rt. 
Je 
S 
1S 
via 
trail, 
ma ted 
se 
om 
eds, 
le 
id 
B 
ty 
comes 
| of 
jSSary 
lote 
Ing 
    
   
   
  
  
  
  
   
   
   
  
  
  
  
  
    
  
   
  
   
     
  
   
  
   
   
  
  
  
  
  
  
  
  
  
    
  
  
   
  
   
  
  
which can support the full range of stated requirements. The key com- 
ponent of an automated data base is its stability and usefulness over 
time. The APU and its data elements. provide this stability. The APU 
design rationale is directly transferable to a data base design con- 
cept. This rationale is to identify and utilize those independent 
variables within the agriculture environment upon which remote sens- 
ing technology and data analysis can be baselined. There are three 
such variables: 1) soils, 2) climatic regions, and 3) historical 
events/data. 
Within this information requirements environment, efficient automated 
support to data storage and retrieval becomes a mandatory element 
for exploitation of remotely sensed data. Given this statement of 
fact, and a baseline of independent variables (APU's), the USDA 
developed the functional and operational design concept for an 
application test data base. 
B. Design 
1. Functional Design Concept 
Following the cost and utility requirements defined in the preced- 
ing section, the functional design architecture of an APU- 
oriented data base is summarized. 
a. Logical Structure. The entire agriculture and potential 
agricultural universe is delineated by a fixed grid cell with 
a unique latitude and longitude address (x and Y). The 
agricultural universe is defined by APU's. This master data 
set contains soils, climatic, geological, and administrative 
boundaries, historical agricultural statistics, and current 
meteorological data. 
Linked to this master data set via stored indices are 1 through 
N commodity/country data sets. These secondary sets contain 
the following generic elements: 1) Statistical sample alloca- 
tion by commodity/region, 2) commodity kind, 3) multispectral 
data used in commodity analysis, 4) yield model forms by 
crop/region, 5) statistical aggregation and/or probability 
parameters by crop and region, 6) multispectral coverage 
requirements by crop/region/year, and 7) crop calendars by 
region. 
It has been argued that this design concept is extremely 
costly, unnecessarily complex, and defies efficient reorganiza- 
tion. This is only true if one pursues the logic that all 
data sets are maintained on-line until needed in the analysis 
process. The physical data structure concept of our dis- 
tributed network also embodies a physical structure concept 
of multiple storage levels activated upon demand and deacti- 
-vated upon completion of specified tasks.