Population growth, inadequate land management, climate change, extreme stochastic events and changes in land use increase rates of soil erosion and degradation and decrease the quantity and quality of water available 4, 5. Such changes generate negative impacts on human health and natural ecosystems, decrease agricultural and industrial production and increase the occurrence of social conflicts 6, 7.

The WHO Global Strategy on Diet, Physical Activity and Health (2004) and the WHO Global Plan of Action for the Prevention and Control of Chronic Non-Communicable Diseases (2013-2020) call on governments to take measures to increase the availability, affordability and consumption of fruits and vegetables in the population. This aims to reduce the impact of non-communicable diseases and protect health. It is stated that there is convincing evidence that fruits and vegetables reduce the risk of obesity and cardiovascular disease and are also likely to reduce the risk of diabetes. A high-level international review of fruit and vegetable consumption and cancer risk coordinated by the International Agency for Research on Cancer (IARC) concluded that consumption of fruits and vegetables may reduce the risk of cancer, and in particular gastrointestinal cancers. The IARC estimates that the preventable fraction of cancers due to insufficient fruit and vegetable intake worldwide ranges from 5% to 12%, and from 20% to 30% for gastrointestinal cancers. additionally the American Heart Association and other international organizations recommend that our diet include 5 or more servings of fruits and vegetables a day 8, 9, 10.

The city green belts, are a critical interface zones between rural and urban, constitute a heritage territory 11, which holds food production systems of proximity, with low environmental footprint associated with transport. GB generate ecosystem services of local scope, acting in many cases as true green lungs in densely populated areas, contribute to the capture of water and water regulation of urban territory, host biodiversity, carbon, local genetic resources, and generate employment for a significant number of people 12.

In the last 40 years, the study of periurban agriculture has gained importance. In 1979, the OECD Committee for Agriculture organized in Paris a discussion among its Member States on “Agriculture in the planning and management of periurban areas”. From there, the strategic, urban and regional planning is recovered, thinking of these spaces in conflict with the threat of the disappearance of agricultural lands and open spaces 13.

As a near example, in Co´rdoba city there are reported between 1,780 ha 14 and 3,167 ha 15 dedicated to the cultivation of vegetables (Figure 1) 14 due to an overreaching advance of the urban frontier, which grows on strategic natural and agricultural systems for the sustainability of the cities.

In general there are not enough data about the food supply that periurban productive area contributes to the cities, relating this lack of information with deficient land planning, which does not take into account local food security.

Productive ecosystems (i.e. agro-systems) suffer a severe degradation of its natural resources and biodiversity, mainly due to the high degree of brittleness and difficult reversibility of some biological and socioeconomic processes affected, that risk the food supply and food security. Little has been done to enhance the adaptability of industrial agriculture to changing and extreme climatic events 16.

Generally, periurban agrosystems are also at this crossroads, where the short-term economic benefits are frequently linked to the deterioration of productive resources in the medium and long term 17. In a large numbers of large cities of developing countries, the GB still preserves, in general, the outdoor fruit-horticultural production strategy, which allows the use of remote sensing of its production and for the analysis of its ecosystemic relationships at a landscape level.

Along with the food supply, other emerging properties of agrosystems form services for society, such as biological regulation of pests, the regulation of water flows and moderation of extreme phenomena 18. These services interact with each, other and between different production systems. The crops under cover are the ones that use the greatest variety of agrochemicals in their cycle, and that in the bulk of crops, the greatest variety of active ingredients used are linked to the control of animal pests (insects, mites and / or nematodes) 19. Al this make this moment significant to generate information regarding the provisioning and regulation services generated by the GBs, for the purpose of their ordering and optimal planning.

Biological pest regulation is a way to reduce pesticide use and enhance biodiversity 20, 21, 22 guarantees the crop production and food supply. The regulation of water flows, it is important to analyze and assess, given that agriculture is a large consumer of water and has an important influence on it. Agrosystem management can contribute to floods and be considered a regional scope mechanism to control water flows 23. In recent years, the increasing variability of the climate has caused even more severe and frequent drought episodes, which influences agricultural systems in all phases of growth and, consequently, in crop yield 24 so is important the moderation of these extreme phenomena.

Since the 1980s, a promising technology such as remote sensing together with GIS and information technologies, constitute a set of convenient tools for providing large databases on areas of interest. The synthetic indicators allow multi-scale approach for the analysis, from field and landscape, to satellite level.

Here the way that the advances are presented is based in their differences on the scale of the analysis, and their objectives.

At a landscape level, for crop detection and mapping: The US Department of Agriculture (USDA), in 2011, produced a Cropland Data Layer (CDL) product, which is a raster-formatted, geo-referenced, crop-specific, land cover map 25. Their inputs include medium resolution satellite imagery (AWiFS, Landsat TM and ETM, MODIS satellite data), and ancillary data sources. A decision tree-supervised classification method was used. There is an unique class that is vegetables/fruit/nuts.

In 2014 26, proposed a pixel-based cropland classification of small agricultural fields using combined Landsat-8 and RapidEye imagery, for northern Serbia. It is based on a time series of multispectral images acquired at different resolutions. Despite the crops maped are non horticultural, the size of the fields in Vojvodina are like horticultural ones, so the methodology it was and improvement. In spite of not having horticultural species, this work stands out for the scale of identified crop lots.

Actually, the framework of SENSAGRI project is mapping 122 classes land cover based on fusion of Sentinel-2, Landsat 8 and Deimos-1 images and also ancillary data. It used C5.0 algorithm, that generate decision trees from a set of reference data 27.

For crop inventories and production forecasts, the studies are mainly based on high resolution or hyperspectral remote sensors and time series analysis. In particular Hyperspectral sensors allow us to measure the energy reflected from the sun in numerous channels (typically more than 200). In this way it is possible to construct with great precision the spectral signature of the targets of interest and then to characterize them radiometrically. As an example of this it is shown in the Figure 2 spectra signatures of horticultural species measures on field obtained with radiometers, where it is possible to be observed clearly the possibility of differentiating them through these tools. Unfortunately today there are still no operational satellite platforms with hyperspectral sensors, so these tools are not yet available for operational use in the planning and operational management of horticultural crops.

In 2004 28 used the Indian Remote Sensing Satellite (IRS)-ID Linear Imaging Self-Scanning (LISS)-III sensor to estimate the production of fruit and vegetable crops grown. They applied an Supervised Maximum Likelihood Classifier (MLC), and visual interpretation of the texture from PAN sensor.

For hyperspectral, 29, proved an portable hyperspectral camera, and an object oriented software framework with models, that identified crop, soil and weed; they case of study was sugar bet and green citrus. Site-specific crop management (SSCM) is one facet of precision agriculture, widely applied on row crop agriculture, but it has very little application in fruit and nut. For that 30 reviewed those that used high resolution satellites, hyperspectral, LIDARS, UAVs, etc; and GIS spatial modeling for fruit orchards.

Other objective at landscape level is drought and flood damage assessment, 52, mentions it as a potential use, given that it has been employed to describe climate change conditions over large areas and horticultural food provision areas could be considered in the future.

In coordination with land monitoring and management and also territorial planing of food provision, 15, 14, studies the territorial dynamic and distribution of resources in periurban horticultural production area, with SPOT satellite images. In Kenia, 33, studies the dynamics and actual conflicts between to sizes and socio-economic scales of horticultural production ventures.

Other scale of approach of remote sensing in horticulture is at a yield level. For crop monitoring production, as well as the advantages made in 43, 42, 29 and 30 an other work that shows the yield monitoring objective is 34 for vineyards canopy management, the GRS is useful to detect canopy, and this is elemental to improve production and/or wine quality, reduce disease incidence, and facilitate mechanization. With RGB and stereoscopic images, segmentation and plant quality models, 53 proposed and improvement for the assessment in complex decisions, the scheduling and quality management of a growing plant crop, especially in protected crop environments in glasshouses.

About the irrigation management, 44, working with Landsat 5 TM, found evapotranspiration (ET) - NDVI consistent relations in grape, almond and citrus crops, demonstrates the value of using GRS as a practical tool for improved irrigation systems and irrigation water management. A review for this task is 54. Near-range and remote sensing techniques have demonstrated a high potential in detecting diseases and in monitoring crop stands for sub-areas with infected plants 55. Most promising sensor types are thermography, chlorophyll fluorescence and hyperspectral sensors 56, 57, 52.

So we can summarize that application following their scale and objectives like:

At Landscape Level

1. Crop detection and mapping

2. Crop inventories and production forecasts

3. Drought and flood damage assessment

4. Land monitoring and management

At Yield Level

1. Crop Monitoring (production)

2. Irrigation Management

3. Growth and health of plants (disease, stress and pests)

All those advances are briefly reviewed on Figure 3 and a nice chart on Fruit and nut crop classification using remote sensing from 52.

The study of the remote sensing in horticultural activity is challenging to researchers given its particular characteristics. Context and fragmentation. Added to that, the size (Figure 4) and the diversity of crops (Figure 5) that make the orchards up, those interfere on the spectral reception of the sensor and make even more important that it have high spatial, spectral and radiometric resolution 52.

The challenges facing by the horticultural RS technology can basically be summed up in three questions: Where?, What? and How much?. By WHERE we mean to try to map the area under horticultural cultivation in a certain region, for example the Green Belt of a big cities. Since the horticultural crop is basically a land use and not a land cover, it does not have a unique radiometric response, and then its classification is not a simple issue. Increasing the level of complexity we are interested in knowing, WHAT i.e., differentiate the type of horticultural crop. Basically it mean to know if it is lettuce, cabbage, beet, or others. These different species of vegetables can have different spectral signatures, but the spatial scale of each crop is too small for the multispectral or hyperspectral remote sensing and the necessary spatial resolution is until now not easily accessible. Finally and even more complex is to try to quantify the production of each crop. That is to say HOW MUCH or how many kg per square meter of a certain vegetable is available to be harvested.

Other characteristic important for GRS in horticulture are the difference in harvest dates, derived from the crops diversity and market strategies. The growth stages and phenology of many horticultural crops are not well understood and tend to be difficult to generalize as a result of wide variations in varieties, planting densities, and cultural practices 52. Despite that, the RS studies seems a useful tool for identifying optimal harvest dates 28.

However, the reliance of optical systems on the Sun’s energy limits image acquisition in cloudy or hazy conditions (Figure 5). Images acquired during these periods are normally of little use in mapping due to high cloud/haze cover.

Nevertheless 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 Earth’s surface 58. Data fusion between SAR, optical high resolution systems and ancillary data is explore to overcome this problem 52, 32.

Another issue is the lack of field survey for horticulture planing and territorial management, the multi-date nature of satellite imagery permits monitoring dynamic features of landscape and thus provides a means to detect major land cover changes and quantify the rates of change 52. Prior knowledge and ancillary information obtained by RS and non-destructive soil and plant sensors can be used to design protocols that provide an appropriate balance between statistical precision and cost 59.

More advanced and accessible tools from a computational perspective open also more opportunities to use the new remote sensing data for operative applications on horticulture. In particular methodologies such as object based image analysis (OBIA), and machine and deep learning usually used in the field of the computer vision are closely related to the access to very high spatial resolution images (VHRI) (2cm of pixel size) for example from Drone. As an example, in the Figure 6 we present an image of a drone on a broccoli orchard. This image have spatial resolution of 2.5cm where each plant can be seen individually and so tools such us convolutionals neural networks used in computer vision, and easily accessible from, for example, python image processing libs, could be applied here for an exhaustive count of this crop.