35. Memory Types

35.1. Introduction to Earth Memory Types

The Earth Memory framework organizes environmental data into specialized memory types, each designed to capture different aspects of Earth’s systems. This guide introduces the various memory types available and explains how to work with them.

35.2. Core Memory Concepts

In the Earth Memory framework, memories are structured representations of environmental data that:

Capture specific phenomena or patterns observed in Earth systems
Include rich metadata about spatial and temporal context
Maintain information about data provenance and quality
Support various resolutions and levels of detail
Can be combined, compared, and analyzed together

35.3. Memory Type Hierarchy

Memory types are organized in a hierarchical structure:

BaseMemory
├── ClimaticMemory
│   ├── TemperatureMemory
│   ├── PrecipitationMemory
│   ├── WindMemory
│   └── AtmosphericCompositionMemory
├── HydrologicMemory
│   ├── SurfaceWaterMemory
│   ├── GroundwaterMemory
│   ├── SoilMoistureMemory
│   └── SnowIceMemory
├── BiologicMemory
│   ├── VegetationMemory
│   ├── BiodiversityMemory
│   ├── EcosystemMemory
│   └── PhenologyMemory
├── GeologicMemory
│   ├── LandformMemory
│   ├── SoilMemory
│   ├── TectonicMemory
│   └── GeochemicalMemory
├── AnthropogenicMemory
│   ├── LandUseMemory
│   ├── InfrastructureMemory
│   ├── PollutionMemory
│   └── ResourceExtractionMemory
└── IntegratedMemory
    ├── LandscapeMemory
    ├── EventMemory
    ├── CarbonCycleMemory
    └── WaterCycleMemory

35.4. Working with Memory Types

35.4.1. Creating Memory Instances

Here’s how to create instances of different memory types:

from memories.earth import MemoryCreator
from memories.earth.types import VegetationMemory, TemperatureMemory

# Create a memory creator
creator = MemoryCreator()

# Create a vegetation memory from NDVI data
vegetation_memory = creator.create_memory(
    memory_type=VegetationMemory,
    name="amazon-vegetation-2023",
    data=ndvi_data,  # NumPy array or xarray.Dataset
    spatial_reference="EPSG:4326",
    time_range=("2023-01-01", "2023-12-31"),
    resolution="30m",
    metadata={
        "sensor": "Sentinel-2",
        "index": "NDVI",
        "processing_level": "L2A",
        "cloud_cover_max": 20
    }
)

# Create a temperature memory
temperature_memory = creator.create_memory(
    memory_type=TemperatureMemory,
    name="global-temperature-2020-2023",
    data=temperature_data,
    spatial_reference="EPSG:4326",
    time_range=("2020-01-01", "2023-12-31"),
    resolution="0.25deg",
    units="celsius",
    metadata={
        "source": "ERA5",
        "variable": "2m_temperature",
        "aggregation": "monthly_mean"
    }
)

35.4.2. Memory Tier Assignment

Memory instances can be assigned to different tiers based on their access patterns and temporal relevance:

# Assign to hot memory tier for active use
vegetation_memory.assign_to_tier("hot")

# Assign to warm memory tier for medium-term storage
older_vegetation_memory.assign_to_tier("warm")

# Assign to cold memory tier for long-term archival
historical_vegetation_memory.assign_to_tier("cold")

# Assign to glacier memory tier for indefinite preservation
ancient_pollen_record.assign_to_tier("glacier")

35.4.3. Memory Operations

Memories support various operations for analysis and transformation:

# Extract a subset of a memory
amazon_subset = vegetation_memory.subset(
    bbox=[-73.5, -9.0, -60.0, 0.0],  # Western Amazon
    time_range=("2023-06-01", "2023-08-31")  # Dry season
)

# Combine two memories
combined_memory = vegetation_memory.combine(temperature_memory)

# Transform a memory
monthly_aggregated = vegetation_memory.transform(
    operation="temporal_aggregation",
    params={"frequency": "1M", "method": "mean"}
)

# Extract statistics
stats = vegetation_memory.statistics()
print(f"Mean NDVI: {stats['mean']}")
print(f"Standard deviation: {stats['std']}")
print(f"Minimum: {stats['min']}, Maximum: {stats['max']}")

# Analyze trends
trend_analysis = vegetation_memory.analyze_trend(
    method="linear_regression",
    significance_level=0.05
)

35.5. Memory Type Descriptions

35.5.1. Climatic Memories

Memory Type	Description
TemperatureMemory	Records air, land, and sea temperature patterns over time
PrecipitationMemory	Captures rainfall, snowfall, and other forms of precipitation
WindMemory	Represents wind speed, direction, and patterns
AtmosphericCompositionMemory	Tracks greenhouse gases, aerosols, and other atmospheric constituents

35.5.2. Hydrologic Memories

Memory Type	Description
SurfaceWaterMemory	Represents lakes, rivers, reservoirs, and surface water dynamics
GroundwaterMemory	Captures aquifer levels, recharge rates, and groundwater quality
SoilMoistureMemory	Tracks water content in soil at different depths
SnowIceMemory	Records snow cover, ice extent, glaciers, and polar ice

35.5.3. Biologic Memories

Memory Type	Description
VegetationMemory	Represents plant cover, health, productivity, and structure
BiodiversityMemory	Captures species distributions, richness, and ecosystem diversity
EcosystemMemory	Tracks ecosystem function, services, and resilience
PhenologyMemory	Represents seasonal biological events and cycles

35.5.4. Geologic Memories

Memory Type	Description
LandformMemory	Represents terrain, topography, and geomorphology
SoilMemory	Captures soil composition, structure, and properties
TectonicMemory	Tracks plate movements, earthquakes, and volcanic activity
GeochemicalMemory	Represents chemical composition of rocks, soils, and sediments

35.5.5. Anthropogenic Memories

Memory Type	Description
LandUseMemory	Represents human land use patterns and changes
InfrastructureMemory	Captures built environment and human infrastructure
PollutionMemory	Tracks pollutants in air, water, soil, and their impacts
ResourceExtractionMemory	Represents mining, drilling, harvesting, and extraction activities

35.5.6. Integrated Memories

Memory Type	Description
LandscapeMemory	Holistic representation of landscapes integrating multiple aspects
EventMemory	Captures discrete environmental events like floods, fires, or storms
CarbonCycleMemory	Integrates carbon fluxes and stocks across Earth systems
WaterCycleMemory	Represents the complete hydrological cycle in a region

35.6. Creating Custom Memory Types

You can create custom memory types to represent specialized environmental data:

from memories.earth.types import BiologicMemory

class CoralReefMemory(BiologicMemory):
    """Memory type specialized for coral reef ecosystems."""

    def __init__(self, name, data, **kwargs):
        super().__init__(name, data, **kwargs)
        self.ecosystem_type = "coral_reef"
        self.required_attributes = [
            "coral_cover", "species_diversity", "health_index"
        ]

    def calc_bleaching_risk(self, temperature_memory):
        """Calculate coral bleaching risk based on temperature anomalies."""
        # Implementation details...
        return bleaching_risk_index

    def identify_resilient_areas(self):
        """Identify areas of the reef showing resilience."""
        # Implementation details...
        return resilient_zones

35.7. Using Memory Types in Analysis

Different memory types can be used together in analysis pipelines:

from memories.earth.analysis import CorrelationAnalysis, RegimeShiftDetection

# Analyze relationship between temperature and vegetation
correlation = CorrelationAnalysis.run(
    memories=[temperature_memory, vegetation_memory],
    method="pearson",
    lag_range=(-3, 3),  # Months
    significance_level=0.05
)

# Detect regime shifts in ecosystem
shifts = RegimeShiftDetection.run(
    memory=ecosystem_memory,
    method="sequential_t_test",
    parameters={
        "cut_off_length": 10,
        "significance_level": 0.01,
        "huber_weight": 1
    }
)

35.8. Advanced Memory Features

35.8.1. Memory Ensemble

Create and work with ensembles of memories:

from memories.earth.ensemble import MemoryEnsemble

# Create an ensemble from multiple vegetation memories
vegetation_ensemble = MemoryEnsemble(
    memories=[vegetation_memory_1, vegetation_memory_2, vegetation_memory_3],
    name="vegetation-ensemble-2023",
    weights=[0.5, 0.3, 0.2]  # Optional weights for each memory
)

# Calculate ensemble statistics
ensemble_mean = vegetation_ensemble.mean()
ensemble_uncertainty = vegetation_ensemble.uncertainty()

# Find optimal subset of memories
optimal_subset = vegetation_ensemble.optimize(
    target_variable="primary_productivity",
    optimization_metric="rmse",
    reference_data=ground_truth_data
)

35.8.2. Memory Fusion

Fuse different memory types to create integrated understanding:

from memories.earth.fusion import MemoryFusion

# Create a fusion of vegetation, temperature, and precipitation memories
drought_memory = MemoryFusion.create(
    memories=[vegetation_memory, temperature_memory, precipitation_memory],
    fusion_type="drought_index",
    parameters={
        "veg_weight": 0.4,
        "temp_weight": 0.3,
        "precip_weight": 0.3,
        "method": "weighted_integration"
    }
)

# Use the fused memory for analysis
drought_severity = drought_memory.extract_index("severity")
drought_duration = drought_memory.extract_index("duration")
drought_impact = drought_memory.analyze_impact(
    target="vegetation_productivity"
)

35.9. Next Steps

After learning about memory types:

Explore memory architecture in Memory Architecture
Learn about memory retrieval and query in Memory Querying
Discover AI integration capabilities in AI Integration with memories-dev