Proceedings of the Workshop on Remote Sensing for Agriculture in the 21st Century

October 23-25th, 1996 
 

Wesley W. Wallender

UC Davis
 
| Ag 21 Agenda |

Remote Sensing Data, Agricultural and Ecological Models, and the Balance Between Profitability and the Environment
 
by Wesley W. Wallender, Professor
wwwallender@ucdavis.edu
Departments of Land, Air and Water Resources and
Biological and Agricultural Engineering
University of California Davis
 
During the past two decades, improved hardware, from Global Positioning Systems (GPS) to high technology sensors, and improved software, from Geographic Information Systems (GIS) to crop simulators, have enhanced understanding and management of natural systems.  Rapid high resolution measurement allows understanding processes better at short time and space scales.  Low cost sensors can be placed at numerous locations to measure the variability within and between fields.  More expensive remote sensors can measure large areas at high resolution to reduce cost per unit area.  Through improved telemetry and networking, information is transferred rapidly over long distances.  With a better understanding of processes and with better measurement in real time, control via Variable Rate Treatment (VRT) technology is superior.  Improved control manifests itself in higher quality products and reduced environmental impact.

Systems
Ideas and mass are transported and transformed across and within system boundaries.  An event at one location influences events elsewhere and an event at one time affects later events.  In the past, an imaginary boundary was drawn around a subsystem or region of interest and only substances of utility were considered part of the circumscribed subsystem; pollutants were blissfully unknown or conveniently ignored.  From a pollution point of view, the world was considered to be a series of disconnected systems, domains or islands of life not influenced by or influencing one another.  Equipped with better measurements and models, the transformation and flow of byproducts as well as products is considered in balancing productivity and pollution.

Control at boundaries is driven by incentives.  Products having value in the market are carefully measured and controlled during production to maximize value to the consumer and profit to the producer.  Conversely, byproducts may be measured and controlled but often to a lesser degree depending on their cost to production.  For example, unless pollution is somehow penalized, there is little interest in controlling and measuring its level within the system or its movement across boundaries.  Byproducts, such as nitrates, herbicides, pesticides in the groundwater, are non-point source and especially difficult to measure and control.  In some cases, the production process is moved away from where the product is consumed and cost is temporarily externalized.  Economic and regulatory incentives and disincentives are used to reduce pollution without jeopardizing profitability.  The task is to manage the system to achieve the economic and environmental goals.  Systems analysis is a logical tool to organize the collection of data and development and implementation of simulation models used to predict the economic and environmental outcomes.

Space and time scales
Agroecosystem inputs and outputs are inherently variable in space. Measurements on the landscape are intended to represent the mean and variability of particular input variables such as a soil texture or crop yield. Variation generally decreases as the sample volume, area or length increases (space scale). Soil variation causes variation in infiltration, soil moisture, biological activity and crop yield to mention a few. If one were to simulated yield as a function of soil properties, the variation in soil would translate into yield variation. At the appropriate space scale, variation in the input variable predicts the variation in measured output. Unfortunately the mean and variation in yield are influenced by other factors such as disease but at a different space scale.

Discussion questions

Agroecosystems not only vary in space but also in time. Time scales vary from seconds to years or even decades and beyond, depending on the phenomenon. Transport of nutrients through the root zone and into the groundwater can occur over a few seconds in cracking soils but clay content changes over centuries. Insect infestation might be tracked daily or weekly and leaf water potential should be sensed hourly or daily depending on intended use.

Discussion questions

System Management
The life cycle of system management (Figure 1) is the measurement of input data, manipulation of input data, simulation and optimization, treatment application, measurement of output data and manipulation of output data. Remote sensing (RS), geographic information systems (GIS), decision support systems (DSS) and variable rate treatment (VRT) technologies underpin system management (Figure 2).

1. Measure input data (RS, GPS)
The cycle begins with measuring the input data using remote sensing (RS) or with ground-based equipment and referencing the data using a global positioning systems (GPS). Examples of data collected for precision farming include:

Physical data: Field boundaries, slope and aspect, particle size distribution, rooting volume, soil moisture, drainage

Chemical data: Cation exchange capacity, nutrient levels, pH, salinity, plant tissue, element levels, leaf water potential

Biological data: Disease distribution, insect distribution, weed distribution, organic matter content

Discussion questions

2. Manipulate input data (GIS)
Multiple data layers are generally required for a simulation. Because data come from a number of sources, the format and coordinate systems, for example, must be reconciled for each location on the landscape. This preprocessing makes the input data available for the simulation model. Common tools for data manipulation and quality control are:

Rectification algorithm to correct geometry of digital images
Classification algorithms
Spatial interpolation algorithms
Time series analysis
Visualization

Discussion questions

3. Simulate and optimize system (DSS)
Models are used to simulation the effect of particular treatments. A harvest plan might be evaluated according to profit and pollution. The optimized plan prescribes the variable rate treatment technology which is guided by a GPS. Examples of modeling approaches include:

Expert systems
Linear and dynamic programming
Deterministic and stochastic modeling
Visualization
Economic analysis

Discussion questions

4. Apply treatments (VRT, GPS)
Variable rate treatment technology delivers the treatment according to the map produced by the decision support system. If variable rate technology is not available, the decision support system calculates optimized uniform treatments. In both cases the DSS constrains input levels according to environmental standards or costs. Variable rate inputs include:

Fertilizer
Insecticide
Nematocide
Herbicide
Fungicide
Crop variety
Crop species
Seed depth
Seed rate
Tillage
Water

Discussion questions

5. Measure response data (RS, GPS)
The system produces marketable physical and biological products as well as pollution which must be carefully monitored. Georeferenced environmental data are measured either directly or indirectly using remote sensing. The following outputs are commonly used to judge the success of a management system:

Biomass or yield quantity and quality
Land quality and quantity
Water quality and quantity
Air quality
Input data (See step 1)

Discussion questions

6. Manipulate response data (GIS)
Measured response data can be compared with itself spatially and temporally but it can also be compared with output from the simulation models mentioned in step 3. If agreement is unsatisfactory, the model can be recalibrated or replaced with a superior model. The tools mentioned in step 1 are useful.
Rectification algorithm to correct geometry of digital image,
Classification algorithms
Spatial interpolation algorithms
Time series analysis
Visualization

Discussion questions