-oriented

approach showed a higher level in accuracy compared to the pixel-level approach. The results of this

study showed that evergreen tree species around the uMfolozi, uMsunduzi and St Lucia Rivers in

KwaZulu-Natal, South Africa, are highly separable over all four seasons.



INTRODUCTION

Mangrove swamp forests are considered to be one of the most threatened forest types in the

world (Valiela et al. 2001). The exact percentage loss of mangrove forests over the past 50 years is

globally estimated to range between 25% and 50% (Alongi 2002; Spalding et al. 2010). Losses are

largely attributed to the clearance of forests for aquaculture and agriculture, and to a lesser degree

to the impacts of climate change including sea-level rise, nitrification and drought (Alongi 2002;

Mucina and Rutherford 2006). In Sub-Saharan Africa, where more than 48% of people live below the

poverty line (http://data.worldbank.org/topic/poverty), mangrove forests provide essential

ecosystem services, including livelihood resources such as construction material, fuel, food and

medicine (Alongi 2002; Mucina and Rutherford 2006). Along other coastal, estuarine and riverine

forests, mangroves also provide inter alia watershed protection, coastal erosion control, and habitats

for some rare and endangered animal species (Alongi 2002). The World Atlas of Mangroves shows

the extent and diversity of mangroves across the globe (Spalding et al. 2010). However, further work

is required to acquire and monitor the distribution of other coastal forest types, which often co-occur

alongside mangrove species in estuarine systems.

Coastal forests, such as mangrove and swamp forests, are however often difficult to access for

traditional vegetation mapping surveys, owing to being waterlogged or occupied by dangerous

animals (United States Department of Energy (US DOE) 2012). Remote sensing provides an

alternative method to map these species at regional level and particularly for inaccessible areas. The

advances of new sensors, such as RapidEye, Sentinel-2 and WorldView-2 and 3, offer increased

spatial resolution to a level where individual tree canopies can be mapped. These satellite images

remain expensive, particularly for developing countries, hence choosing the best season to

discriminate between tree species associated with the different forest types, can allow for cost

savings while enabling effective monitoring.

This study investigated which season, amongst winter, spring, summer and autumn months,

would provide the best discrimination of six evergreen tree species, associated with swamp (Ficus

Trichopoda), mangrove (Avicennia marina, Bruguiera gymnorrhiza, Hibiscus tiliaceus), wetlands in

adjacent woodlands (Syzygium cordatum) and coastal floodplain systems (Ficus sycomorus), using

leaf-level hyperspectral data. Species discrimination was assessed at leaf level as well as at canopy

level, simulating the per-pixel and object-orientation classification of tree species at image level. Leaf

spectra were collected from 113 trees for the winter, spring, summer and autumn between 2011-

2012 in the subtropical estuarine systems of the uMfolozi, uMsunduzi and St Lucia Rivers, on the east

coast of KwaZulu-Natal, South Africa. The Random Forest (RF) classification algorithm has been

proven successful in tree species classification (Adelabu and Dube 2014; Naidoo et al. 2012), and was

therefore applied to assess the best season for discriminating amongst the six species.


MATERIALS AND METHODS

Study area

The iSimangaliso Wetland Park (28S, 3230E) is situated in a sub-tropical climate zone on the

east coast of South Africa in the KwaZulu-Natal province. Mean Annual Precipitation (MAP) ranges

between 1 000 1 500 mm on the coast (Middleton and Bailey 2008). The mean temperatures in

summer range between 23 30C and are approximately 10C in winter (Sokolic 2006). The channel

of the uMfolozi has been artificially joined to the uMsunduzi River, close to its estuary, while the St

Lucia River estuary is located about 1 km north of the estuary of the other two Rivers (Figure 1).

Figure 1: The estuaries of the St Lucia, uMfolozi and uMsunduzi are located within 1 km of each other, and situated within the iSimangaliso Wetland Park on the east coast of South Africa.

A number of indigenous forest patches are found around the estuaries. To the south of the

uMfolozi and uMsunduzi estuaries are coastal dune forests stretching north-south along the coast.

Ficus sycomorus is found on the floodplain system of the uMfolozi and uMsunduzi Rivers to the west

of the estuaries as remnants between the sugarcane farms. Dryland coastal forests are situated near

the St Lucia town with the largest patch remaining at the DukuDuku forest southwest of Khula

village. In the swamp forests, located between the uMfolozi and uMsunduzi Rivers, Ficus trichopoda

and Barringtonia racemosa species are found, whereas two mangrove species (Avicenna marina &


Bruguiera gymnorrhiza), Hibiscus tiliaceus and Syzyguim cordatum is primarily found along the

channels of the estuarine systems (Figure 2).

Figure 2: Distribution of coastal and inland indigenous forests, as well as other land uses, near the three estuaries of the St Lucia, uMfolozi and uMsunduzi Rivers (SANSA 2011).

Data collection

Six evergreen tree species, associated with the estuarine system, were sampled (Table 1) over

four seasons (winter, spring, summer and autumn) between 2011 and 2012. Five green and fully

expanded leaves were collected from across sun-exposed canopies of mature trees. The leaves were

placed in zip-lock bags in a cooled container and transported back to the laboratory. One spectral

measurement was recorded of the adiaxal surface of each of the five leaves using the leaf-clip device

of an Analytical Spectral Device spectroradiometer (FieldSpec Pro FR, Analytical Spectral Device, Inc,

USA.) within 3 to 5 hours of collection. The ASD covers the spectral range between 350 to 2500 nm

with a 1.4 nm sampling interval between 350-1050 nm range, and a 2 nm between 1 050 2 500

nm. Radiance was converted to reflectance against the scans of a spectralon reference panel. The

reflectance spectra of the five leaves were averaged to a single reflectance signature per tree.

Table 1. Number of tree species sampled for each season. The number of trees is derived as

the average of the number of usable leave spectra in brackets.

Tree species Common Acronym Winter Spring Summer Autumn Total number


name (n) (n) (n) (n) per species (n)

Avicennia marina

White mangrove AM

21 21 21 21 84 (105) (104) (104) (105) (418)

Bruguiera gymnorrhiza

Black mangrove BG

19 19 19 19 76 (95) (94) (95) (94) (378)

Ficus sycamores Sycamore fig FSYC 15 15 15 15 60 (75) (75) (75) (75) (300) Ficus trichopoda Swamp fig

FT 11 11 11 11 44 (55) (55) (55) (55) (220)

Hibiscus tiliaceus

Lagoon hibiscus HT

30 30 30 30 120 (150) (150) (150) (150) (600)

Syzigium cordatum Waterberry SC

17 17 17 17 68 (85) (85) (85) (85) (340)

Total per season: 113 113 113 113 452 (565) (563) (564) (564) (2 256)

Statistical Analysis

In order to establish the applicability of parametric methods for modelling the relevant spectra, the

distribution assumptions of the spectra were assessed using a variety of tools. Firstly, graphical

methods such as histograms and quantile-quantile (Q-Q) plots were used and showed departures

from normality. Further, numeric tools including Mardias and Henze-Zirklers tests were also

performed and indicated that the relevant spectra did not meet the multivariate normality

assumption. Therefore, a non-parametric classifier, Random Forest (RF) (Breiman 2001), which has

been proven successful in handling a large number of factors for classification, has therefore been

chosen to assess the species separability (Grossmann et al. 2010; Prasad et al. 2006). For each

season, the dataset was split into 66% training (2/3rds of the data) and 34% validation (1/3rd of the

data) for the leaf-level classification, using the WEKA software (Waikato Environment for Knowledge

Analysis v 3.6.11, 1999-2014, The University of Waikato, New Zealand, available from

http://www.cs.waikato.ac.nz/~ml/weka/). The default settings of WEKA were maintained in the

analysis for the number of trees (10) and features (12). Thereafter, the average of the five leaves was

calculated to simulate canopy level, and classification repeated in a similar manner. The overall

classification accuracy, kappa statistic and users and producers accuracies were calculated for each

species in each season.

RESULTS

The results of the leaf-level assessment showed that all four seasons produced overall accuracies

of > 83% (Table 2). In spring the classification accuracy was the highest (88%), compared to the

summer season which had the lowest accuracy (83%). The producers and users accuracy for all

species in spring was > 78%, whereas the other three seasons had accuracies ranging between 47

and 100%. In summer, Ficus trihopoda was also classified as Hibiscus tiliaceus and

Syzygium cordatum (results are not shown), resulting in producers and users accuracy of 47%. A

large number of Ficus trihopoda was incorrectly classified as Hibiscus tiliaceus in autumn (producers

accuracy = 54%). Syzygium cordatum also had producers accuracies between 63-65% in winter, and

a user accuracy of 63% in summer.

http://www.cs.waikato.ac.nz/~ml/weka/


Table 2: Classification accuracies of the six evergreen tree species at leaf-level.

Winter Spring Summer Autumn

Overall accuracy (%)

85.4 88.0 83.3 86.5

Kappa statistic

0.82 0.85 0.80 0.83

Accuracy (%) Producer User Producer User Producer User Producer User Avicennia marina 89.1 97.6 81.1 90.9 91.1 100.0 95.2 97.6

Bruguiera gymnorrhiza 95.8 79.3 100.0 88.9 97.1 97.1 94.3 97.1

Ficus sycamores 95.8 95.8 85.7 94.7 77.8 87.5 90.5 79.2

Ficus trichopoda 75.0 75.0 78.6 91.7 47.1 47.1 54.2 76.5

Hibiscus tiliaceus 87.0 87.0 91.7 84.6 87.0 81.6 85.4 83.7

Syzigium cordatum 62.5 65.2 88.0 81.5 73.9 63.0 90.9 74.1

In comparison to the leaf-level results, the upscaled canopy-level results showed an overall

improvement in accuracy by 5% (Table 3). The spring, summer and autumn seasons also produced a

much higher overall accuracy of 94.7%, compared to the varying results of the leaf-level analysis (83-

88%). The canopy-level results also indicated that the winter season had the lowest overall accuracy

(90%) of all four seasons, whereas the season with the lowest overall accuracy for the leaf-level

results was in summer. Most species showed producers and users accuracies above 75% at the

canopy-level analysis, except for Ficus sycamorus in spring, where the producers accuracy was

66.7%. Ficus trihopoda showed the lowest users accuracies for winter, spring and autumn at 50%. A

small number of Ficus sycamores have been misclassified as Ficus trichopoda in spring (results are

not shown), resulting in a producers accuracy of 67%.

Table 3: Classification accuracies of the six evergreen tree species at canopy level.

Winter Spring Summer Autumn

Overall accuracy (%)

89.5 94.7 94.7 94.7

Kappa statistic

0.87 0.94 0.94 0.94

Accuracy (%) Producer User Producer User Producer User Producer User Avicennia marina 88.9 100.0 100.0 100.0 100.0 100.0 100.0 100.0

Bruguiera gymnorrhiza 100.0 85.7 100.0 100.0 100.0 100.0 100.0 100.0

Ficus 100.0 75.0 66.7 100.0 75.0 75.0 80.0 100.0


sycamores Ficus trichopoda 100.0 50.0 100.0 50.0 80.0 100.0 100.0 50.0

Hibiscus tiliaceus 83.3 100.0 100.0 100.0 100.0 90.0 100.0 100.0

Syzigium cordatum 83.3 100.0 100.0 100.0 100.0 100.0 83.3 100.0

DISCUSSION

A number of studies have used multispectral Landsat and SPOT imagery to map mangrove species

in Africa: in Ghana (Mensah 2013), Kenya (Brakel 1984; Gang and Agatsiva 1992), Senegal

(Conchedda et al. 2008) and Tanzania (Wang et al. 2003). We could not find published work on

hyperspectral studies done in Africa in species discrimination of mangrove or estuarine tree species.

Hyperspectral species discrimination studies done in Africa included those on other wetland

vegetation (Adam and Mutanga 2009; Mafuratidze 2010) or savanna trees (Naidoo et al. 2012).

This study is the first study which assessed the separability of evergreen tree species associated

with estuarine systems on the east coast of South Africa, in the KwaZulu-Natal province. The selected

tree species are associated with mangroves, swamps, floodplain and other estuarine species. The

results of this study showed that evergreen tree species around the uMfolozi, uMsunduzi and St

Lucia Rivers in KwaZulu-Natal, South Africa, are highly separable at leaf spectral level over all four

seasons (>83%). At canopy level the accuracy of the classification increased by 5% for the spring,

summer and autumn seasons (95%), compared to the winter season (90%).

The upscaled canopy-level analysis produced higher accuracies across all seasons, improving the

classification between 4 11%. The canopy-level results also showed fewer accuracies below 75%,

and none < 50%. These findings concur with the findings of other species discrimination studies

where improvement in accuracies was found at canopy level, compared to the individual pixel-level

classification (Kamal and Phinn 2011). The use of object-oriented classification of mangroves, which

represents canopy level classification, using SPOT imagery also resulted in accuracies > 75%

(Conchedda et al. 2008; Vo et al. 2013).

Further analysis can include the assessment of specific regions of the spectrum which contributed

significantly to the classification, and assessing whether the classification accuracy remains high if

degraded to the spectral resolution of the bands of commercially available space-borne sensors. This

study will also be extended in future to assess whether these species are separable using RapidEye

images for four seasons. The work can also be extended to include more tree species of this

estuarine environment and elsewhere in Africa.

CONCLUSIONS

This study found that six evergreen tree species, associated with estuarine forests in the east

coast of South Africa, were highly separable at individual canopy level (95%) over three of the four

seasons (spring, summer and autumn). The accuracy of the classification was slightly less in the

winter (90%). The results of the leaf-level analysis showed a decrease in accuracy of between 4 11%

for the four seasons. Similar to other studies, our results showed that the canopy-level analysis,


which is comparable to an object-oriented approach, showed a higher level in accuracy compared to

the pixel-level approach.

ACKNOWLEDGEMENTS

This work was funded by the South African Department of Science and Technology (DST), the

National Research Foundation (NRF), the Council for Scientific and Industrial Research (CSIR) and the

Water Research Commission (WRC). We thank the iSimangaliso Wetland Park Authority and

Ezemvelo KwaZulu-Natal (KZN) Wildlife for granting access and logistical support. Dr Ricky Taylor,

former ecologist at Ezemvelo KZN Wildlife for guidance and advice. We are grateful to the following

fieldwork assistants: Mr S. Gumede, Dr F. Durand, Ms R. van Deventer, Mr A. van Deventer, Mr S.

Mfeka, Dr. E. Adam, Mr T. Colins, Mr K. Barker, Mr O. Malahlela and the iSimangaliso bursary

students Ms N. Nkosi, Ms T. Sibiya, Ms N. Gumede, Mr B. Mdamba, Mr B. Gumbi.

REFERENCES

Adam, E., Mutanga, O., 2009. Spectral discrimination of papyrus vegetation (Cyperus papyrus L.) in swamp wetlands using field spectrometry. ISPRS Journal of Photogrammetry and Remote Sensing, 64 (6), 612-620.

Adelabu, S., Dube, T., 2014. Employing ground and satellite-based QuickBird data and random forest to discriminate five tree species in a Southern African woodland. Geocarto International, DOI: 10.1080/10106049.2014.885589 (http://dx.doi.org/10.1080/10106049.2014.885589).

Alongi, D.M., 2002. Present state and future of the world's mangrove forests. Environmental Conservation, 29 (3), 331-349.

Brakel, W.H., 1984. Seasonal dynamics of the suspended sediment plumes from Tano and Sabaki Rivers, Kenya: analysis of coastal imagery. Remote Sens. Environ., 18 165-173.

Breiman, L., 2001. Random Forests. Machine Learning, 45 5-32.

Conchedda, G., Durieux, L., Mayaux, P., 2008. An object-based method for mapping and change analysis in mangrove ecosystems. ISPRS Journal of Photogrammetry and Remote Sensing, 63 (5), 578-589.

Gang, P.O., Agatsiva, J.L., 1992. The current status of mangroves along the Kenyan coast: a case study of Mida Creek mangroves based on remote sensing. Hydrobiologia, 247 29-36.

Grossmann, E., Ohmann, J., Kagan, J., May, H., Gregory, M., 2010. Mapping ecological systems with a random forest model: tradeoffs between errors and bias. GAP Analysis Bulletin, 17 (1), 16-22.

Kamal, M., Phinn, S., 2011. Hyperspectral data for mangrove species mapping: a comparison of pixel-based and object-based approach. Remote Sens., 3 222-2242.

Mafuratidze, P., 2010. Discriminating wetland vegetation species in an African savanna using hyperspectral data. University of KwaZulu-Natal, Pietermaritzburg, South Africa.

http://dx.doi.org/10.1080/10106049.2014.885589


Mensah, J.C., 2013. Remote sensing application for mangrove mapping in the Ellembelle District in Ghana. USAID Integrated Coastal and Fisheries Governance Program for the Western Region of Ghana. Narragansett RI: Coastal Resources Centre, Graduate School of Oceanography, University of Rhode Island. University of Rhode Island, United States of America.

Middleton, B.J., Bailey, A.K., 2008. Water resources of South Africa, 2005 Study (WR2005) and Book of Maps. Research Reports No.TT381/08 & TT382/08.

Mucina, L., Rutherford, M.C., 2006. The Vegetation of South Africa, Lesotho and Swaziland. South African National Biodiversity Institute (Strelizia), Pretoria.

Naidoo, L., Cho, M.A., Mathieu, R., Asner, G., 2012. Classification of savanna tree species, in the Greater Kruger National Park region, by integrating hyperspectral and LiDAR data in a Random Forest data mining environment. ISPRS Journal of Photogrammetry and Remote Sensing, 69 (0), 167-179.

Prasad, A.M., Iverson, L.R., Liaw, A., 2006. Newer classification and regression tree techniques: bagging and Random Forest for ecological prediction. Ecosystems, 9 181-199.

South African National Space Agency (SANSA), 2011. SPOT 5 national mosaic for 2011. 1.5 m spatial resolution (Republic of South Africa). Pretoria: SANSA.

Sokolic, F., 2006. The use of satellite remote sensing to determine the spatial and temporal distribution of surface water on the Eastern Shores of Lake St Lucia. M.Sc. thesis, UKZN, Durban, South Africa.

Spalding, M., Kaunuma, M., Collins, L., 2010. World Atlas of Mangroves. A collaborative project of ITTO, ISME, FAO, UNEP-WCMC, UNESCO-MAB, UNU-INWEH and TNC. Earthscan, London, United Kingdom.

United States Department of Energy (US DOE), 2012. Research Priorities for Tropical Ecosystems Under Climate Change Workshop Report. US Department of Energy Office of Science, DOE/SC-0153 Chambers, J.; Fisher, R.; Hall, J.; Norby, R.; Wofsy, S.

Valiela, I., Bowen, J.L., York, J.K., 2001. Mangrove Forests: One of the world's threatened major tropical environments. Bioscience, 51 807-815.

Vo, Q.T., Oppelt, N., Leinenkugel, P., Kuenzer, C., 2013. Remote sensing in mapping mangrove ecosystems - an object-based approach. Remote Sens., 5 183-201.

Wang, Y., Bonynge, G., Nugranad, J., Traber, M., Ngusaru, A., Tobey, J., et al., 2003. Remote sensing of mangrove change along the Tanzania coast. Marine Geodesy, 26 (1-2), 35-48.


ANALYSIS OF DIFFERENT EMPIRICAL RELATIONS FOR THE COMPUTATION OF THE REGRESSION

COEFFICIENTS FOR THE CLIMATIC CONDITIONS OF WESTERN CAPE PROVINCE OF SOUTH AFRICA

N.E. Maluta, V. Sankaran, T.S. Mulaudzi, F. Nemangwele, T.T. Khedzi, L. Takalani [email protected]

Department of Physics, University of Venda, South Africa

KEYWORDS: Solar energy, Global solar radiation, Regression coefficient, Actual Sunshine hour

ABSTRACT

Solar radiation is a crucial parameter in designing solar power devices, installation of solar

technologies systems, etc. These data are usually not available for most areas due to the lack of

weather stations especially developing countries. An attempt has been made to study the suitability

of different empirical formulae to determine the regression coefficients for the climatic conditions of

the Western Cape Province of South Africa. Most of the empirical relations available in the literature

use weather station data such as sunshine hours, precipitation, air temperature, etc. Theoretical

estimation of the global solar radiation needs the basic understanding of the regression coefficients

(a and b). The weather station data from the Agricultural Research Council (ARC) and South African

Weather Services (SAWS) are used in the present study to evaluate the regression coefficients for

one station. The computed values of this station have been used to evaluate the global solar

radiation for other stations in this region, and the computed global solar radiation was compared to

the experimentally observed solar radiation data received from ARC and SAWS. The computed

regression coefficients obtained in this study show a reasonable agreement with the previously

reported values, and these values were used to estimate the global solar radiation for the climatic

conditions of the Western Cape province of South Africa.

INTRODUCTION

Industrial developments, the conversion of raw natural material into finished products, the

abundance of material goods, our food supply and agricultural practices and the conveniences of our

everyday lives all depend on a ready supply of energy. Energy usage has grown more rapidly than the

population because of the simultaneous rising standards of living and the continuous change to more

energy intensive methods of providing such standard. Hence the use of energy is the most

fundamental requirement for human existence. The use of fossil fuels such as coal, oil and gas has

increased to a great extent and the use of these energy resources is seen as having major

environmental impacts. Hence renewable energy supply becomes the solution to the energy

problem.

The global solar radiation is not widely measured due to the cost of establishing new weather

stations, maintenance and calibration required [Li, et.al, 2014, Liu et.al 2009]. In developing countries

like South Africa, the lack of national grid lines in remote rural areas necessitates the use of solar

energy as an alternative energy source. In this regard, solar radiation at a given location is considered



as one of the primary factors for the installation of renewable energy devices [Maluta et.al 2014a,

Maluta et.al. 2014b]. Some of the well-established sunshine and temperature based models are

widely used to estimate the global solar radiation. In the current investigation, we have considered

the coastal area of South Africa, where we have based our calculation on the Western Cape Province,

to estimate the solar regression coefficient (a and b) from different theoretical empirical relations.

The estimated regression coefficients are then used to compute the global solar radiation using

Angstrom equation. The computed values of the global solar radiation on the horizontal surface were

compared with the experimentally measured values at the ARC weather stations.

STUDY AREA

South Africa is divided into nine provinces, namely Gauteng, Mpumalanga, North West, Kwazulu-

Natal, Eastern Cape, Northern Cape, Western Cape, Free State and Limpopo. South Africa is a

subtropical location, with latitude stretching from 22o to 35o S, while the longitude running from 17o

to 35o E [Maluta et.al 2014a, Maluta et. al, 2014b]. The current study focuses on the Western Cape

Province of South Africa.

The Western Cape Province has a Mediterranean climate characterised by dry summers and wet

winters. This is largely due to the position of the subcontinent relative to the low pressure systems

between 40 and 50 South [5]. These low pressure systems bring winter rainfall to the South-western

part of the country by means of a procession of cold fronts, when the westerly waves shift

northward. Dry conditions are attributed to variation of the westerly wave and high pressure

positions annually [5]. Mountain ranges stretching north-south along the west coast and east-west in

the south act as orographic barriers. These topographical features create a dry interior and, in the

coastal region they also augment the rain through orographic rain [DEA & DP Report 2011]. The

regional climate is also influenced by coastal low pressure systems, resulting in hot, dry bergwinds,

blowing from the interior and causing above normal warm conditions during spring and late winter.

THEORETICAL CONSIDERATIONS

The first correlation proposed for estimating the monthly average daily global solar radiation is

based on the method of Angstrom [Maluta et. al., 2014a, Maluta et al., 2014b]. The Angstrom

regression equation relates monthly average daily radiation to a clear day radiation in a given

location and average fraction of possible sunshine hours, and is given by the following equation:

0= +

0 (1)

H is the monthly average daily global radiation, H0 is the monthly average extra-terrestrial radiation,

a and b are regression coefficients, S is the actual measured sunshine duration and S0 is the

maximum daily sunshine duration.

The extra-terrestrial solar radiation on a horizontal surface can be calculated using the well-

established equation, [Maluta et. al., 2014a, Maluta et al., 2014b, Pandey et. al, 2013, Jamil et. al,

2010]:


0 = (243600)

[1 + 0.033cos (360

365)] [(

2

360) sin cos ] (2)

Isc is the solar constant (1367 Wm-2), is the latitude of the site, is the declination, s is the mean

hour angle for a given month and dn is the day of the year from 01 January to 31 December.

The hour angle s for horizontal surface is given by the equation [Maluta et. al., 2014a, Maluta et al.,

2014b, Pandey et. al, 2013, Jamil et. al, 2010]:

= 1( ) (3)

The declination () angle can be obtained using [Maluta et. al., 2014a, Maluta et al., 2014b, Pandey

et. al, 2013, Jamil et. al, 2010]:

= 23.45 sin (360284

365) (4)

The day length S0 is written as [Maluta et. al., 2014a, Maluta et al., 2014b, Pandey et. al, 2013, Jamil

et. al, 2010]:

0 =2

15 (5)

Most of the empirical relations use the temperature data, sunshine hours and other related

parameters to fix the regressions coefficient for a particular location. The following equations were

employed in the present study and brief descriptions of these equations are given below:

1. Rietveld Model [Pandey et. al, 2013, Jamil et. al, 2010, Muzathik et. al, 2013] Rietveld estimate several published values of a and b using the following equations respectively,

= 0.10 + 0.24 (

0) (6a)

= 0.38 + 0.08 (

0) (6b)

2. Zabara Model [Pandey et. al, 2013, Muzathik et. al, 2013] Zabara evaluated the monthly solar regression coefficients a and b values of the Angstrom-Prescott-

Page model with monthly measured sunshine hours using polynomial regression coefficient. The

equation for a and b coefficient is given by,

= 0.395 1.274 (

0) + 2.680 (

0)

2 1.674 (

0)

3 (7a)

= 0.395 + 1.384 (

0) 3.249 (

0)

2+ 2.055 (

0)

3 (7a)

3. Gopinathan model [Pandey et. al, 2013, Katiyar et.al, 2013, Muzathik et. al, 2013] Gopinathan suggested the Angstrom-Prescott coefficients for estimation of the global solar radiation

as follows:


= 0.265 + 0.07 0.135 (

0) (8a)

= 0.401 + 0.108 0.325 (

0) (8b)

where Z is the altitude in kilometres

4. While Soler model [Jamil et. al, 2010, Katiyar et.al, 2013, Muzathik et. al, 2013] While Soler has given a modification of the Angstrom type equation for each month, then the

regression coefficients of a and b was also found for different stations to be

= 0.179 + 0.099 (

0) (9a)

= 0.1640 + 0.1786 (

0) 1.0935 (

0)

2 (9b)

5. Medugu model [Medugu et. al, 2011, Yakubu et. al. 2012 ] Medugu and Yakubu have proposed the estimation of the coefficient a and b using the combination

of the latitude of the area and the relative sunshine hour duration using the following relationship,

= 0.110 += 0.235 + 0.099 (

0) (10a)

= 1.449 0.553 0.694 (

0) (10b)


The computations of the regression coefficients were performed using the five different models

discussed above. The values of the regression coefficients for Caledon in the Western Cape Province

under Chiltern Damwal ARC station, derived from different empirical formulae are listed in Table 1.

The observed global solar radiation data of the years 2011 and 2012 for the Waterford ARC station

were used to validate the computed global solar radiation data using the current estimated

regression coefficients. As can be seen from this table the values of a and b are in agreement with

each other. However it can be noted that a few models, such as Zabara, Gopinathan and Medugu

yield a larger value for b. It can also be observed that most of the empirical relations for a and b

depend on the sunshine hours throughout the year. It is also evident from Fig. 1 that the average

values of the regression coefficients change in some years. In the case of the Rietveld method, it can

be observed that the values are almost equal through the period of the study. It can also be observed

from the figure that the coefficients of the following models, namely Zabara, Medugu and Whiler

vary in some years during the period of study.

The computed yearly average regression coefficients values of a and b obtained in this study for

different models, are quite consistent with values reported in the literature [6, 12, 13]. We noticed

that the sum of a and b is less than 1, and it ranges between 0.6 to 0.9 as observed by previous

researchers [12]. Although the same actual sunshine hour data are used for all the models, we

observed that they show a different behaviour during the period of study. In some of the models one

of the solar regression coefficients, b, decreases, just like in the case of Zabara, Whiler and Medugu.


Fig. 1 Calculated solar regression coefficients (a and b) using five different model over

10 years period (2001 -2010)

The values of a and b obtained in the present work for the stations under study through each of

the relations, are used to compute the corresponding global solar radiation data for other stations in

this region listed in Table 2. These values are obtained for a period of two years to test the

applicability of the different models by comparing the observed global solar radiation data measured

by ARC with the values obtained through different methods.


Table 1. Average regression coefficients (a and b) calculated over the period of 10 years from

different models

Year Models Rietveld Whiler Zabara Gopinathan Medugu

2001

a 0.246 0.232 0.261 0.209 0.256

b 0.422 0.394 0.436 0.540 0.623

2002

a

0.247 0.239 0.215 0.204 0.269

b

0.426 0.389 0.492 0.553 0.594

2003

a

0.237 0.240 0.210 0.204 0.281

b

0.429 0.381 0.544 0.566 0.569

2004

a

0.232 0.236 0.259 0.209 0.282

b

0.429 0.382 0.549 0.567 0.566

2005

a

0.237 0.233 0.289 0.212 0.270

b

0.426 0.388 0.494 0.554 0.594

2006

a

0.211 0.236 0.260 0.209 0.262

b

0.424 0.391 0.462 0.546 0.610

2007

a

0.218 0.225 0.282 0.224 0.269

b

0.426 0.389 0.493 0.554 0.594

2008

a

0.233 0.226 0.273 0.220 0.234

b

0.417 0.289 0.479 0.518 0.671

2009

a

0.229 0.234 0.256 0.212 0.243

b

0.419 0.311 0.487 0.527 0.650

2010

a

0.232 0.232 0.266 0.214 0.264


b

0.424 0.374 0.500 0.548 0.606

Table 2.

Comparison between measured and average annual global solar irradiance (MJ/m2)

estimated, using different models for the year 2011 and 2012

Table 3. Statistical comparison between measured and average yearly global solar

radiation (MJ/m2) estimated, using different models for Waterford station

Year 2011 2012

Models

Observed 16.2357 12.5732

Rietveld 14.4976 14.6179

Whiler 13.5520 13.6564

Zabara 16.4677 16.6077

Gopinathan 13.8783 13.9986

Medugu 18.6400 18.8123

Models RMSE MPE MBE


Five models were used to compute the regression coefficients a and b, which were then used to

estimate the global solar radiation for the station under study. The respective estimated global solar

radiation is then compared with the measured values. The accuracy of the regression coefficients

used to compute the global solar radiation is tested by calculating the RMSE, MBE and MPE for the

given station. The comparison of the observed to the estimated global solar radiation is illustrated in

Fig. 2. It can be observed from Table 3 that lower values of RMSE, MBE and MPE were obtained in

this study. Overall, the determined annual regression coefficients for the Western Cape Province

appear to be consistent with the reports such as for the location in Lesotho [Mulaudzi et. al, 2013,

Safari et. al, 2009], where a = 0.266 and b = 0.512. Pereira [Mulaudzi et. al, 2013, Safari et. al, 2009]

has recommended that the general values of a and b to be 0.25 and 0.5 respectively, for any given

location.

Observed 2011 2012 2011 2012 2011 2012

Rietveld 0.0144 0.0162 0.4598 0.5765 0.0046 0.0058

Whiler 0.0153 0.0150 0.7200 0.3120 0.0072 0.0031

Zabara 0.0141 0.0195 0.0822 1.1239 8.23*10-4 0.0112

Gopinathan 0.0152 0.0157 0.63020 0.4062 0.0063 0.0041

Medugu 0.0168 0.0248 0.6798 1.7304 0.00680 0.0173


Figure 2.Comparison of the calculated global solar radiation from regression coefficients using

different models, and the measured global solar radiation for Waterford ARC station in

Western Cape Province for year 2011

CONCLUSION

The models available in the literature are applied to predict the solar regression coefficients and

daily global radiation for Western Cape Province over the period of 10 years and a comparison is also

made with measured data. From this study it is found that all the models except for the case of

Medugu, are suitable for the region under study, and they reasonably reproduce the measured solar

radiation data of the weather stations. The Angstrom-Prescott linear regression model can be used

to estimate the global solar radiation using the computed regression coefficients for Western Cape

Province of SA. These methods can assist the energy planners and installers of solar conversion

devices to design suitable devices in this region, particularly to assist the rural communities in this

region.

REFERENCES

Jamil AM, Tiwari GN. 2010. Solar radiation models review. IJEE 2010; 1:513532.

Katiyar AK and Pandey CK, 2013. A Review of Solar Radiation ModelsPart I JRE dx.doi.org/10.1155/2013/168048

Li, H., Cao, F., Wang X. and Weibin M., 2014. A Temperature-Based model for estimating Monthly Average Daily Global Solar Radiation in China. Scientific World Journal, dio10.1155/2014/128754.


Liu, X., Mei X., Li, Y., Wang, Q., and Jesen, J.R., 2009. Evaluation of temperature-based global solar radiation models in China, Agricultural and Forest Meteorology, 149, pp. 1433 1446.

Maluta, N.E., Mulaudzi, T.S., and Sankaran, V., 2014a. Estimation of the global solar radiation on the horizontal surface from Temperature data for the Vhembe District in the Limpopo Province of South Africa. International Journal of Green Energy, 11(5), pp. 454 464.

Maluta, N.E., Sankaran, V., and Mulaudzi, T.S., 2014b. Comparative analysis of different empirical relations for the computation of the regression coefficients and the global solar radiation for the climatic conditions of Vhembe District in Limpopo province of South Africa EEST Part A, 32(1), pp. 413-428.

Medugu, D.W., Yakubu, D., 2011. Estimation of mean monthly global solar radiation in Yola, Nigeria using Angstrom model. Advances in Applied Science Research, 2, pp. 414 421.

Mulaudzi T.S., Sankaran, V. and Lysko, M.D., 2013. Solar radiation analysis and regression coefficients for the Vhembe Region, Limpopo Province, South Africa. Journal of Energy in Southern Africa, pp. 24:27.12.

Muzathik AB., 2013. Potential of Global Solar Radiation in Terengganu, Malaysia. International Journal of Energy Engineering. 3, pp.130 136.

Olayinka, M.S., Nzeako, 2011. Estimation of global and diffuse solar radiations for selected cities in Nigeria, ISRN Renewable Energy, 22, pp. 13 33. doi:10.5402/2011/439410

Pandey, C.K., Katiyar, A.K., 2013. Solar Radiation: Models and Measurement Techniques. Journal of Energy, dx.doi.org/10.1155/2013/305207.

Safari, B., and Gasore, J., 2009. Estimation of global solar radiation in Rwanda using empirical models. Asian Journal of Scientific Research, 2(2), pp.68-75

The Department of Environmental Affairs and Development Planning Report, 2011. (DEA &DP Report 2011) http://www.westerncape.gov.za/other/2011/8/chapter_13_climate_change.pdf

Yakubu D., Medugu D.W., 2012. Relationship between the global solar radiation and the sunshine duration in Abuja, Nigeria. Ozean Journal of Applied Sciences 5(3), pp. 221228.


COMPARISON OF SUMMER AND WINTER CARBON DIOXIDE VERTICAL AND SPATIAL DISTRIBUTION

OVER SOUTH AFRICA

Xolile Ncipha and Venkataraman Sivakumar

School of Chemistry and Physics, University of KwaZulu-Natal, Durban 4000, South Africa, [email protected]

KEY WORDS: Carbon dioxide, carbon cycle, climate change, greenhouse gas, radiative forcing

ABSTRACT

Atmospheric carbon dioxide (CO2) is a long-lived greenhouse gas (LLGHG) with an atmospheric lifetime

of 50 to 200 years. Its contribution to radiative forcing is greater than the contribution of the other LLGHGs

combined. Emissions of CO2 from human activities are considered the single largest anthropogenic factor

contributing to climate change.The African continent plays a big and growing role in the carbon global

cycle, with potentially important climate change implications. However, current knowledge about Africas

role in the global carbon cycle remains limited because of sparse observation networks in and around the

African continent. South Africa, an industrialized country, has only one station with long-term ambient CO2

monitoring records. In this study, the data from the Tropospheric Emission Spectrometer (TES) instrument

on-board the Aura satellite are used to determine the vertical distribution of CO2 throughout the

troposphere over several areas in South Africa, to locate CO2 hotspots over the country and to infer the

types of CO2 sources at different locations during the summer and winter seasons. The wet and dry season

atmospheric loading of CO2 is compared. TES surface level CO2 data is also compared with the ground

based CO2 data from the Global Atmospheric Watch (GAW) Cape Point station. It was found that there is

a general shift to high concentrations from the summer to the winter seasons, with Cape Town being an

exception. Over the Highveld region, ground level emissions appear to be stronger in winter than industrial emissions. TES surface data was found to be in generally good agreement with Cape Point GAW ground

based station CO2 data.

INTRODUCTION

Atmospheric CO2 is a long-lived greenhouse gas with an atmospheric lifetime of 50 to 200 years. In

2005 it contributed about 63% of the gaseous radiative forcing responsible for anthropogenic climate

change. It had a contribution to radiative forcing of 1.66 Wm-2 as compared to a combined LLGHGs

radiative forcing of 2.63 Wm-2. Over the past 250 years human activities have changed the natural global

carbon cycle significantly: the average atmospheric mixing ratio of CO2 increased globally by about 100

ppm (36%), from a range of 275 to 285 ppm in the pre-industrial era (AD 1000-1750) to 379 ppm in 2005.

Among the LLGHGs, CO2 increases have caused the largest sustained radiative forcing over this period.

Emissions of CO2 from human activities are considered the single largest anthropogenic factor contributing

to climate change (Forster et al, 2007).



The African continent plays a big and growing role in the global carbon cycle, with potentially important

climate change implications. However current knowledge about Africas role in the global carbon cycle

remains limited because of sparse observation networks in and around the African continent. The

industrial CO2 emissions from Africa are small at continental scale, but they are significant for an

industrialized country such as South Africa (Williams et al., 2007). In South Africa, coal power plant

operations are responsible for the countrys power sector being the 9th highest CO2 emitter in the world

(Osman et al., 2014). Figure 1 shows CO2 emission sources from various industries in South Africa. Despite

this, the Global Atmospheric Watch (GAW) station at Cape Point is the only station in the country with

long-term ambient CO2 monitoring records (Brunke et al., 2011, DEA, 2009). This study aims at responding

to this limited information on ambient CO2 concentration in South Africa, by providing a 3-dimensional

picture of CO2 distribution in the country using the Tropospheric Emission Spectrometer (TES) instrument

on-board the Aura satellite. TES is used to determine the vertical distribution of CO2 throughout the

troposphere over several areas in South Africa, to locate CO2 hotspots over the country and to infer the

types of CO2 sources at different locations during the summer and winter seasons. The wet and dry season

atmospheric loading of CO2 is compared. This study presents the first attempt in determining a 3-

dimensional distribution of CO2 at national scale in South Africa. TES surface level CO2 data are also

compared with the ground based CO2 data from the GAW Cape Point station.

Figure 1. Main CO2 emission point sources in South Africa (Osman et al., 2014)

MATERIALS AND METHODS

The data used in this study were collected from the TES instrument on-board the Aura satellite. The

spacecraft was launched on 15/07/2004, in a near polar, sun synchronous, 705 km altitude orbit with

equator crossing at 1:45 pm local solar time. Aura makes near same location observations every 2 days

and a repeat cycle every 16 days. Its objective is to make comprehensive global stratospheric and

tropospheric composition measurements from the 4 instruments onboard. The over-passes over South


Africa occurred at midday and midnight. TES is an infrared, high resolution Fourier Transforms

Spectrometer (FTS) and it operates in both nadir (downward view) and limb (side view) modes to measure

the vertical distribution of atmospheric composition and surface parameters. It covers the spectral range

6503050 cm at a spectral resolution of 0.1 cm (nadir viewing) or 0.025 cm (limb viewing) (Beer, 2006).

Averaged CO2 vertical profiles over industrial source areas in Figure 1 were calculated by averaging

atmospheric CO2 horizontally at different altitudes up to 18 km altitude during the summer (DJF 2004-

2009) and winter (JJA 2005-2009) seasons. Table 1 shows the vertically profiled areas and their

demarcations. The boundaries of these areas are based on inclusion of urbanized and industrialized towns.

The central Northern Cape area is selected as a terrestrial background site. The summer (DJF 2004-2009)

and winter (JJA 2005-2009) seasons spatial distribution maps were plotted using GIS. The CO2 data were

collected from a column representing the boundary layer with the top at 700 hpa (approximately 3400 m),

the level of the semi-permanent stable layer (Garstang et al., 1996; Tyson et al., 1996). The CO2 column

data were averaged vertically from the surface to 700 hpa at first, the averaged data at different locations

were mapped and the spatial data gaps were filled by interpolation.

Table 1. Vertically profiled areas and their geographic demarcations

Area Latitude range Longitude range

Cape Town (CT) -32.5 to -35.0 17.5 to 20.5

Central Freestate (FS) -27.5 to -29.5 25.5 to 29.0

Central Northern Cape (NC) -28.0 to -31.0 21.0 to 25.0

Coastal Eastern Cape (EC) -32.0 to -34.5 25.0 to 29.0

Coastal KwaZulu-Natal (KZN) -28.0 to -31.0 30.0 to 32.7

Industrial Highveld (IHV) -24.8 to -27.2 26.8 to 30.2

Limpopo (L) -22.0 to -25.0 27.0 to 32.0

Ground based CO2 monitoring was done at the GAW station at the Cape Point which is located at the

southern end of the Cape Peninsula, South Africa, (-3421' , 1829'). The monitoring station is exposed to

the sea on top of a cliff 230 m a.s.l., about 60 km south from the city of Cape Town. The station is

dominated by SE - S SW wind, hence it is subjected to maritime air from the South Atlantic most of the

time (Brunke et al., 2011). TES surface level data were compared with the ground based data from

23/05/2005 to 20/12/2006 when the TES data were within 0.5 latitude and longitude radius from the

ground station.


There is generally good agreement between surface CO2 data from the TES and the ground based GAW

station data. The highest difference of 16.41 ppm (4.17% difference from larger value) was recorded on


23/03/2006 at 12:50. Data whose difference is less than 6 ppm (1.5% difference from the bigger value)

between the two data sets make 73.3% of the whole data series. Figure 2 shows the comparison of TES

surface and ground based GAW station CO2 data.

Figure 2. Comparison of TES surface and ground based monitored GAW Cape Point station CO2

data

Figure 3 and 4 show the spatial distribution of CO2 in summer (2004 - 2009) and winter (2005 - 2009)

seasons respectively, over South Africa. The areas of high CO2 concentrations or hotspots during both

seasons are consistent. However the atmospheric loading of CO2 is generally higher in winter than in

summer. The Mpumalanga and Limpopo Provinces are showing the largest hotspots, Durban and the

adjacent inland area also showing a notable CO2 signature in both seasons. The eastern part of the North

West province adjacent to Gauteng Province shows high CO2 concentration in winter. The Eastern Cape

Province shows a signature of high CO2 concentration in the north and northeast part of the province

adjacent to its provincial borders, in both seasons. The industrial point source at the northeast of the

Eastern Cape Province is accounted for in Figure 1, but the area source and some point sources north of

this province are not included in the inventory in Figure 1. Instances of trans-provincial boundary pollution

flow over several provinces are detected in both seasons. The concentration of CO2 over Cape Town is

higher in summer than in winter. Low background atmospheric CO2 loading in the summer season makes

the isolated small spots of high CO2 concentration over the country discernible, the spots suggesting point

source emissions during the summer season.


Figure 3. Summer (DJF 2004 - 2009) CO2 spatial distribution over South Africa

Figure 4. Winter (JJA 2005 - 2009) CO2 spatial distribution over South Africa

Figure 5 and 6 show the vertical distribution of CO2 in summer (2004 - 2009) and winter (2005 - 2009)

seasons respectively, over South Africa. There is a clear influence of the semi-permanent stable layers at

about 3400m (700 hpa) and 5500 m (550 hpa) on the vertical atmospheric loading of CO2. These stable

layers decouple the air aloft from the surface air (Garstang et al., 1996; Tyson et al., 1996). During the DJF

(2004 - 009) season both these stable layers influence is clearly visible and during the JJA (2005 - 2009)


season only the 700 hpa stable layer influence is notable. Above the stable layers, the CO2 profile

discontinues its trend with altitude from the one at the surface, it increases or decreases with altitude. If

the profile above the stable layer is the continuation of the surface profile, the rate of change of CO2 with

altitude becomes more rapid, whether it is increasing or decreasing. The profiles above the stable layers

decrease with altitude over areas with high surface CO2 concentration and increase over areas with low

surface CO2 concentration, showing the insulation of the air aloft from surface forcing.

Figure 5. Summer (DJF) 2004 - 2009 CO2 vertical profiles over selected areas

Figure 6. Winter (JJA) 2005 -2009 CO2 vertical profiles over selected areas


The Southern Africa transport modes influence on the CO2 atmospheric loading below the stable layers

is also discernible. During summer there is a high frequency of ridging high pressure systems at 850 hpa

level along the South African east coast bringing in clean maritime air. In winter there is an occurrence of

westerly waves and continental high pressure system driving continental air out into the Indian Ocean

(Garstang et al., 1996; Tyson et al., 1996). The spatial maps and the profiles show a general shift to high

concentrations from the summer to winter season, which is both caused by meteorology and change in

emission strength. In both seasons the central Free State (FS) area has the lowest atmospheric loading of

CO2 and Limpopo (L) province and the Industrial Highveld (IHV) region has the highest loading. The sharp

increase of CO2 concentration with altitude at the surface (Figure 5 and 6) is a signature of the influence

of industrial emissions, and this feature is more prominent in winter in many of these study areas, however

over the IHV region the surface emissions appear to be stronger in winter than industrial emissions. The

Eastern Cape (EC) study area profiles show high concentration at sea level in both seasons, this may be

due to the COEGA harbor activities.

CONCLUSION

This study highlights the importance of remote sensed data as a tool in air quality management.

Satellite data can offer a comprehensive 3-dimensional picture of air pollution over a wide space, and can

give useful information particularly in places where there is a lack of in-situ monitoring like in the South

African case. TES surface data is generally in good agreement with Cape Point GAW ground based station

CO2 data. The spatial plots, particularly during the summer season, show several small spots of high CO2

concentration. These spots could be industrial sources and they are not included in the industrial CO2

emissions inventory in Figure 1. Carbon dioxide hotspots in both summer and winter seasons coincide with

CO2 emitting industries in Figure 1, their spatial extent is expanded in winter when there is a contribution

from ground level sources like biomass and domestic emissions. There is a CO2 hotspot north of the Eastern

Cape Province. This source area may not be known because of inadequate CO2 monitoring in the country.

There is a general shift to high concentrations from summer to winter seasons with Cape Town as an

exception. This change in atmospheric CO2 loading is both caused by meteorology and change in emission

strength. In both seasons, the Limpopo Province and the Industrial Highveld region have the highest CO2

atmospheric loading, with the Free State Province having the lowest CO2 concentrations. Over the

Highveld region ground level emissions appear to be stronger in winter than industrial emissions. The

vertical profiles and spatial distribution plots from the satellite data makes it possible to infer whether the

tall stack industrial sources or ground level sources are dominating emissions at the source area.

ACKNOWLEDGEMENTS

The authors wish to thank the University of KwaZulu-Natal, School of Chemistry and Physics, for the

enabling environment, the Bureau Ocan Indien of the Agence Universitaire de la Francophonie (AUF) for

the training opportunities and all the support in general, and lastly the South African Weather Service Cape

Point GAW station for providing data.


REFERENCES

Beer, R.., 2006. TES on the Aura Mission: Scientific Objectives, Measurements, and Analysis Overview, IEE Transactions on Geoscience and Remote Sensing, 44, 5, 1102-1105.

Brunke, E-G., Labuschagne, C., Parker, B., and Scheel, H.E., 2011. Recent results from measurements of CO2, CH4, CO and N2O at the GAW station Cape Point, 15th WMO/IAEA Meeting of experts on Carbon Dioxide, Other Greenhouse Gases and Related Tracers Measurement Techniques, Jena, Germany , 7-10 September 2009.

Cias, P., Bombelli, A., Williams, M., Piao, S.L., Chave, J., Ryan, C.M., Henry, M., Brender, P., and Valentini, R., 2011.The carbon balance of Africa: synthesis of recent research studies, Philosophical Transactions of The Royal Society, 369, 1-20, doi:10.1098/rsta.2010.0328.

Department of Environmental Affairs, 2009. State of Air Report 2005. A report on the state of air in South Africa.

Forster, P., Ramaswamy, V., Artaxo, P., Berntsen, T., Betts, R., Fahey, D.W., Haywood, J., Lean, J., Lowe, D.C., Myhre, G., Nganga, J., Prinn, R., Raga, G., Schulz, M., and Van Dorland, R., 2007. Changes in Atmospheric Constituents and in Radiative Forcing. In: Change 2007: The Physical Science Basis.

Garstang M., Tyson, P.D., Swap, R., Edwards, M., Kllberg, P. and Lindesay, J.A., 1996. Horizontal and Vertical Transport of Air over Southern Africa, Journal of Geophysical Research, 101, D19, 23,721-23,736.

Osman, K., Coquelet, C., and Ramjugernath, D., 2014: Review of carbon dioxide capture and storage with relevance to the South African power sector, South African Journal of Science, 2014;110(5/6), Art.#2013-0188, 12 pages. http://dx.doi.org/10.1590/sajs.2014/20130188.

Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M.Tignor and H.L. Miller (eds.), Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge, United Kingdom and New York, USA: Cambridge University Press.

Tyson, P.D., Garstang, M., Swap, R., Kllberg, P. and Edwards, M., 1996. An air transport climatology for subtropical southern Africa, International Journal of Climatology, 16, 265-291.

Williams, C.A., Hanan, N.P., Neff, J.C., Scholes, J.R., Berry, J.A., Denning, A.S., and Baker, D.F., 2007. Africa and the global carbon cycle, Carbon Balance and Management, 2:3, doi: 10.1186/1750-0680-2-3, http://www.cbmjournal.com/content/2/1/3

http://dx.doi.org/10.1590/sajs.2014/20130188http://www.cbmjournal.com/content/2/1/3


INVESTIGATION OF URBAN HEAT ISLAND USING LANDSAT DATA

K. Abutaleb1, 3; Adeline, Ngie1; F. Ahmed2; M. H. Ahmed3; S.B. Elkafrawy3; S. M. Arafat3; A. Darwish3

1. University of Johannesburg, P.O. Box 524 Auckland Park 2006, South Africa: [email protected]. 2. School of Geography, Archaeology and Environmental Studies, Wits University, South Africa.

3. National Authority for Remote Sensing and Space Sciences, Egypt.

KEYWORDS: Urban heat island, Landsat, Egypt, Mono window, NDVI, Remote sensing.

ABSTRACT

Urban Heat Island (UHI) has become one of the worlds leading urban environmental issues.

Urbanization has increased concerns about the UHI, particularly in terms of human health and a

healthy environment. UHI results in a significant and sometimes dramatic increase in air temperature

change between the urban environment and its surrounding, which will likely alter how much energy

we consume. In a warmer condition, people would use more electricity for air conditioning. If the air

warms by 1.8C, the demand for energy used for conditioning and cooling would increase by about 5-

20%. The most important type of UHIs is the one that extends from the ground to the top of roofs

and canopy levels (canopy layer heat island) which cause different thermal projections throughout

the city. Urban heat island is a contributing factor to enormous health problems which will be

aggravated by the explosive growth of population and increasing impact of climate change. Urban

heat island studies are generally conducted in two ways; measuring air temperature by use of

automobile transects and weather station network and measuring surface temperature using

airborne or satellite remotely sensed thermal data. The aim of this study is to investigate UHI over

two cities in Egypt by applying the mono-window algorithm. Four Landsat scenes were selected,

processed and analysed to retrieve land surface temperature during 2002 and 2012 for each city.

Surface emissivity was calculated based on NDVI and atmospheric transmittance calculated based on

water vapour content which was estimated from ground weather data.

INTRODUCTION

Urban heat island (UHI) is a climatic phenomenon where urban areas have higher air temperature

than their peripheries due to anthropogenic modifications of land surfaces. Urbanization has

negative influences on the environment, including air, temperature and alteration of landscape.

Building patterns, structures and materials are related to urban temperature increase, energy

consumption, pollution and the production of waste. (Santamouris et. al., 2001). A combination of

factors leads to the development of the UHI; some factors could be controlled and some could not be

controlled. UHI is measured by the differences in the surface temperature, which can vary between

city centres and the peripheries by up to 5C (United Nations, 2010).

Building materials thermal properties, urban design geometry, roads width and directions,

existence of canyons, anthropogenic factors, existing land use/cover and altitudes are some of the

factors that contribute to UHI phenomenon. Building materials reflectance is low so they reflect less

and absorb more energy, which lead to temperature increase at surface level. Building materials

have high remittance values so they release heat quickly and stay cooler. Common building materials


such as tar, asphalt, brick and concrete store solar energy during the day and release it at night, so

UHI intensity is reported to be stronger at night time. Urban design geometry where urban canyons

are created by narrow streets and tall buildings decreases wind speed and increase impervious

surfaces that trap heat. Anthropogenic factors such as waste heat from vehicles and altered land

surface cover like porous vegetation is replaced with non-porous materials, thus restricting

evaporative cooling (Vereda and Cynthia, 2007; Li-Wei and Wan-Li, 2009).

Satellite thermal infrared (TIR) sensors measure top of the atmosphere (TOA) radiances, from

which at sensor temperature can be calculated by Planks law (Dash et al., 2002). The TOA radiance is

a mix

Proceedings p186-256

Documents

Transcript of Proceedings p186-256