Introduction

Agriculture is one of the key socio-economic activities substantially affected by

climate variability and change globally [1]. The impact of weather and climate
variability and change is more remarkable in the Arid and Semi Arid Lands (ASALs).
In Kenya , maize is the most widely grown crop and maize farmers in small-scale
farms grow maize under rain-fed condition. Maize production is expected to
decrease by 3-10 % until 2050 due to climate change effects (Thornton and Cramer
2012). Increasing risk of invasion by vectors of crops virus diseases in eastern Africa
is one of the evidences for the climate change effect.
This study investigates the use of two different optical sensors, the multispectral imager (MSI) onboard the RapidEye satellites and the operational land imager (OLI)
onboard the Landsat-8 for mapping within-field variability of crop growth conditions and tracking the seasonal growth
dynamics.

1. Adamgbe EM, Ujoh F. Effect of variability in rainfall characteristics on

maize yield in Gboko. Nigeria J Environ Prot. 2013; 4:881-7.
Publisher Full Text

With increasing population pressure throughout the world and the need for increased agricultural
production there is a definite need for improved management of the world's agricultural
resources. To make this happen it is first necessary to obtain reliable data on not only the types,
but also the cropping system.

To address this challenge, we exploited use of different high spatial resolution multitemporal optical
it is believed that the red edge band (690- 730 nm) of the RapidEye sensor allows
better estimates of chlorophyll content for crop monitoring.

Crop mapping in West Africa is challenging, due to the unavailability of adequate

satellite images (as a result of excessive cloud cover), small agricultural fields
and a heterogeneous landscape. To address this challenge, we integrated high
spatial resolution multi-temporal optical (RapidEye)

In recent years, agricultural land use has experienced high expansion rates in many
parts of the world [1]. This expansion is mainly due to high population growth
(especially in developing countries) and the need to grow more food to meet the
rising food demand. Accurate and up-to-date information on agricultural land use is
essential to appropriately monitor these changes and assess their impacts on water
and soil quality, biodiversity and other environmental factors at various scales [2
4]. This is particularly important considering the looming effects of climate change
and variability. Updated information on agricultural land use can help in monitoring
changes in cropping systems and gauge farmers reaction to the changing climate.

Additionally, a wide range of biophysical and economic models can benefit from
this information and improve decision-making based on their results. Remotely
sensed (RS) data provide useful information for agricultural land use mapping.
Periodic acquisition of RS data enables analysis to be conducted at regular intervals,
which aids in identifying changes. Optical systems, which have largely been relied
upon for agricultural land use mapping [5,6], measure reflectance from objects in
the visible and infrared portions of the electromagnetic spectrum. The amount of
reflectance is a function of the bio-physical characteristics of the reflecting feature
(e.g., canopy moisture, leaf area and level of greenness of vegetation). Since
different crops at varying vegetative stages exhibit different bio-physical
characteristics, optical images have been useful in previous crop mapping studies
[79]. However, the reliance of optical systems on the Suns energy limits image
acquisition in cloudy or hazy conditions. Images acquired during these periods are
normally of little use in mapping due to high cloud/haze cover. Whereas on irrigated
land under arid conditions, the entire growing period can be easily covered by
optical data [10,11], agricultural land use mapping efforts in rainfed dominated
agricultural regions, like West Africa (WA), are hampered, because the rainfall
season coincides with the cropping season. Consequently, little or no in-season
images are available for agricultural land use mapping, leading to challenges in
discriminating between different crop types or crop groups [1214]. For example, a
number of land use studies [1517] in WA have had to lump all crop classes into one
thematic class (cropland), due to a poor image temporal sequence. Synthetic
aperture radar (SAR) systems are nearly independent of weather conditions. Unlike
optical sensors, active radar systems have their own source of energy, transmitting
radio waves and receiving the reflected echoes from objects on the Earths surface.
The longer wavelengths of radio waves enable transmitted signals to penetrate
clouds and other atmospheric conditions [18], which make radar systems highly
reliable in terms of data provision, especially during periods in which optical sensors
fail [1921]. Moreover, the information content of radar imagery differs from that of

optical data owing to differences in how transmitted signals from the two systems
interact with features on the ground. A radar sensor transmits an electromagnetic
signal to an object and receives/records a reflected echo (backscatter) from the
object. Backscatter intensities recorded by radar systems are largely a function
Remote Sens. 2014, 6 6474 of the size, shape, orientation and dielectric constant of
the scatterer [22]. Thus, in vegetation studies, radar backscatter intensities will
differ based on the size, shape and orientation of the canopy components (e.g.,
leaves, stalks, fruit, etc.). Crops with different canopy architecture and cropping
characteristics (e.g., planting in mounds) can be distinguished based on their
backscatter intensities [2325]. The recent introduction of dual and quadpolarization acquisition modes in many radar satellites (e.g., Radarsat-2, PALSAR,
TerraSAR-X) further increases the information content in radar data. Owing to the
differences in imaging and information content, data from optical and radar systems
have been found to be complementary [26]. Several studies have shown that
integrating data from the two sources improves classification accuracies over the
use of either of them [27]. The authors of [23] tested the integration of Landsat TM
and SAR data (Radarsat, ENVISAT ASAR) for five regions in Canada. They concluded
that in the absence of a good time series of optical imagery, the integration of two
SAR images and a single optical image is sufficient to deliver operational accuracies
(>85% overall accuracy). The authors of [28,29] noted an increase of 20% and 25%,
respectively, in overall accuracy when radar and optical imagery were integrated in
crop mapping. Other studies found percentage increases between 5% and 8% when
the two data sources were merged [13,3034]. In this study, high resolution multitemporal optical (RapidEye) and dual polarimetric (VV and VH) radar data (TerraSARX) have been combined to map crops and crop groups in northwestern Benin, West
Africa. Excessive cloud cover during the main cropping season in West Africa has,
for many years, hindered crop mapping efforts in the sub-region due to the
unavailability of satellite images. A recent study [12] conducted in the sub-region
with multi-temporal RapidEye images identified poor image temporal coverage as
the limiting factor in accurately discrimination between certain crop types. A further
limiting factor is the heterogeneity (small patches of different land use and land
cover types) of the landscape [35], which leads to spectral confusion between
classes, especially when per-pixel approaches are employed [36]. In order to reduce
this confusion, a field-based classification approach was employed [37,38]. Vector
field boundaries were derived through image segmentation. A per-pixel
classification result was then overlaid and the modal class within each field
assigned to it. The aim of this study was to combine optical and radar data to
ascertain the contribution of radar data to crop mapping in WA. The specific
research question addressed is: can dual polarized radar images acquired during
peak cropping season months complement optical data to improve classification
accuracies in crop mapping?

High-spatial resolution optical remote sensing observations can provide crop

information at a spatial scale suitable for field to subfield level studies. The
capability for simultaneous acquisition over a large area allows for capturing spatial
variability due to underlying soil properties and management practices. It can
greatly alleviate the workload for conducting crop surveys or field measurements.
The time series observation is especially useful for tracking the seasonal trend of
crop growth and improving our understanding of canopy functioning. Multiple
optical remote sensing products over a growing season have been used for crop
biomass and yield estimation with a radiation use efficiency model (RUE)1 and have
proven to be useful in reducing the uncertainty of several input descriptors of crop
models using the data assimilation approach.2,3 Unlike the moderate-resolution
satellite sensors such as the MODIS and AVHRR, the relatively longer revisiting cycle
of a high-resolution satellite sensor is largely affected by cloud contamination and
hence leads to missed acquisitions during part of the key growth stages. For
continuous monitoring of crop seasonal development trends, it is advantageous to
be able to use data available from different sensors to shorten the revisit cycle.

The automatic cropland classification based on the data from spaceborne imagery is one of the most important sources of valuable information about the
composition and the development of a variety of crops grown in different agricultural regions around the world. A general goal is to be able to estimate the
area of specific crops,1,2 monitor their health, and predict their yield.3,4 A fundamental task in such applications is to determine the type of crop that is
grown on a specific stretch of land.57 The result is that the cropland classification has a significant role in the proper monitoring and management of land
use on a local as well as a global level and represents an important factor in the overall agricultural policy making. 8 Therefore, it is essential to make such
tools more accessible to different parties involved in the agricultural market and offer them the expertise acquired from the case studies of their
deployment and development.
The recent advances in the satellite imaging technology provide researchers and practitioners with ever more data in both quantitative and qualitative
ways. This opens new opportunities for the extraction of meaningful and useful information, but it also creates new challenges regarding the choice and
the development of appropriate methods for their processing. While satellite imagery for crop identification and crop-covered area estimation is a practice
with more than 30 years of tradition,9,10 the sensors have only recently gained resolution that allows for precision monitoring in the case of agriculture
practices based on small land parcels. 6,11,12 The fact that most of these satellite imagery systems are commercial represents one of the obstacles for their
widespread use and produces the need for the effective and efficient use of a vast amount of freely available data, like the high-quality data already
available through the Landsat program13 or the data that are planned to be published under similar terms by the upcoming Sentinel missions. 14
In line with these efforts, this paper presents a pixel-based cropland classification study that utilizes a time series of multispectral images with different
properties which were acquired at different resolutions by different imaging instrumentsLandsat-8 15 and RapidEye.16 It also explores the capabilities of
the proposed data fusion method for their combination with the aim of overcoming the shortcomings of different instruments in the particular cropland
classification scenario characterized by the specific size of crop fields over the chosen agricultural region situated in the plains of Vojvodina in northern
Serbia (southeastern Europe). It can be said that this scenario, where agriculture is based on very small areas dedicated to specific crops, is quite common
in different parts of the world. One of the problems that arises is the presence of crop fields with very small areas and small dimensions compared to the
spatial resolution of the freely available multispectral imagery. Additionally, if there is no available information about the crop field boundaries or cadastral
data (such as in this study) which would enable the application of an object-based classification approach, it is of great significance to develop methods
that effectively exploit available data and provide an improvement over the existing pixel-based classification approaches through the combination of
different data sources. Therefore, this paper proposes a data fusion method that is successfully utilized in combination with a robust random forest
classifier17 in improving overall classification performance, as well as in enabling application of satellite imagery with a coarser spatial resolution in the
given specific cropland classification task. A random forest classifier was chosen due to its recent use in remote sensing and because it has been accepted
as an efficient tool in crop classification.1820
The proposed method exploits different spectral and spatial resolutions of two different data sources in order to mitigate the described problem. Through
feature-level fusion where composite features are extracted from all available multisensor data, 21 a data integration, which is considered as a low-level
data fusion, is employed. It addresses the problem by using one mosaic multispectral image of the observed area, which is formed by mosaicking images
acquired in the short time interval by the constellation of commercial satellites with a higher spatial resolution as an addition to the freely available
Landsat-8 multispectral time series with a coarser spatial resolution which is acquired over a longer time interval that covers the phenological
development of all observed crop types. This approach is related to a recent study, 22 which also tried to exploit the advantages of different data sources,

but with application in the detection of vegetation changes. Through data fusion, it extends the applicability of the pixel-based classification using freely
available satellite imagery with a coarser spatial resolution to the classification scenarios that demand finer spatial resolution due to the agricultural
practices that are characterized by the growth of different crop types on the very small parcels of farmland.

In line with these efforts, this paper presents a pixel-based cropland classification study that
utilizes a time series of multispectral images with different properties which were acquired at
different resolutions by different sensors i.e. Landsat-8 and RapidEye
The fact that most of these satellite imagery systems are commercial represents
one of the obstacles for their widespread use and produces the need for the
effective and efficient use of a vast amount of freely available data, like the highquality data already available through the Landsat program 13 o

13 C. Reynold, Utilizing Landsat sidelaps to monitor global agriculture by a multiplatform system, in Proc. of the Global Geospatial Conf. Africa GIS, Addis Ababa,
Ethiopia, Global Spatial Data Infrastructure (GSDI) Association, Needham,
Massachusetts (2013).

It also explores the capabilities of the proposed data fusion method for their
combination with the aim of overcoming the shortcomings of different instruments
in the particular cropland classification scenario characterized by the specific size of
crop fields over the chosen agricultural region situated in the plains of Vojvodina in
northern Serbia (southeastern Europe). It can be said that this scenario, where
agriculture is based on very small areas dedicated to specific crops, is quite
common in different parts of the world. One of the problems that arises is the
presence of crop fields with very small areas and small dimensions compared to the
spatial resolution of the freely available multispectral imagery. Additionally, if there
is no available information about the crop field boundaries or cadastral data (such
as in this study) which would enable the application of an object-based classification
approach, it is of great significance to develop methods that effectively exploit
available data and provide an improvement over the existing pixel-based
classification approaches through the combination of different data sources.
Therefore, this paper proposes a data fusion method that is successfully utilized in
combination with a robust random forest classifier17 in improving overall
classification performance, as well as in enabling application of satellite imagery
with a coarser spatial resolution in the given specific cropland classification task. A
random forest classifier was chosen due to its recent use in remote sensing and
because it has been accepted as an efficient tool in crop classification.1820