A3.1: Crop Yield

Model Description: 

A Bayesian network model for crop yield is shown below. This is a multi-target model which captures primarily the factors related to the production, the area sown, and the yield of two primary crops (rice and wheat). Dependency relationships between the nodes are established from government reports and published literature.

Resources:  https://drive.google.com/drive/folders/1jdj6F22DoHDeGBJTZh_gVTWs8NAwgSB7?usp=sharing  



The above model is based on the following nodes. The color scheme used represents the type of node/attribute. (Purple – Independent node; Orange – Dependent node; Red – Target node)

Agro-Climatic ZoneClimate suitable for major types of crops KAG 2020-21 
RainfallAnnual rainfall recorded in mm KAG 2020-21 
LandholdingsPossession of land  KAG 2020-21 
Agriculture loan distributedAverage of agriculture loans distributed by public, private and regional banks KAG 2020-21 
KCC distributedKisan credit cards distributed KAG 2020-21 
SHC distributedSoil health cards distributed KAG 2020-21 
NPK distributedFertilizers distributed KAG 2020-21 
Irrigated areaIrrigated Land Area (hectares)KAG 2020-21 
Crop 1 – Area Sown (Rice/Paddy)Area sown for crop 1 (say paddy) hectares KAG 2020-21 
Crop 2- Area Sown (Wheat)Area sown for crop 1 (say paddy) hectares KAG 2020-21 
Crop production 1 (Rice/Paddy)Production (kgs) of crop1KAG 2020-21 
Crop production 2 (Wheat)Production (kgs) of crop2KAG 2020-21 
Crop Yield 1 (Rice/Paddy) Yield of crop 1 in kgs/hectareKAG 2020-21 
Crop Yield 2 (Wheat)Yield of crop 2 in kgs/hectareKAG 2020-21 
Node Description Table
  • Note: The attributes are chosen from KAG 20-21 dataset. The model can be improvised with additional data and facts accordingly.

Attribute Correlation Matrix : 

The correlation Matrix shows that attributes are correlated with many others, and provides a strong statistical basis to support further investigation to unravel potential linear and non-linear relationships

Intervention Modeling :

It sets the network variables to specific levels, to track the changes in the target variable in the network. For instance: If we fix the value of the variable ‘Fertilizer consumption’ to ‘low’, then we would study the impact of this intervention on the target variable i.e., “Crop Yield” of the network across various geo-entities of the state.

Intervention CodeIntervening NodeDescriptionEase of Intervention
[1-most difficult, 5-easy]
Total Models ImpactedModels ImpactedDomains ImpactedData Stories impacted upon intervention
I1NPK distributedIntervening in the distribution of synthetic fertilizers to study the impact on crop yield40001. Modeling The Intervention Of NPK Consumption On Crop Yield
2. Stress Modelling 
I2Agriculture loans distributedIntervening in the distribution of agricultural loans from public, private and regional banks3000
I3Land holdingsIncreasing marginal and small farmer land holdings area 1000
I4Land area sownIncreasing the area of land to sow crops during Rabi and Kharif seasons4000

DISCLAIMER: AI predictive models are meant to support and augment expert decision-making, and not a replacement for the same. It is important for AI model predictions to be vetted by domain experts before committing to action